linux/drivers/iommu/exynos-iommu.c

1531 lines
42 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
/*
* Copyright (c) 2011,2016 Samsung Electronics Co., Ltd.
* http://www.samsung.com
*/
#ifdef CONFIG_EXYNOS_IOMMU_DEBUG
#define DEBUG
#endif
#include <linux/clk.h>
#include <linux/dma-mapping.h>
#include <linux/err.h>
#include <linux/io.h>
#include <linux/iommu.h>
#include <linux/interrupt.h>
headers: untangle kmemleak.h from mm.h Currently <linux/slab.h> #includes <linux/kmemleak.h> for no obvious reason. It looks like it's only a convenience, so remove kmemleak.h from slab.h and add <linux/kmemleak.h> to any users of kmemleak_* that don't already #include it. Also remove <linux/kmemleak.h> from source files that do not use it. This is tested on i386 allmodconfig and x86_64 allmodconfig. It would be good to run it through the 0day bot for other $ARCHes. I have neither the horsepower nor the storage space for the other $ARCHes. Update: This patch has been extensively build-tested by both the 0day bot & kisskb/ozlabs build farms. Both of them reported 2 build failures for which patches are included here (in v2). [ slab.h is the second most used header file after module.h; kernel.h is right there with slab.h. There could be some minor error in the counting due to some #includes having comments after them and I didn't combine all of those. ] [akpm@linux-foundation.org: security/keys/big_key.c needs vmalloc.h, per sfr] Link: http://lkml.kernel.org/r/e4309f98-3749-93e1-4bb7-d9501a39d015@infradead.org Link: http://kisskb.ellerman.id.au/kisskb/head/13396/ Signed-off-by: Randy Dunlap <rdunlap@infradead.org> Reviewed-by: Ingo Molnar <mingo@kernel.org> Reported-by: Michael Ellerman <mpe@ellerman.id.au> [2 build failures] Reported-by: Fengguang Wu <fengguang.wu@intel.com> [2 build failures] Reviewed-by: Andrew Morton <akpm@linux-foundation.org> Cc: Wei Yongjun <weiyongjun1@huawei.com> Cc: Luis R. Rodriguez <mcgrof@kernel.org> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Mimi Zohar <zohar@linux.vnet.ibm.com> Cc: John Johansen <john.johansen@canonical.com> Cc: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:25:34 +00:00
#include <linux/kmemleak.h>
#include <linux/list.h>
#include <linux/of.h>
#include <linux/of_platform.h>
#include <linux/platform_device.h>
#include <linux/pm_runtime.h>
#include <linux/slab.h>
#include "iommu-pages.h"
typedef u32 sysmmu_iova_t;
typedef u32 sysmmu_pte_t;
static struct iommu_domain exynos_identity_domain;
/* We do not consider super section mapping (16MB) */
#define SECT_ORDER 20
#define LPAGE_ORDER 16
#define SPAGE_ORDER 12
#define SECT_SIZE (1 << SECT_ORDER)
#define LPAGE_SIZE (1 << LPAGE_ORDER)
#define SPAGE_SIZE (1 << SPAGE_ORDER)
#define SECT_MASK (~(SECT_SIZE - 1))
#define LPAGE_MASK (~(LPAGE_SIZE - 1))
#define SPAGE_MASK (~(SPAGE_SIZE - 1))
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
#define lv1ent_fault(sent) ((*(sent) == ZERO_LV2LINK) || \
((*(sent) & 3) == 0) || ((*(sent) & 3) == 3))
#define lv1ent_zero(sent) (*(sent) == ZERO_LV2LINK)
#define lv1ent_page_zero(sent) ((*(sent) & 3) == 1)
#define lv1ent_page(sent) ((*(sent) != ZERO_LV2LINK) && \
((*(sent) & 3) == 1))
#define lv1ent_section(sent) ((*(sent) & 3) == 2)
#define lv2ent_fault(pent) ((*(pent) & 3) == 0)
#define lv2ent_small(pent) ((*(pent) & 2) == 2)
#define lv2ent_large(pent) ((*(pent) & 3) == 1)
/*
* v1.x - v3.x SYSMMU supports 32bit physical and 32bit virtual address spaces
* v5.0 introduced support for 36bit physical address space by shifting
* all page entry values by 4 bits.
* All SYSMMU controllers in the system support the address spaces of the same
* size, so PG_ENT_SHIFT can be initialized on first SYSMMU probe to proper
* value (0 or 4).
*/
static short PG_ENT_SHIFT = -1;
#define SYSMMU_PG_ENT_SHIFT 0
#define SYSMMU_V5_PG_ENT_SHIFT 4
static const sysmmu_pte_t *LV1_PROT;
static const sysmmu_pte_t SYSMMU_LV1_PROT[] = {
((0 << 15) | (0 << 10)), /* no access */
((1 << 15) | (1 << 10)), /* IOMMU_READ only */
((0 << 15) | (1 << 10)), /* IOMMU_WRITE not supported, use read/write */
((0 << 15) | (1 << 10)), /* IOMMU_READ | IOMMU_WRITE */
};
static const sysmmu_pte_t SYSMMU_V5_LV1_PROT[] = {
(0 << 4), /* no access */
(1 << 4), /* IOMMU_READ only */
(2 << 4), /* IOMMU_WRITE only */
(3 << 4), /* IOMMU_READ | IOMMU_WRITE */
};
static const sysmmu_pte_t *LV2_PROT;
static const sysmmu_pte_t SYSMMU_LV2_PROT[] = {
((0 << 9) | (0 << 4)), /* no access */
((1 << 9) | (1 << 4)), /* IOMMU_READ only */
((0 << 9) | (1 << 4)), /* IOMMU_WRITE not supported, use read/write */
((0 << 9) | (1 << 4)), /* IOMMU_READ | IOMMU_WRITE */
};
static const sysmmu_pte_t SYSMMU_V5_LV2_PROT[] = {
(0 << 2), /* no access */
(1 << 2), /* IOMMU_READ only */
(2 << 2), /* IOMMU_WRITE only */
(3 << 2), /* IOMMU_READ | IOMMU_WRITE */
};
#define SYSMMU_SUPPORTED_PROT_BITS (IOMMU_READ | IOMMU_WRITE)
#define sect_to_phys(ent) (((phys_addr_t) ent) << PG_ENT_SHIFT)
#define section_phys(sent) (sect_to_phys(*(sent)) & SECT_MASK)
#define section_offs(iova) (iova & (SECT_SIZE - 1))
#define lpage_phys(pent) (sect_to_phys(*(pent)) & LPAGE_MASK)
#define lpage_offs(iova) (iova & (LPAGE_SIZE - 1))
#define spage_phys(pent) (sect_to_phys(*(pent)) & SPAGE_MASK)
#define spage_offs(iova) (iova & (SPAGE_SIZE - 1))
#define NUM_LV1ENTRIES 4096
#define NUM_LV2ENTRIES (SECT_SIZE / SPAGE_SIZE)
static u32 lv1ent_offset(sysmmu_iova_t iova)
{
return iova >> SECT_ORDER;
}
static u32 lv2ent_offset(sysmmu_iova_t iova)
{
return (iova >> SPAGE_ORDER) & (NUM_LV2ENTRIES - 1);
}
#define LV1TABLE_SIZE (NUM_LV1ENTRIES * sizeof(sysmmu_pte_t))
#define LV2TABLE_SIZE (NUM_LV2ENTRIES * sizeof(sysmmu_pte_t))
#define SPAGES_PER_LPAGE (LPAGE_SIZE / SPAGE_SIZE)
#define lv2table_base(sent) (sect_to_phys(*(sent) & 0xFFFFFFC0))
#define mk_lv1ent_sect(pa, prot) ((pa >> PG_ENT_SHIFT) | LV1_PROT[prot] | 2)
#define mk_lv1ent_page(pa) ((pa >> PG_ENT_SHIFT) | 1)
#define mk_lv2ent_lpage(pa, prot) ((pa >> PG_ENT_SHIFT) | LV2_PROT[prot] | 1)
#define mk_lv2ent_spage(pa, prot) ((pa >> PG_ENT_SHIFT) | LV2_PROT[prot] | 2)
#define CTRL_ENABLE 0x5
#define CTRL_BLOCK 0x7
#define CTRL_DISABLE 0x0
#define CFG_LRU 0x1
#define CFG_EAP (1 << 2)
#define CFG_QOS(n) ((n & 0xF) << 7)
#define CFG_ACGEN (1 << 24) /* System MMU 3.3 only */
#define CFG_SYSSEL (1 << 22) /* System MMU 3.2 only */
#define CFG_FLPDCACHE (1 << 20) /* System MMU 3.2+ only */
#define CTRL_VM_ENABLE BIT(0)
#define CTRL_VM_FAULT_MODE_STALL BIT(3)
#define CAPA0_CAPA1_EXIST BIT(11)
#define CAPA1_VCR_ENABLED BIT(14)
/* common registers */
#define REG_MMU_CTRL 0x000
#define REG_MMU_CFG 0x004
#define REG_MMU_STATUS 0x008
#define REG_MMU_VERSION 0x034
#define MMU_MAJ_VER(val) ((val) >> 7)
#define MMU_MIN_VER(val) ((val) & 0x7F)
#define MMU_RAW_VER(reg) (((reg) >> 21) & ((1 << 11) - 1)) /* 11 bits */
#define MAKE_MMU_VER(maj, min) ((((maj) & 0xF) << 7) | ((min) & 0x7F))
/* v1.x - v3.x registers */
#define REG_PAGE_FAULT_ADDR 0x024
#define REG_AW_FAULT_ADDR 0x028
#define REG_AR_FAULT_ADDR 0x02C
#define REG_DEFAULT_SLAVE_ADDR 0x030
/* v5.x registers */
#define REG_V5_FAULT_AR_VA 0x070
#define REG_V5_FAULT_AW_VA 0x080
/* v7.x registers */
#define REG_V7_CAPA0 0x870
#define REG_V7_CAPA1 0x874
#define REG_V7_CTRL_VM 0x8000
#define has_sysmmu(dev) (dev_iommu_priv_get(dev) != NULL)
static struct device *dma_dev;
static struct kmem_cache *lv2table_kmem_cache;
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
static sysmmu_pte_t *zero_lv2_table;
#define ZERO_LV2LINK mk_lv1ent_page(virt_to_phys(zero_lv2_table))
static sysmmu_pte_t *section_entry(sysmmu_pte_t *pgtable, sysmmu_iova_t iova)
{
return pgtable + lv1ent_offset(iova);
}
static sysmmu_pte_t *page_entry(sysmmu_pte_t *sent, sysmmu_iova_t iova)
{
return (sysmmu_pte_t *)phys_to_virt(
lv2table_base(sent)) + lv2ent_offset(iova);
}
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
struct sysmmu_fault {
sysmmu_iova_t addr; /* IOVA address that caused fault */
const char *name; /* human readable fault name */
unsigned int type; /* fault type for report_iommu_fault() */
};
struct sysmmu_v1_fault_info {
unsigned short addr_reg; /* register to read IOVA fault address */
const char *name; /* human readable fault name */
unsigned int type; /* fault type for report_iommu_fault */
};
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
static const struct sysmmu_v1_fault_info sysmmu_v1_faults[] = {
{ REG_PAGE_FAULT_ADDR, "PAGE", IOMMU_FAULT_READ },
{ REG_AR_FAULT_ADDR, "MULTI-HIT", IOMMU_FAULT_READ },
{ REG_AW_FAULT_ADDR, "MULTI-HIT", IOMMU_FAULT_WRITE },
{ REG_DEFAULT_SLAVE_ADDR, "BUS ERROR", IOMMU_FAULT_READ },
{ REG_AR_FAULT_ADDR, "SECURITY PROTECTION", IOMMU_FAULT_READ },
{ REG_AR_FAULT_ADDR, "ACCESS PROTECTION", IOMMU_FAULT_READ },
{ REG_AW_FAULT_ADDR, "SECURITY PROTECTION", IOMMU_FAULT_WRITE },
{ REG_AW_FAULT_ADDR, "ACCESS PROTECTION", IOMMU_FAULT_WRITE },
};
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
/* SysMMU v5 has the same faults for AR (0..4 bits) and AW (16..20 bits) */
static const char * const sysmmu_v5_fault_names[] = {
"PTW",
"PAGE",
"MULTI-HIT",
"ACCESS PROTECTION",
"SECURITY PROTECTION"
};
static const char * const sysmmu_v7_fault_names[] = {
"PTW",
"PAGE",
"ACCESS PROTECTION",
"RESERVED"
};
/*
* This structure is attached to dev->iommu->priv of the master device
* on device add, contains a list of SYSMMU controllers defined by device tree,
* which are bound to given master device. It is usually referenced by 'owner'
* pointer.
*/
struct exynos_iommu_owner {
struct list_head controllers; /* list of sysmmu_drvdata.owner_node */
struct iommu_domain *domain; /* domain this device is attached */
struct mutex rpm_lock; /* for runtime pm of all sysmmus */
};
/*
* This structure exynos specific generalization of struct iommu_domain.
* It contains list of SYSMMU controllers from all master devices, which has
* been attached to this domain and page tables of IO address space defined by
* it. It is usually referenced by 'domain' pointer.
*/
struct exynos_iommu_domain {
struct list_head clients; /* list of sysmmu_drvdata.domain_node */
sysmmu_pte_t *pgtable; /* lv1 page table, 16KB */
short *lv2entcnt; /* free lv2 entry counter for each section */
spinlock_t lock; /* lock for modyfying list of clients */
spinlock_t pgtablelock; /* lock for modifying page table @ pgtable */
struct iommu_domain domain; /* generic domain data structure */
};
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
struct sysmmu_drvdata;
/*
* SysMMU version specific data. Contains offsets for the registers which can
* be found in different SysMMU variants, but have different offset values.
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
* Also contains version specific callbacks to abstract the hardware.
*/
struct sysmmu_variant {
u32 pt_base; /* page table base address (physical) */
u32 flush_all; /* invalidate all TLB entries */
u32 flush_entry; /* invalidate specific TLB entry */
u32 flush_range; /* invalidate TLB entries in specified range */
u32 flush_start; /* start address of range invalidation */
u32 flush_end; /* end address of range invalidation */
u32 int_status; /* interrupt status information */
u32 int_clear; /* clear the interrupt */
u32 fault_va; /* IOVA address that caused fault */
u32 fault_info; /* fault transaction info */
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
int (*get_fault_info)(struct sysmmu_drvdata *data, unsigned int itype,
struct sysmmu_fault *fault);
};
/*
* This structure hold all data of a single SYSMMU controller, this includes
* hw resources like registers and clocks, pointers and list nodes to connect
* it to all other structures, internal state and parameters read from device
* tree. It is usually referenced by 'data' pointer.
*/
struct sysmmu_drvdata {
struct device *sysmmu; /* SYSMMU controller device */
struct device *master; /* master device (owner) */
struct device_link *link; /* runtime PM link to master */
void __iomem *sfrbase; /* our registers */
struct clk *clk; /* SYSMMU's clock */
struct clk *aclk; /* SYSMMU's aclk clock */
struct clk *pclk; /* SYSMMU's pclk clock */
struct clk *clk_master; /* master's device clock */
spinlock_t lock; /* lock for modyfying state */
bool active; /* current status */
struct exynos_iommu_domain *domain; /* domain we belong to */
struct list_head domain_node; /* node for domain clients list */
struct list_head owner_node; /* node for owner controllers list */
phys_addr_t pgtable; /* assigned page table structure */
unsigned int version; /* our version */
struct iommu_device iommu; /* IOMMU core handle */
const struct sysmmu_variant *variant; /* version specific data */
/* v7 fields */
bool has_vcr; /* virtual machine control register */
};
#define SYSMMU_REG(data, reg) ((data)->sfrbase + (data)->variant->reg)
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
static int exynos_sysmmu_v1_get_fault_info(struct sysmmu_drvdata *data,
unsigned int itype,
struct sysmmu_fault *fault)
{
const struct sysmmu_v1_fault_info *finfo;
if (itype >= ARRAY_SIZE(sysmmu_v1_faults))
return -ENXIO;
finfo = &sysmmu_v1_faults[itype];
fault->addr = readl(data->sfrbase + finfo->addr_reg);
fault->name = finfo->name;
fault->type = finfo->type;
return 0;
}
static int exynos_sysmmu_v5_get_fault_info(struct sysmmu_drvdata *data,
unsigned int itype,
struct sysmmu_fault *fault)
{
unsigned int addr_reg;
if (itype < ARRAY_SIZE(sysmmu_v5_fault_names)) {
fault->type = IOMMU_FAULT_READ;
addr_reg = REG_V5_FAULT_AR_VA;
} else if (itype >= 16 && itype <= 20) {
fault->type = IOMMU_FAULT_WRITE;
addr_reg = REG_V5_FAULT_AW_VA;
itype -= 16;
} else {
return -ENXIO;
}
fault->name = sysmmu_v5_fault_names[itype];
fault->addr = readl(data->sfrbase + addr_reg);
return 0;
}
static int exynos_sysmmu_v7_get_fault_info(struct sysmmu_drvdata *data,
unsigned int itype,
struct sysmmu_fault *fault)
{
u32 info = readl(SYSMMU_REG(data, fault_info));
fault->addr = readl(SYSMMU_REG(data, fault_va));
fault->name = sysmmu_v7_fault_names[itype % 4];
fault->type = (info & BIT(20)) ? IOMMU_FAULT_WRITE : IOMMU_FAULT_READ;
return 0;
}
/* SysMMU v1..v3 */
static const struct sysmmu_variant sysmmu_v1_variant = {
.flush_all = 0x0c,
.flush_entry = 0x10,
.pt_base = 0x14,
.int_status = 0x18,
.int_clear = 0x1c,
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
.get_fault_info = exynos_sysmmu_v1_get_fault_info,
};
/* SysMMU v5 */
static const struct sysmmu_variant sysmmu_v5_variant = {
.pt_base = 0x0c,
.flush_all = 0x10,
.flush_entry = 0x14,
.flush_range = 0x18,
.flush_start = 0x20,
.flush_end = 0x24,
.int_status = 0x60,
.int_clear = 0x64,
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
.get_fault_info = exynos_sysmmu_v5_get_fault_info,
};
/* SysMMU v7: non-VM capable register layout */
static const struct sysmmu_variant sysmmu_v7_variant = {
.pt_base = 0x0c,
.flush_all = 0x10,
.flush_entry = 0x14,
.flush_range = 0x18,
.flush_start = 0x20,
.flush_end = 0x24,
.int_status = 0x60,
.int_clear = 0x64,
.fault_va = 0x70,
.fault_info = 0x78,
.get_fault_info = exynos_sysmmu_v7_get_fault_info,
};
/* SysMMU v7: VM capable register layout */
static const struct sysmmu_variant sysmmu_v7_vm_variant = {
.pt_base = 0x800c,
.flush_all = 0x8010,
.flush_entry = 0x8014,
.flush_range = 0x8018,
.flush_start = 0x8020,
.flush_end = 0x8024,
.int_status = 0x60,
.int_clear = 0x64,
.fault_va = 0x1000,
.fault_info = 0x1004,
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
.get_fault_info = exynos_sysmmu_v7_get_fault_info,
};
static struct exynos_iommu_domain *to_exynos_domain(struct iommu_domain *dom)
{
return container_of(dom, struct exynos_iommu_domain, domain);
}
static void sysmmu_unblock(struct sysmmu_drvdata *data)
{
writel(CTRL_ENABLE, data->sfrbase + REG_MMU_CTRL);
}
static bool sysmmu_block(struct sysmmu_drvdata *data)
{
int i = 120;
writel(CTRL_BLOCK, data->sfrbase + REG_MMU_CTRL);
while ((i > 0) && !(readl(data->sfrbase + REG_MMU_STATUS) & 1))
--i;
if (!(readl(data->sfrbase + REG_MMU_STATUS) & 1)) {
sysmmu_unblock(data);
return false;
}
return true;
}
static void __sysmmu_tlb_invalidate(struct sysmmu_drvdata *data)
{
writel(0x1, SYSMMU_REG(data, flush_all));
}
static void __sysmmu_tlb_invalidate_entry(struct sysmmu_drvdata *data,
sysmmu_iova_t iova, unsigned int num_inv)
{
unsigned int i;
if (MMU_MAJ_VER(data->version) < 5 || num_inv == 1) {
for (i = 0; i < num_inv; i++) {
writel((iova & SPAGE_MASK) | 1,
SYSMMU_REG(data, flush_entry));
iova += SPAGE_SIZE;
}
} else {
writel(iova & SPAGE_MASK, SYSMMU_REG(data, flush_start));
writel((iova & SPAGE_MASK) + (num_inv - 1) * SPAGE_SIZE,
SYSMMU_REG(data, flush_end));
writel(0x1, SYSMMU_REG(data, flush_range));
}
}
static void __sysmmu_set_ptbase(struct sysmmu_drvdata *data, phys_addr_t pgd)
{
u32 pt_base;
if (MMU_MAJ_VER(data->version) < 5)
pt_base = pgd;
else
pt_base = pgd >> SPAGE_ORDER;
writel(pt_base, SYSMMU_REG(data, pt_base));
__sysmmu_tlb_invalidate(data);
}
static void __sysmmu_enable_clocks(struct sysmmu_drvdata *data)
{
BUG_ON(clk_prepare_enable(data->clk_master));
BUG_ON(clk_prepare_enable(data->clk));
BUG_ON(clk_prepare_enable(data->pclk));
BUG_ON(clk_prepare_enable(data->aclk));
}
static void __sysmmu_disable_clocks(struct sysmmu_drvdata *data)
{
clk_disable_unprepare(data->aclk);
clk_disable_unprepare(data->pclk);
clk_disable_unprepare(data->clk);
clk_disable_unprepare(data->clk_master);
}
static bool __sysmmu_has_capa1(struct sysmmu_drvdata *data)
{
u32 capa0 = readl(data->sfrbase + REG_V7_CAPA0);
return capa0 & CAPA0_CAPA1_EXIST;
}
static void __sysmmu_get_vcr(struct sysmmu_drvdata *data)
{
u32 capa1 = readl(data->sfrbase + REG_V7_CAPA1);
data->has_vcr = capa1 & CAPA1_VCR_ENABLED;
}
static void __sysmmu_get_version(struct sysmmu_drvdata *data)
{
u32 ver;
__sysmmu_enable_clocks(data);
ver = readl(data->sfrbase + REG_MMU_VERSION);
/* controllers on some SoCs don't report proper version */
if (ver == 0x80000001u)
data->version = MAKE_MMU_VER(1, 0);
else
data->version = MMU_RAW_VER(ver);
dev_dbg(data->sysmmu, "hardware version: %d.%d\n",
MMU_MAJ_VER(data->version), MMU_MIN_VER(data->version));
if (MMU_MAJ_VER(data->version) < 5) {
data->variant = &sysmmu_v1_variant;
} else if (MMU_MAJ_VER(data->version) < 7) {
data->variant = &sysmmu_v5_variant;
} else {
if (__sysmmu_has_capa1(data))
__sysmmu_get_vcr(data);
if (data->has_vcr)
data->variant = &sysmmu_v7_vm_variant;
else
data->variant = &sysmmu_v7_variant;
}
__sysmmu_disable_clocks(data);
}
static void show_fault_information(struct sysmmu_drvdata *data,
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
const struct sysmmu_fault *fault)
{
sysmmu_pte_t *ent;
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
dev_err(data->sysmmu, "%s: [%s] %s FAULT occurred at %#x\n",
dev_name(data->master),
fault->type == IOMMU_FAULT_READ ? "READ" : "WRITE",
fault->name, fault->addr);
dev_dbg(data->sysmmu, "Page table base: %pa\n", &data->pgtable);
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
ent = section_entry(phys_to_virt(data->pgtable), fault->addr);
dev_dbg(data->sysmmu, "\tLv1 entry: %#x\n", *ent);
if (lv1ent_page(ent)) {
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
ent = page_entry(ent, fault->addr);
dev_dbg(data->sysmmu, "\t Lv2 entry: %#x\n", *ent);
}
}
static irqreturn_t exynos_sysmmu_irq(int irq, void *dev_id)
{
struct sysmmu_drvdata *data = dev_id;
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
unsigned int itype;
struct sysmmu_fault fault;
int ret = -ENOSYS;
WARN_ON(!data->active);
spin_lock(&data->lock);
clk_enable(data->clk_master);
itype = __ffs(readl(SYSMMU_REG(data, int_status)));
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
ret = data->variant->get_fault_info(data, itype, &fault);
if (ret) {
dev_err(data->sysmmu, "Unhandled interrupt bit %u\n", itype);
goto out;
}
show_fault_information(data, &fault);
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
if (data->domain) {
ret = report_iommu_fault(&data->domain->domain, data->master,
fault.addr, fault.type);
}
if (ret)
panic("Unrecoverable System MMU Fault!");
out:
writel(1 << itype, SYSMMU_REG(data, int_clear));
iommu/exynos: Abstract getting the fault info Fault info obtaining is implemented for SysMMU v1..v5 in a very hardware specific way, as it relies on: - interrupt bits being tied to read or write access - having separate registers for the fault address w.r.t. AR/AW ops Newer SysMMU versions (like SysMMU v7) have different way of providing the fault info via registers: - the transaction type (read or write) should be read from the register (instead of hard-coding it w.r.t. corresponding interrupt status bit) - there is only one single register for storing the fault address Because of that, it is not possible to add newer SysMMU support into existing paradigm. Also it's not very effective performance-wise: - checking SysMMU version in ISR each time is not necessary - performing linear search to find the fault info by interrupt bit can be replaced with a single lookup operation Pave the way for adding support for new SysMMU versions by abstracting the getting of fault info in ISR. While at it, do some related style cleanups as well. This is mostly a refactoring patch, but there are some minor functional changes: - fault message format is a bit different; now instead of AR/AW prefixes for the fault's name, the request direction is printed as [READ]/[WRITE]. It has to be done to prepare an abstraction for SysMMU v7 support - don't panic on unknown interrupts; print corresponding message and continue - if fault wasn't recovered, panic with some sane message instead of just doing BUG_ON() The whole fault message looks like this now: [READ] PAGE FAULT occurred at 0x12341000 Signed-off-by: Sam Protsenko <semen.protsenko@linaro.org> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Link: https://lore.kernel.org/r/20220726200739.30017-2-semen.protsenko@linaro.org Signed-off-by: Joerg Roedel <jroedel@suse.de>
2022-07-26 20:07:38 +00:00
/* SysMMU is in blocked state when interrupt occurred */
sysmmu_unblock(data);
clk_disable(data->clk_master);
spin_unlock(&data->lock);
return IRQ_HANDLED;
}
static void __sysmmu_disable(struct sysmmu_drvdata *data)
{
unsigned long flags;
clk_enable(data->clk_master);
spin_lock_irqsave(&data->lock, flags);
writel(CTRL_DISABLE, data->sfrbase + REG_MMU_CTRL);
writel(0, data->sfrbase + REG_MMU_CFG);
data->active = false;
spin_unlock_irqrestore(&data->lock, flags);
__sysmmu_disable_clocks(data);
}
static void __sysmmu_init_config(struct sysmmu_drvdata *data)
{
unsigned int cfg;
if (data->version <= MAKE_MMU_VER(3, 1))
cfg = CFG_LRU | CFG_QOS(15);
else if (data->version <= MAKE_MMU_VER(3, 2))
cfg = CFG_LRU | CFG_QOS(15) | CFG_FLPDCACHE | CFG_SYSSEL;
else
cfg = CFG_QOS(15) | CFG_FLPDCACHE | CFG_ACGEN;
cfg |= CFG_EAP; /* enable access protection bits check */
writel(cfg, data->sfrbase + REG_MMU_CFG);
}
static void __sysmmu_enable_vid(struct sysmmu_drvdata *data)
{
u32 ctrl;
if (MMU_MAJ_VER(data->version) < 7 || !data->has_vcr)
return;
ctrl = readl(data->sfrbase + REG_V7_CTRL_VM);
ctrl |= CTRL_VM_ENABLE | CTRL_VM_FAULT_MODE_STALL;
writel(ctrl, data->sfrbase + REG_V7_CTRL_VM);
}
static void __sysmmu_enable(struct sysmmu_drvdata *data)
{
unsigned long flags;
__sysmmu_enable_clocks(data);
spin_lock_irqsave(&data->lock, flags);
writel(CTRL_BLOCK, data->sfrbase + REG_MMU_CTRL);
__sysmmu_init_config(data);
__sysmmu_set_ptbase(data, data->pgtable);
__sysmmu_enable_vid(data);
writel(CTRL_ENABLE, data->sfrbase + REG_MMU_CTRL);
data->active = true;
spin_unlock_irqrestore(&data->lock, flags);
/*
* SYSMMU driver keeps master's clock enabled only for the short
* time, while accessing the registers. For performing address
* translation during DMA transaction it relies on the client
* driver to enable it.
*/
clk_disable(data->clk_master);
}
static void sysmmu_tlb_invalidate_flpdcache(struct sysmmu_drvdata *data,
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
sysmmu_iova_t iova)
{
unsigned long flags;
spin_lock_irqsave(&data->lock, flags);
if (data->active && data->version >= MAKE_MMU_VER(3, 3)) {
clk_enable(data->clk_master);
if (sysmmu_block(data)) {
if (data->version >= MAKE_MMU_VER(5, 0))
__sysmmu_tlb_invalidate(data);
else
__sysmmu_tlb_invalidate_entry(data, iova, 1);
sysmmu_unblock(data);
}
clk_disable(data->clk_master);
}
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
spin_unlock_irqrestore(&data->lock, flags);
}
static void sysmmu_tlb_invalidate_entry(struct sysmmu_drvdata *data,
sysmmu_iova_t iova, size_t size)
{
unsigned long flags;
spin_lock_irqsave(&data->lock, flags);
if (data->active) {
unsigned int num_inv = 1;
clk_enable(data->clk_master);
/*
* L2TLB invalidation required
* 4KB page: 1 invalidation
* 64KB page: 16 invalidations
* 1MB page: 64 invalidations
* because it is set-associative TLB
* with 8-way and 64 sets.
* 1MB page can be cached in one of all sets.
* 64KB page can be one of 16 consecutive sets.
*/
if (MMU_MAJ_VER(data->version) == 2)
num_inv = min_t(unsigned int, size / SPAGE_SIZE, 64);
if (sysmmu_block(data)) {
__sysmmu_tlb_invalidate_entry(data, iova, num_inv);
sysmmu_unblock(data);
}
clk_disable(data->clk_master);
}
spin_unlock_irqrestore(&data->lock, flags);
}
static const struct iommu_ops exynos_iommu_ops;
static int exynos_sysmmu_probe(struct platform_device *pdev)
{
int irq, ret;
struct device *dev = &pdev->dev;
struct sysmmu_drvdata *data;
struct resource *res;
data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
if (!data)
return -ENOMEM;
res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
data->sfrbase = devm_ioremap_resource(dev, res);
if (IS_ERR(data->sfrbase))
return PTR_ERR(data->sfrbase);
irq = platform_get_irq(pdev, 0);
if (irq <= 0)
return irq;
ret = devm_request_irq(dev, irq, exynos_sysmmu_irq, 0,
dev_name(dev), data);
if (ret) {
dev_err(dev, "Unabled to register handler of irq %d\n", irq);
return ret;
}
data->clk = devm_clk_get_optional(dev, "sysmmu");
if (IS_ERR(data->clk))
return PTR_ERR(data->clk);
data->aclk = devm_clk_get_optional(dev, "aclk");
if (IS_ERR(data->aclk))
return PTR_ERR(data->aclk);
data->pclk = devm_clk_get_optional(dev, "pclk");
if (IS_ERR(data->pclk))
return PTR_ERR(data->pclk);
if (!data->clk && (!data->aclk || !data->pclk)) {
dev_err(dev, "Failed to get device clock(s)!\n");
return -ENOSYS;
}
data->clk_master = devm_clk_get_optional(dev, "master");
if (IS_ERR(data->clk_master))
return PTR_ERR(data->clk_master);
data->sysmmu = dev;
spin_lock_init(&data->lock);
__sysmmu_get_version(data);
ret = iommu_device_sysfs_add(&data->iommu, &pdev->dev, NULL,
dev_name(data->sysmmu));
if (ret)
return ret;
platform_set_drvdata(pdev, data);
if (PG_ENT_SHIFT < 0) {
if (MMU_MAJ_VER(data->version) < 5) {
PG_ENT_SHIFT = SYSMMU_PG_ENT_SHIFT;
LV1_PROT = SYSMMU_LV1_PROT;
LV2_PROT = SYSMMU_LV2_PROT;
} else {
PG_ENT_SHIFT = SYSMMU_V5_PG_ENT_SHIFT;
LV1_PROT = SYSMMU_V5_LV1_PROT;
LV2_PROT = SYSMMU_V5_LV2_PROT;
}
}
if (MMU_MAJ_VER(data->version) >= 5) {
ret = dma_set_mask(dev, DMA_BIT_MASK(36));
if (ret) {
dev_err(dev, "Unable to set DMA mask: %d\n", ret);
goto err_dma_set_mask;
}
}
/*
* use the first registered sysmmu device for performing
* dma mapping operations on iommu page tables (cpu cache flush)
*/
if (!dma_dev)
dma_dev = &pdev->dev;
pm_runtime_enable(dev);
ret = iommu_device_register(&data->iommu, &exynos_iommu_ops, dev);
if (ret)
goto err_dma_set_mask;
return 0;
err_dma_set_mask:
iommu_device_sysfs_remove(&data->iommu);
return ret;
}
static int __maybe_unused exynos_sysmmu_suspend(struct device *dev)
{
struct sysmmu_drvdata *data = dev_get_drvdata(dev);
struct device *master = data->master;
if (master) {
struct exynos_iommu_owner *owner = dev_iommu_priv_get(master);
mutex_lock(&owner->rpm_lock);
if (&data->domain->domain != &exynos_identity_domain) {
dev_dbg(data->sysmmu, "saving state\n");
__sysmmu_disable(data);
}
mutex_unlock(&owner->rpm_lock);
}
return 0;
}
static int __maybe_unused exynos_sysmmu_resume(struct device *dev)
{
struct sysmmu_drvdata *data = dev_get_drvdata(dev);
struct device *master = data->master;
if (master) {
struct exynos_iommu_owner *owner = dev_iommu_priv_get(master);
mutex_lock(&owner->rpm_lock);
if (&data->domain->domain != &exynos_identity_domain) {
dev_dbg(data->sysmmu, "restoring state\n");
__sysmmu_enable(data);
}
mutex_unlock(&owner->rpm_lock);
}
return 0;
}
static const struct dev_pm_ops sysmmu_pm_ops = {
SET_RUNTIME_PM_OPS(exynos_sysmmu_suspend, exynos_sysmmu_resume, NULL)
SET_SYSTEM_SLEEP_PM_OPS(pm_runtime_force_suspend,
pm_runtime_force_resume)
};
static const struct of_device_id sysmmu_of_match[] = {
{ .compatible = "samsung,exynos-sysmmu", },
{ },
};
static struct platform_driver exynos_sysmmu_driver __refdata = {
.probe = exynos_sysmmu_probe,
.driver = {
.name = "exynos-sysmmu",
.of_match_table = sysmmu_of_match,
.pm = &sysmmu_pm_ops,
.suppress_bind_attrs = true,
}
};
static inline void exynos_iommu_set_pte(sysmmu_pte_t *ent, sysmmu_pte_t val)
{
dma_sync_single_for_cpu(dma_dev, virt_to_phys(ent), sizeof(*ent),
DMA_TO_DEVICE);
*ent = cpu_to_le32(val);
dma_sync_single_for_device(dma_dev, virt_to_phys(ent), sizeof(*ent),
DMA_TO_DEVICE);
}
static struct iommu_domain *exynos_iommu_domain_alloc_paging(struct device *dev)
{
struct exynos_iommu_domain *domain;
dma_addr_t handle;
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
int i;
/* Check if correct PTE offsets are initialized */
BUG_ON(PG_ENT_SHIFT < 0 || !dma_dev);
domain = kzalloc(sizeof(*domain), GFP_KERNEL);
if (!domain)
return NULL;
domain->pgtable = iommu_alloc_pages(GFP_KERNEL, 2);
if (!domain->pgtable)
goto err_pgtable;
domain->lv2entcnt = iommu_alloc_pages(GFP_KERNEL, 1);
if (!domain->lv2entcnt)
goto err_counter;
/* Workaround for System MMU v3.3 to prevent caching 1MiB mapping */
for (i = 0; i < NUM_LV1ENTRIES; i++)
domain->pgtable[i] = ZERO_LV2LINK;
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
handle = dma_map_single(dma_dev, domain->pgtable, LV1TABLE_SIZE,
DMA_TO_DEVICE);
/* For mapping page table entries we rely on dma == phys */
BUG_ON(handle != virt_to_phys(domain->pgtable));
if (dma_mapping_error(dma_dev, handle))
goto err_lv2ent;
spin_lock_init(&domain->lock);
spin_lock_init(&domain->pgtablelock);
INIT_LIST_HEAD(&domain->clients);
domain->domain.geometry.aperture_start = 0;
domain->domain.geometry.aperture_end = ~0UL;
domain->domain.geometry.force_aperture = true;
return &domain->domain;
err_lv2ent:
iommu_free_pages(domain->lv2entcnt, 1);
err_counter:
iommu_free_pages(domain->pgtable, 2);
err_pgtable:
kfree(domain);
return NULL;
}
static void exynos_iommu_domain_free(struct iommu_domain *iommu_domain)
{
struct exynos_iommu_domain *domain = to_exynos_domain(iommu_domain);
struct sysmmu_drvdata *data, *next;
unsigned long flags;
int i;
WARN_ON(!list_empty(&domain->clients));
spin_lock_irqsave(&domain->lock, flags);
list_for_each_entry_safe(data, next, &domain->clients, domain_node) {
spin_lock(&data->lock);
__sysmmu_disable(data);
data->pgtable = 0;
data->domain = NULL;
list_del_init(&data->domain_node);
spin_unlock(&data->lock);
}
spin_unlock_irqrestore(&domain->lock, flags);
dma_unmap_single(dma_dev, virt_to_phys(domain->pgtable), LV1TABLE_SIZE,
DMA_TO_DEVICE);
for (i = 0; i < NUM_LV1ENTRIES; i++)
if (lv1ent_page(domain->pgtable + i)) {
phys_addr_t base = lv2table_base(domain->pgtable + i);
dma_unmap_single(dma_dev, base, LV2TABLE_SIZE,
DMA_TO_DEVICE);
kmem_cache_free(lv2table_kmem_cache,
phys_to_virt(base));
}
iommu_free_pages(domain->pgtable, 2);
iommu_free_pages(domain->lv2entcnt, 1);
kfree(domain);
}
static int exynos_iommu_identity_attach(struct iommu_domain *identity_domain,
struct device *dev)
{
struct exynos_iommu_owner *owner = dev_iommu_priv_get(dev);
struct exynos_iommu_domain *domain;
phys_addr_t pagetable;
struct sysmmu_drvdata *data, *next;
unsigned long flags;
if (owner->domain == identity_domain)
return 0;
domain = to_exynos_domain(owner->domain);
pagetable = virt_to_phys(domain->pgtable);
mutex_lock(&owner->rpm_lock);
list_for_each_entry(data, &owner->controllers, owner_node) {
pm_runtime_get_noresume(data->sysmmu);
if (pm_runtime_active(data->sysmmu))
__sysmmu_disable(data);
pm_runtime_put(data->sysmmu);
}
spin_lock_irqsave(&domain->lock, flags);
list_for_each_entry_safe(data, next, &domain->clients, domain_node) {
spin_lock(&data->lock);
data->pgtable = 0;
data->domain = NULL;
list_del_init(&data->domain_node);
spin_unlock(&data->lock);
}
owner->domain = identity_domain;
spin_unlock_irqrestore(&domain->lock, flags);
mutex_unlock(&owner->rpm_lock);
dev_dbg(dev, "%s: Restored IOMMU to IDENTITY from pgtable %pa\n",
__func__, &pagetable);
return 0;
}
static struct iommu_domain_ops exynos_identity_ops = {
.attach_dev = exynos_iommu_identity_attach,
};
static struct iommu_domain exynos_identity_domain = {
.type = IOMMU_DOMAIN_IDENTITY,
.ops = &exynos_identity_ops,
};
static int exynos_iommu_attach_device(struct iommu_domain *iommu_domain,
struct device *dev)
{
struct exynos_iommu_domain *domain = to_exynos_domain(iommu_domain);
struct exynos_iommu_owner *owner = dev_iommu_priv_get(dev);
struct sysmmu_drvdata *data;
phys_addr_t pagetable = virt_to_phys(domain->pgtable);
unsigned long flags;
int err;
err = exynos_iommu_identity_attach(&exynos_identity_domain, dev);
if (err)
return err;
mutex_lock(&owner->rpm_lock);
spin_lock_irqsave(&domain->lock, flags);
list_for_each_entry(data, &owner->controllers, owner_node) {
spin_lock(&data->lock);
data->pgtable = pagetable;
data->domain = domain;
list_add_tail(&data->domain_node, &domain->clients);
spin_unlock(&data->lock);
}
owner->domain = iommu_domain;
spin_unlock_irqrestore(&domain->lock, flags);
list_for_each_entry(data, &owner->controllers, owner_node) {
pm_runtime_get_noresume(data->sysmmu);
if (pm_runtime_active(data->sysmmu))
__sysmmu_enable(data);
pm_runtime_put(data->sysmmu);
}
mutex_unlock(&owner->rpm_lock);
dev_dbg(dev, "%s: Attached IOMMU with pgtable %pa\n", __func__,
&pagetable);
return 0;
}
static sysmmu_pte_t *alloc_lv2entry(struct exynos_iommu_domain *domain,
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
sysmmu_pte_t *sent, sysmmu_iova_t iova, short *pgcounter)
{
if (lv1ent_section(sent)) {
WARN(1, "Trying mapping on %#08x mapped with 1MiB page", iova);
return ERR_PTR(-EADDRINUSE);
}
if (lv1ent_fault(sent)) {
dma_addr_t handle;
sysmmu_pte_t *pent;
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
bool need_flush_flpd_cache = lv1ent_zero(sent);
pent = kmem_cache_zalloc(lv2table_kmem_cache, GFP_ATOMIC);
BUG_ON((uintptr_t)pent & (LV2TABLE_SIZE - 1));
if (!pent)
return ERR_PTR(-ENOMEM);
exynos_iommu_set_pte(sent, mk_lv1ent_page(virt_to_phys(pent)));
kmemleak_ignore(pent);
*pgcounter = NUM_LV2ENTRIES;
handle = dma_map_single(dma_dev, pent, LV2TABLE_SIZE,
DMA_TO_DEVICE);
if (dma_mapping_error(dma_dev, handle)) {
kmem_cache_free(lv2table_kmem_cache, pent);
return ERR_PTR(-EADDRINUSE);
}
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
/*
* If pre-fetched SLPD is a faulty SLPD in zero_l2_table,
* FLPD cache may cache the address of zero_l2_table. This
* function replaces the zero_l2_table with new L2 page table
* to write valid mappings.
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
* Accessing the valid area may cause page fault since FLPD
* cache may still cache zero_l2_table for the valid area
* instead of new L2 page table that has the mapping
* information of the valid area.
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
* Thus any replacement of zero_l2_table with other valid L2
* page table must involve FLPD cache invalidation for System
* MMU v3.3.
* FLPD cache invalidation is performed with TLB invalidation
* by VPN without blocking. It is safe to invalidate TLB without
* blocking because the target address of TLB invalidation is
* not currently mapped.
*/
if (need_flush_flpd_cache) {
struct sysmmu_drvdata *data;
spin_lock(&domain->lock);
list_for_each_entry(data, &domain->clients, domain_node)
sysmmu_tlb_invalidate_flpdcache(data, iova);
spin_unlock(&domain->lock);
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
}
}
return page_entry(sent, iova);
}
static int lv1set_section(struct exynos_iommu_domain *domain,
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
sysmmu_pte_t *sent, sysmmu_iova_t iova,
phys_addr_t paddr, int prot, short *pgcnt)
{
if (lv1ent_section(sent)) {
WARN(1, "Trying mapping on 1MiB@%#08x that is mapped",
iova);
return -EADDRINUSE;
}
if (lv1ent_page(sent)) {
if (*pgcnt != NUM_LV2ENTRIES) {
WARN(1, "Trying mapping on 1MiB@%#08x that is mapped",
iova);
return -EADDRINUSE;
}
kmem_cache_free(lv2table_kmem_cache, page_entry(sent, 0));
*pgcnt = 0;
}
exynos_iommu_set_pte(sent, mk_lv1ent_sect(paddr, prot));
spin_lock(&domain->lock);
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
if (lv1ent_page_zero(sent)) {
struct sysmmu_drvdata *data;
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
/*
* Flushing FLPD cache in System MMU v3.3 that may cache a FLPD
* entry by speculative prefetch of SLPD which has no mapping.
*/
list_for_each_entry(data, &domain->clients, domain_node)
sysmmu_tlb_invalidate_flpdcache(data, iova);
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
}
spin_unlock(&domain->lock);
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
return 0;
}
static int lv2set_page(sysmmu_pte_t *pent, phys_addr_t paddr, size_t size,
int prot, short *pgcnt)
{
if (size == SPAGE_SIZE) {
if (WARN_ON(!lv2ent_fault(pent)))
return -EADDRINUSE;
exynos_iommu_set_pte(pent, mk_lv2ent_spage(paddr, prot));
*pgcnt -= 1;
} else { /* size == LPAGE_SIZE */
int i;
dma_addr_t pent_base = virt_to_phys(pent);
dma_sync_single_for_cpu(dma_dev, pent_base,
sizeof(*pent) * SPAGES_PER_LPAGE,
DMA_TO_DEVICE);
for (i = 0; i < SPAGES_PER_LPAGE; i++, pent++) {
if (WARN_ON(!lv2ent_fault(pent))) {
if (i > 0)
memset(pent - i, 0, sizeof(*pent) * i);
return -EADDRINUSE;
}
*pent = mk_lv2ent_lpage(paddr, prot);
}
dma_sync_single_for_device(dma_dev, pent_base,
sizeof(*pent) * SPAGES_PER_LPAGE,
DMA_TO_DEVICE);
*pgcnt -= SPAGES_PER_LPAGE;
}
return 0;
}
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
/*
* *CAUTION* to the I/O virtual memory managers that support exynos-iommu:
*
* System MMU v3.x has advanced logic to improve address translation
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
* performance with caching more page table entries by a page table walk.
* However, the logic has a bug that while caching faulty page table entries,
* System MMU reports page fault if the cached fault entry is hit even though
* the fault entry is updated to a valid entry after the entry is cached.
* To prevent caching faulty page table entries which may be updated to valid
* entries later, the virtual memory manager should care about the workaround
* for the problem. The following describes the workaround.
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
*
* Any two consecutive I/O virtual address regions must have a hole of 128KiB
* at maximum to prevent misbehavior of System MMU 3.x (workaround for h/w bug).
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
*
* Precisely, any start address of I/O virtual region must be aligned with
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
* the following sizes for System MMU v3.1 and v3.2.
* System MMU v3.1: 128KiB
* System MMU v3.2: 256KiB
*
* Because System MMU v3.3 caches page table entries more aggressively, it needs
* more workarounds.
* - Any two consecutive I/O virtual regions must have a hole of size larger
* than or equal to 128KiB.
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
* - Start address of an I/O virtual region must be aligned by 128KiB.
*/
static int exynos_iommu_map(struct iommu_domain *iommu_domain,
unsigned long l_iova, phys_addr_t paddr, size_t size,
size_t count, int prot, gfp_t gfp, size_t *mapped)
{
struct exynos_iommu_domain *domain = to_exynos_domain(iommu_domain);
sysmmu_pte_t *entry;
sysmmu_iova_t iova = (sysmmu_iova_t)l_iova;
unsigned long flags;
int ret = -ENOMEM;
BUG_ON(domain->pgtable == NULL);
prot &= SYSMMU_SUPPORTED_PROT_BITS;
spin_lock_irqsave(&domain->pgtablelock, flags);
entry = section_entry(domain->pgtable, iova);
if (size == SECT_SIZE) {
ret = lv1set_section(domain, entry, iova, paddr, prot,
&domain->lv2entcnt[lv1ent_offset(iova)]);
} else {
sysmmu_pte_t *pent;
pent = alloc_lv2entry(domain, entry, iova,
&domain->lv2entcnt[lv1ent_offset(iova)]);
if (IS_ERR(pent))
ret = PTR_ERR(pent);
else
ret = lv2set_page(pent, paddr, size, prot,
&domain->lv2entcnt[lv1ent_offset(iova)]);
}
if (ret)
pr_err("%s: Failed(%d) to map %#zx bytes @ %#x\n",
__func__, ret, size, iova);
else
*mapped = size;
spin_unlock_irqrestore(&domain->pgtablelock, flags);
return ret;
}
static void exynos_iommu_tlb_invalidate_entry(struct exynos_iommu_domain *domain,
sysmmu_iova_t iova, size_t size)
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
{
struct sysmmu_drvdata *data;
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
unsigned long flags;
spin_lock_irqsave(&domain->lock, flags);
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
list_for_each_entry(data, &domain->clients, domain_node)
sysmmu_tlb_invalidate_entry(data, iova, size);
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
spin_unlock_irqrestore(&domain->lock, flags);
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
}
static size_t exynos_iommu_unmap(struct iommu_domain *iommu_domain,
unsigned long l_iova, size_t size, size_t count,
struct iommu_iotlb_gather *gather)
{
struct exynos_iommu_domain *domain = to_exynos_domain(iommu_domain);
sysmmu_iova_t iova = (sysmmu_iova_t)l_iova;
sysmmu_pte_t *ent;
size_t err_pgsize;
unsigned long flags;
BUG_ON(domain->pgtable == NULL);
spin_lock_irqsave(&domain->pgtablelock, flags);
ent = section_entry(domain->pgtable, iova);
if (lv1ent_section(ent)) {
if (WARN_ON(size < SECT_SIZE)) {
err_pgsize = SECT_SIZE;
goto err;
}
/* workaround for h/w bug in System MMU v3.3 */
exynos_iommu_set_pte(ent, ZERO_LV2LINK);
size = SECT_SIZE;
goto done;
}
if (unlikely(lv1ent_fault(ent))) {
if (size > SECT_SIZE)
size = SECT_SIZE;
goto done;
}
/* lv1ent_page(sent) == true here */
ent = page_entry(ent, iova);
if (unlikely(lv2ent_fault(ent))) {
size = SPAGE_SIZE;
goto done;
}
if (lv2ent_small(ent)) {
exynos_iommu_set_pte(ent, 0);
size = SPAGE_SIZE;
domain->lv2entcnt[lv1ent_offset(iova)] += 1;
goto done;
}
/* lv1ent_large(ent) == true here */
if (WARN_ON(size < LPAGE_SIZE)) {
err_pgsize = LPAGE_SIZE;
goto err;
}
dma_sync_single_for_cpu(dma_dev, virt_to_phys(ent),
sizeof(*ent) * SPAGES_PER_LPAGE,
DMA_TO_DEVICE);
memset(ent, 0, sizeof(*ent) * SPAGES_PER_LPAGE);
dma_sync_single_for_device(dma_dev, virt_to_phys(ent),
sizeof(*ent) * SPAGES_PER_LPAGE,
DMA_TO_DEVICE);
size = LPAGE_SIZE;
domain->lv2entcnt[lv1ent_offset(iova)] += SPAGES_PER_LPAGE;
done:
spin_unlock_irqrestore(&domain->pgtablelock, flags);
exynos_iommu_tlb_invalidate_entry(domain, iova, size);
return size;
err:
spin_unlock_irqrestore(&domain->pgtablelock, flags);
pr_err("%s: Failed: size(%#zx) @ %#x is smaller than page size %#zx\n",
__func__, size, iova, err_pgsize);
return 0;
}
static phys_addr_t exynos_iommu_iova_to_phys(struct iommu_domain *iommu_domain,
dma_addr_t iova)
{
struct exynos_iommu_domain *domain = to_exynos_domain(iommu_domain);
sysmmu_pte_t *entry;
unsigned long flags;
phys_addr_t phys = 0;
spin_lock_irqsave(&domain->pgtablelock, flags);
entry = section_entry(domain->pgtable, iova);
if (lv1ent_section(entry)) {
phys = section_phys(entry) + section_offs(iova);
} else if (lv1ent_page(entry)) {
entry = page_entry(entry, iova);
if (lv2ent_large(entry))
phys = lpage_phys(entry) + lpage_offs(iova);
else if (lv2ent_small(entry))
phys = spage_phys(entry) + spage_offs(iova);
}
spin_unlock_irqrestore(&domain->pgtablelock, flags);
return phys;
}
static struct iommu_device *exynos_iommu_probe_device(struct device *dev)
{
struct exynos_iommu_owner *owner = dev_iommu_priv_get(dev);
struct sysmmu_drvdata *data;
if (!has_sysmmu(dev))
return ERR_PTR(-ENODEV);
list_for_each_entry(data, &owner->controllers, owner_node) {
/*
* SYSMMU will be runtime activated via device link
* (dependency) to its master device, so there are no
* direct calls to pm_runtime_get/put in this driver.
*/
data->link = device_link_add(dev, data->sysmmu,
DL_FLAG_STATELESS |
DL_FLAG_PM_RUNTIME);
}
/* There is always at least one entry, see exynos_iommu_of_xlate() */
data = list_first_entry(&owner->controllers,
struct sysmmu_drvdata, owner_node);
return &data->iommu;
}
static void exynos_iommu_release_device(struct device *dev)
{
struct exynos_iommu_owner *owner = dev_iommu_priv_get(dev);
struct sysmmu_drvdata *data;
WARN_ON(exynos_iommu_identity_attach(&exynos_identity_domain, dev));
list_for_each_entry(data, &owner->controllers, owner_node)
device_link_del(data->link);
}
static int exynos_iommu_of_xlate(struct device *dev,
const struct of_phandle_args *spec)
{
struct platform_device *sysmmu = of_find_device_by_node(spec->np);
struct exynos_iommu_owner *owner = dev_iommu_priv_get(dev);
struct sysmmu_drvdata *data, *entry;
if (!sysmmu)
return -ENODEV;
data = platform_get_drvdata(sysmmu);
if (!data) {
put_device(&sysmmu->dev);
return -ENODEV;
}
if (!owner) {
owner = kzalloc(sizeof(*owner), GFP_KERNEL);
if (!owner) {
put_device(&sysmmu->dev);
return -ENOMEM;
}
INIT_LIST_HEAD(&owner->controllers);
mutex_init(&owner->rpm_lock);
owner->domain = &exynos_identity_domain;
dev_iommu_priv_set(dev, owner);
}
list_for_each_entry(entry, &owner->controllers, owner_node)
if (entry == data)
return 0;
list_add_tail(&data->owner_node, &owner->controllers);
data->master = dev;
return 0;
}
static const struct iommu_ops exynos_iommu_ops = {
.identity_domain = &exynos_identity_domain,
.domain_alloc_paging = exynos_iommu_domain_alloc_paging,
.device_group = generic_device_group,
.probe_device = exynos_iommu_probe_device,
.release_device = exynos_iommu_release_device,
.pgsize_bitmap = SECT_SIZE | LPAGE_SIZE | SPAGE_SIZE,
.of_xlate = exynos_iommu_of_xlate,
.default_domain_ops = &(const struct iommu_domain_ops) {
.attach_dev = exynos_iommu_attach_device,
.map_pages = exynos_iommu_map,
.unmap_pages = exynos_iommu_unmap,
.iova_to_phys = exynos_iommu_iova_to_phys,
.free = exynos_iommu_domain_free,
}
};
static int __init exynos_iommu_init(void)
{
struct device_node *np;
int ret;
np = of_find_matching_node(NULL, sysmmu_of_match);
if (!np)
return 0;
of_node_put(np);
lv2table_kmem_cache = kmem_cache_create("exynos-iommu-lv2table",
LV2TABLE_SIZE, LV2TABLE_SIZE, 0, NULL);
if (!lv2table_kmem_cache) {
pr_err("%s: Failed to create kmem cache\n", __func__);
return -ENOMEM;
}
iommu/exynos: Apply workaround of caching fault page table entries This patch contains 2 workaround for the System MMU v3.x. System MMU v3.2 and v3.3 has FLPD cache that caches first level page table entries to reduce page table walking latency. However, the FLPD cache is filled with a first level page table entry even though it is not accessed by a master H/W because System MMU v3.3 speculatively prefetches page table entries that may be accessed in the near future by the master H/W. The prefetched FLPD cache entries are not invalidated by iommu_unmap() because iommu_unmap() only unmaps and invalidates the page table entries that is mapped. Because exynos-iommu driver discards a second level page table when it needs to be replaced with another second level page table or a first level page table entry with 1MB mapping, It is required to invalidate FLPD cache that may contain the first level page table entry that points to the second level page table. Another workaround of System MMU v3.3 is initializing the first level page table entries with the second level page table which is filled with all zeros. This prevents System MMU prefetches 'fault' first level page table entry which may lead page fault on access to 16MiB wide. System MMU 3.x fetches consecutive page table entries by a page table walking to maximize bus utilization and to minimize TLB miss panelty. Unfortunately, functional problem is raised with the fetching behavior because it fetches 'fault' page table entries that specifies no translation information and that a valid translation information will be written to in the near future. The logic in the System MMU generates page fault with the cached fault entries that is no longer coherent with the page table which is updated. There is another workaround that must be implemented by I/O virtual memory manager: any two consecutive I/O virtual memory area must have a hole between the two that is larger than or equal to 128KiB. Also, next I/O virtual memory area must be started from the next 128KiB boundary. 0 128K 256K 384K 512K |-------------|---------------|-----------------|----------------| |area1---------------->|.........hole...........|<--- area2 ----- The constraint is depicted above. The size is selected by the calculation followed: - System MMU can fetch consecutive 64 page table entries at once 64 * 4KiB = 256KiB. This is the size between 128K ~ 384K of the above picture. This style of fetching is 'block fetch'. It fetches the page table entries predefined consecutive page table entries including the entry that is the reason of the page table walking. - System MMU can prefetch upto consecutive 32 page table entries. This is the size between 256K ~ 384K. Signed-off-by: Cho KyongHo <pullip.cho@samsung.com> Signed-off-by: Shaik Ameer Basha <shaik.ameer@samsung.com> Signed-off-by: Joerg Roedel <jroedel@suse.de>
2014-05-12 06:15:04 +00:00
zero_lv2_table = kmem_cache_zalloc(lv2table_kmem_cache, GFP_KERNEL);
if (zero_lv2_table == NULL) {
pr_err("%s: Failed to allocate zero level2 page table\n",
__func__);
ret = -ENOMEM;
goto err_zero_lv2;
}
ret = platform_driver_register(&exynos_sysmmu_driver);
if (ret) {
pr_err("%s: Failed to register driver\n", __func__);
goto err_reg_driver;
}
return 0;
err_reg_driver:
kmem_cache_free(lv2table_kmem_cache, zero_lv2_table);
err_zero_lv2:
kmem_cache_destroy(lv2table_kmem_cache);
return ret;
}
core_initcall(exynos_iommu_init);