linux/drivers/net/spider_net.c

2604 lines
70 KiB
C
Raw Normal View History

/*
* Network device driver for Cell Processor-Based Blade and Celleb platform
*
* (C) Copyright IBM Corp. 2005
* (C) Copyright 2006 TOSHIBA CORPORATION
*
* Authors : Utz Bacher <utz.bacher@de.ibm.com>
* Jens Osterkamp <Jens.Osterkamp@de.ibm.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2, or (at your option)
* any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#include <linux/compiler.h>
#include <linux/crc32.h>
#include <linux/delay.h>
#include <linux/etherdevice.h>
#include <linux/ethtool.h>
#include <linux/firmware.h>
#include <linux/if_vlan.h>
#include <linux/in.h>
#include <linux/init.h>
#include <linux/interrupt.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 08:04:11 +00:00
#include <linux/gfp.h>
#include <linux/ioport.h>
#include <linux/ip.h>
#include <linux/kernel.h>
#include <linux/mii.h>
#include <linux/module.h>
#include <linux/netdevice.h>
#include <linux/device.h>
#include <linux/pci.h>
#include <linux/skbuff.h>
#include <linux/tcp.h>
#include <linux/types.h>
#include <linux/vmalloc.h>
#include <linux/wait.h>
#include <linux/workqueue.h>
#include <linux/bitops.h>
#include <asm/pci-bridge.h>
#include <net/checksum.h>
#include "spider_net.h"
MODULE_AUTHOR("Utz Bacher <utz.bacher@de.ibm.com> and Jens Osterkamp " \
"<Jens.Osterkamp@de.ibm.com>");
MODULE_DESCRIPTION("Spider Southbridge Gigabit Ethernet driver");
MODULE_LICENSE("GPL");
MODULE_VERSION(VERSION);
MODULE_FIRMWARE(SPIDER_NET_FIRMWARE_NAME);
static int rx_descriptors = SPIDER_NET_RX_DESCRIPTORS_DEFAULT;
static int tx_descriptors = SPIDER_NET_TX_DESCRIPTORS_DEFAULT;
module_param(rx_descriptors, int, 0444);
module_param(tx_descriptors, int, 0444);
MODULE_PARM_DESC(rx_descriptors, "number of descriptors used " \
"in rx chains");
MODULE_PARM_DESC(tx_descriptors, "number of descriptors used " \
"in tx chain");
char spider_net_driver_name[] = "spidernet";
static DEFINE_PCI_DEVICE_TABLE(spider_net_pci_tbl) = {
{ PCI_VENDOR_ID_TOSHIBA_2, PCI_DEVICE_ID_TOSHIBA_SPIDER_NET,
PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0UL },
{ 0, }
};
MODULE_DEVICE_TABLE(pci, spider_net_pci_tbl);
/**
* spider_net_read_reg - reads an SMMIO register of a card
* @card: device structure
* @reg: register to read from
*
* returns the content of the specified SMMIO register.
*/
static inline u32
spider_net_read_reg(struct spider_net_card *card, u32 reg)
{
/* We use the powerpc specific variants instead of readl_be() because
* we know spidernet is not a real PCI device and we can thus avoid the
* performance hit caused by the PCI workarounds.
*/
return in_be32(card->regs + reg);
}
/**
* spider_net_write_reg - writes to an SMMIO register of a card
* @card: device structure
* @reg: register to write to
* @value: value to write into the specified SMMIO register
*/
static inline void
spider_net_write_reg(struct spider_net_card *card, u32 reg, u32 value)
{
/* We use the powerpc specific variants instead of writel_be() because
* we know spidernet is not a real PCI device and we can thus avoid the
* performance hit caused by the PCI workarounds.
*/
out_be32(card->regs + reg, value);
}
/** spider_net_write_phy - write to phy register
* @netdev: adapter to be written to
* @mii_id: id of MII
* @reg: PHY register
* @val: value to be written to phy register
*
* spider_net_write_phy_register writes to an arbitrary PHY
* register via the spider GPCWOPCMD register. We assume the queue does
* not run full (not more than 15 commands outstanding).
**/
static void
spider_net_write_phy(struct net_device *netdev, int mii_id,
int reg, int val)
{
struct spider_net_card *card = netdev_priv(netdev);
u32 writevalue;
writevalue = ((u32)mii_id << 21) |
((u32)reg << 16) | ((u32)val);
spider_net_write_reg(card, SPIDER_NET_GPCWOPCMD, writevalue);
}
/** spider_net_read_phy - read from phy register
* @netdev: network device to be read from
* @mii_id: id of MII
* @reg: PHY register
*
* Returns value read from PHY register
*
* spider_net_write_phy reads from an arbitrary PHY
* register via the spider GPCROPCMD register
**/
static int
spider_net_read_phy(struct net_device *netdev, int mii_id, int reg)
{
struct spider_net_card *card = netdev_priv(netdev);
u32 readvalue;
readvalue = ((u32)mii_id << 21) | ((u32)reg << 16);
spider_net_write_reg(card, SPIDER_NET_GPCROPCMD, readvalue);
/* we don't use semaphores to wait for an SPIDER_NET_GPROPCMPINT
* interrupt, as we poll for the completion of the read operation
* in spider_net_read_phy. Should take about 50 us */
do {
readvalue = spider_net_read_reg(card, SPIDER_NET_GPCROPCMD);
} while (readvalue & SPIDER_NET_GPREXEC);
readvalue &= SPIDER_NET_GPRDAT_MASK;
return readvalue;
}
/**
* spider_net_setup_aneg - initial auto-negotiation setup
* @card: device structure
**/
static void
spider_net_setup_aneg(struct spider_net_card *card)
{
struct mii_phy *phy = &card->phy;
u32 advertise = 0;
u16 bmsr, estat;
bmsr = spider_net_read_phy(card->netdev, phy->mii_id, MII_BMSR);
estat = spider_net_read_phy(card->netdev, phy->mii_id, MII_ESTATUS);
if (bmsr & BMSR_10HALF)
advertise |= ADVERTISED_10baseT_Half;
if (bmsr & BMSR_10FULL)
advertise |= ADVERTISED_10baseT_Full;
if (bmsr & BMSR_100HALF)
advertise |= ADVERTISED_100baseT_Half;
if (bmsr & BMSR_100FULL)
advertise |= ADVERTISED_100baseT_Full;
if ((bmsr & BMSR_ESTATEN) && (estat & ESTATUS_1000_TFULL))
advertise |= SUPPORTED_1000baseT_Full;
if ((bmsr & BMSR_ESTATEN) && (estat & ESTATUS_1000_THALF))
advertise |= SUPPORTED_1000baseT_Half;
mii_phy_probe(phy, phy->mii_id);
phy->def->ops->setup_aneg(phy, advertise);
}
/**
* spider_net_rx_irq_off - switch off rx irq on this spider card
* @card: device structure
*
* switches off rx irq by masking them out in the GHIINTnMSK register
*/
static void
spider_net_rx_irq_off(struct spider_net_card *card)
{
u32 regvalue;
regvalue = SPIDER_NET_INT0_MASK_VALUE & (~SPIDER_NET_RXINT);
spider_net_write_reg(card, SPIDER_NET_GHIINT0MSK, regvalue);
}
/**
* spider_net_rx_irq_on - switch on rx irq on this spider card
* @card: device structure
*
* switches on rx irq by enabling them in the GHIINTnMSK register
*/
static void
spider_net_rx_irq_on(struct spider_net_card *card)
{
u32 regvalue;
regvalue = SPIDER_NET_INT0_MASK_VALUE | SPIDER_NET_RXINT;
spider_net_write_reg(card, SPIDER_NET_GHIINT0MSK, regvalue);
}
/**
* spider_net_set_promisc - sets the unicast address or the promiscuous mode
* @card: card structure
*
* spider_net_set_promisc sets the unicast destination address filter and
* thus either allows for non-promisc mode or promisc mode
*/
static void
spider_net_set_promisc(struct spider_net_card *card)
{
u32 macu, macl;
struct net_device *netdev = card->netdev;
if (netdev->flags & IFF_PROMISC) {
/* clear destination entry 0 */
spider_net_write_reg(card, SPIDER_NET_GMRUAFILnR, 0);
spider_net_write_reg(card, SPIDER_NET_GMRUAFILnR + 0x04, 0);
spider_net_write_reg(card, SPIDER_NET_GMRUA0FIL15R,
SPIDER_NET_PROMISC_VALUE);
} else {
macu = netdev->dev_addr[0];
macu <<= 8;
macu |= netdev->dev_addr[1];
memcpy(&macl, &netdev->dev_addr[2], sizeof(macl));
macu |= SPIDER_NET_UA_DESCR_VALUE;
spider_net_write_reg(card, SPIDER_NET_GMRUAFILnR, macu);
spider_net_write_reg(card, SPIDER_NET_GMRUAFILnR + 0x04, macl);
spider_net_write_reg(card, SPIDER_NET_GMRUA0FIL15R,
SPIDER_NET_NONPROMISC_VALUE);
}
}
/**
* spider_net_get_mac_address - read mac address from spider card
* @card: device structure
*
* reads MAC address from GMACUNIMACU and GMACUNIMACL registers
*/
static int
spider_net_get_mac_address(struct net_device *netdev)
{
struct spider_net_card *card = netdev_priv(netdev);
u32 macl, macu;
macl = spider_net_read_reg(card, SPIDER_NET_GMACUNIMACL);
macu = spider_net_read_reg(card, SPIDER_NET_GMACUNIMACU);
netdev->dev_addr[0] = (macu >> 24) & 0xff;
netdev->dev_addr[1] = (macu >> 16) & 0xff;
netdev->dev_addr[2] = (macu >> 8) & 0xff;
netdev->dev_addr[3] = macu & 0xff;
netdev->dev_addr[4] = (macl >> 8) & 0xff;
netdev->dev_addr[5] = macl & 0xff;
if (!is_valid_ether_addr(&netdev->dev_addr[0]))
return -EINVAL;
return 0;
}
/**
* spider_net_get_descr_status -- returns the status of a descriptor
* @descr: descriptor to look at
*
* returns the status as in the dmac_cmd_status field of the descriptor
*/
static inline int
spider_net_get_descr_status(struct spider_net_hw_descr *hwdescr)
{
return hwdescr->dmac_cmd_status & SPIDER_NET_DESCR_IND_PROC_MASK;
}
/**
* spider_net_free_chain - free descriptor chain
* @card: card structure
* @chain: address of chain
*
*/
static void
spider_net_free_chain(struct spider_net_card *card,
struct spider_net_descr_chain *chain)
{
struct spider_net_descr *descr;
descr = chain->ring;
do {
descr->bus_addr = 0;
descr->hwdescr->next_descr_addr = 0;
descr = descr->next;
} while (descr != chain->ring);
dma_free_coherent(&card->pdev->dev, chain->num_desc,
chain->hwring, chain->dma_addr);
}
/**
* spider_net_init_chain - alloc and link descriptor chain
* @card: card structure
* @chain: address of chain
*
* We manage a circular list that mirrors the hardware structure,
* except that the hardware uses bus addresses.
*
* Returns 0 on success, <0 on failure
*/
static int
spider_net_init_chain(struct spider_net_card *card,
struct spider_net_descr_chain *chain)
{
int i;
struct spider_net_descr *descr;
struct spider_net_hw_descr *hwdescr;
dma_addr_t buf;
size_t alloc_size;
alloc_size = chain->num_desc * sizeof(struct spider_net_hw_descr);
chain->hwring = dma_alloc_coherent(&card->pdev->dev, alloc_size,
&chain->dma_addr, GFP_KERNEL);
if (!chain->hwring)
return -ENOMEM;
memset(chain->ring, 0, chain->num_desc * sizeof(struct spider_net_descr));
/* Set up the hardware pointers in each descriptor */
descr = chain->ring;
hwdescr = chain->hwring;
buf = chain->dma_addr;
for (i=0; i < chain->num_desc; i++, descr++, hwdescr++) {
hwdescr->dmac_cmd_status = SPIDER_NET_DESCR_NOT_IN_USE;
hwdescr->next_descr_addr = 0;
descr->hwdescr = hwdescr;
descr->bus_addr = buf;
descr->next = descr + 1;
descr->prev = descr - 1;
buf += sizeof(struct spider_net_hw_descr);
}
/* do actual circular list */
(descr-1)->next = chain->ring;
chain->ring->prev = descr-1;
spin_lock_init(&chain->lock);
chain->head = chain->ring;
chain->tail = chain->ring;
return 0;
}
/**
* spider_net_free_rx_chain_contents - frees descr contents in rx chain
* @card: card structure
*
* returns 0 on success, <0 on failure
*/
static void
spider_net_free_rx_chain_contents(struct spider_net_card *card)
{
struct spider_net_descr *descr;
descr = card->rx_chain.head;
do {
if (descr->skb) {
pci_unmap_single(card->pdev, descr->hwdescr->buf_addr,
SPIDER_NET_MAX_FRAME,
PCI_DMA_BIDIRECTIONAL);
dev_kfree_skb(descr->skb);
descr->skb = NULL;
}
descr = descr->next;
} while (descr != card->rx_chain.head);
}
/**
* spider_net_prepare_rx_descr - Reinitialize RX descriptor
* @card: card structure
* @descr: descriptor to re-init
*
* Return 0 on success, <0 on failure.
*
* Allocates a new rx skb, iommu-maps it and attaches it to the
* descriptor. Mark the descriptor as activated, ready-to-use.
*/
static int
spider_net_prepare_rx_descr(struct spider_net_card *card,
struct spider_net_descr *descr)
{
struct spider_net_hw_descr *hwdescr = descr->hwdescr;
dma_addr_t buf;
int offset;
int bufsize;
/* we need to round up the buffer size to a multiple of 128 */
bufsize = (SPIDER_NET_MAX_FRAME + SPIDER_NET_RXBUF_ALIGN - 1) &
(~(SPIDER_NET_RXBUF_ALIGN - 1));
/* and we need to have it 128 byte aligned, therefore we allocate a
* bit more */
/* allocate an skb */
descr->skb = netdev_alloc_skb(card->netdev,
bufsize + SPIDER_NET_RXBUF_ALIGN - 1);
if (!descr->skb) {
if (netif_msg_rx_err(card) && net_ratelimit())
dev_err(&card->netdev->dev,
"Not enough memory to allocate rx buffer\n");
card->spider_stats.alloc_rx_skb_error++;
return -ENOMEM;
}
hwdescr->buf_size = bufsize;
hwdescr->result_size = 0;
hwdescr->valid_size = 0;
hwdescr->data_status = 0;
hwdescr->data_error = 0;
offset = ((unsigned long)descr->skb->data) &
(SPIDER_NET_RXBUF_ALIGN - 1);
if (offset)
skb_reserve(descr->skb, SPIDER_NET_RXBUF_ALIGN - offset);
/* iommu-map the skb */
buf = pci_map_single(card->pdev, descr->skb->data,
SPIDER_NET_MAX_FRAME, PCI_DMA_FROMDEVICE);
dma-mapping: add the device argument to dma_mapping_error() Add per-device dma_mapping_ops support for CONFIG_X86_64 as POWER architecture does: This enables us to cleanly fix the Calgary IOMMU issue that some devices are not behind the IOMMU (http://lkml.org/lkml/2008/5/8/423). I think that per-device dma_mapping_ops support would be also helpful for KVM people to support PCI passthrough but Andi thinks that this makes it difficult to support the PCI passthrough (see the above thread). So I CC'ed this to KVM camp. Comments are appreciated. A pointer to dma_mapping_ops to struct dev_archdata is added. If the pointer is non NULL, DMA operations in asm/dma-mapping.h use it. If it's NULL, the system-wide dma_ops pointer is used as before. If it's useful for KVM people, I plan to implement a mechanism to register a hook called when a new pci (or dma capable) device is created (it works with hot plugging). It enables IOMMUs to set up an appropriate dma_mapping_ops per device. The major obstacle is that dma_mapping_error doesn't take a pointer to the device unlike other DMA operations. So x86 can't have dma_mapping_ops per device. Note all the POWER IOMMUs use the same dma_mapping_error function so this is not a problem for POWER but x86 IOMMUs use different dma_mapping_error functions. The first patch adds the device argument to dma_mapping_error. The patch is trivial but large since it touches lots of drivers and dma-mapping.h in all the architecture. This patch: dma_mapping_error() doesn't take a pointer to the device unlike other DMA operations. So we can't have dma_mapping_ops per device. Note that POWER already has dma_mapping_ops per device but all the POWER IOMMUs use the same dma_mapping_error function. x86 IOMMUs use device argument. [akpm@linux-foundation.org: fix sge] [akpm@linux-foundation.org: fix svc_rdma] [akpm@linux-foundation.org: build fix] [akpm@linux-foundation.org: fix bnx2x] [akpm@linux-foundation.org: fix s2io] [akpm@linux-foundation.org: fix pasemi_mac] [akpm@linux-foundation.org: fix sdhci] [akpm@linux-foundation.org: build fix] [akpm@linux-foundation.org: fix sparc] [akpm@linux-foundation.org: fix ibmvscsi] Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@elte.hu> Cc: Avi Kivity <avi@qumranet.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-26 02:44:49 +00:00
if (pci_dma_mapping_error(card->pdev, buf)) {
dev_kfree_skb_any(descr->skb);
descr->skb = NULL;
if (netif_msg_rx_err(card) && net_ratelimit())
dev_err(&card->netdev->dev, "Could not iommu-map rx buffer\n");
card->spider_stats.rx_iommu_map_error++;
hwdescr->dmac_cmd_status = SPIDER_NET_DESCR_NOT_IN_USE;
} else {
hwdescr->buf_addr = buf;
wmb();
hwdescr->dmac_cmd_status = SPIDER_NET_DESCR_CARDOWNED |
SPIDER_NET_DMAC_NOINTR_COMPLETE;
}
return 0;
}
/**
* spider_net_enable_rxchtails - sets RX dmac chain tail addresses
* @card: card structure
*
* spider_net_enable_rxchtails sets the RX DMAC chain tail addresses in the
* chip by writing to the appropriate register. DMA is enabled in
* spider_net_enable_rxdmac.
*/
static inline void
spider_net_enable_rxchtails(struct spider_net_card *card)
{
/* assume chain is aligned correctly */
spider_net_write_reg(card, SPIDER_NET_GDADCHA ,
card->rx_chain.tail->bus_addr);
}
/**
* spider_net_enable_rxdmac - enables a receive DMA controller
* @card: card structure
*
* spider_net_enable_rxdmac enables the DMA controller by setting RX_DMA_EN
* in the GDADMACCNTR register
*/
static inline void
spider_net_enable_rxdmac(struct spider_net_card *card)
{
wmb();
spider_net_write_reg(card, SPIDER_NET_GDADMACCNTR,
SPIDER_NET_DMA_RX_VALUE);
}
/**
* spider_net_disable_rxdmac - disables the receive DMA controller
* @card: card structure
*
* spider_net_disable_rxdmac terminates processing on the DMA controller
* by turing off the DMA controller, with the force-end flag set.
*/
static inline void
spider_net_disable_rxdmac(struct spider_net_card *card)
{
spider_net_write_reg(card, SPIDER_NET_GDADMACCNTR,
SPIDER_NET_DMA_RX_FEND_VALUE);
}
/**
* spider_net_refill_rx_chain - refills descriptors/skbs in the rx chains
* @card: card structure
*
* refills descriptors in the rx chain: allocates skbs and iommu-maps them.
*/
static void
spider_net_refill_rx_chain(struct spider_net_card *card)
{
struct spider_net_descr_chain *chain = &card->rx_chain;
unsigned long flags;
/* one context doing the refill (and a second context seeing that
* and omitting it) is ok. If called by NAPI, we'll be called again
* as spider_net_decode_one_descr is called several times. If some
* interrupt calls us, the NAPI is about to clean up anyway. */
if (!spin_trylock_irqsave(&chain->lock, flags))
return;
while (spider_net_get_descr_status(chain->head->hwdescr) ==
SPIDER_NET_DESCR_NOT_IN_USE) {
if (spider_net_prepare_rx_descr(card, chain->head))
break;
chain->head = chain->head->next;
}
spin_unlock_irqrestore(&chain->lock, flags);
}
/**
* spider_net_alloc_rx_skbs - Allocates rx skbs in rx descriptor chains
* @card: card structure
*
* Returns 0 on success, <0 on failure.
*/
static int
spider_net_alloc_rx_skbs(struct spider_net_card *card)
{
struct spider_net_descr_chain *chain = &card->rx_chain;
struct spider_net_descr *start = chain->tail;
struct spider_net_descr *descr = start;
/* Link up the hardware chain pointers */
do {
descr->prev->hwdescr->next_descr_addr = descr->bus_addr;
descr = descr->next;
} while (descr != start);
/* Put at least one buffer into the chain. if this fails,
* we've got a problem. If not, spider_net_refill_rx_chain
* will do the rest at the end of this function. */
if (spider_net_prepare_rx_descr(card, chain->head))
goto error;
else
chain->head = chain->head->next;
/* This will allocate the rest of the rx buffers;
* if not, it's business as usual later on. */
spider_net_refill_rx_chain(card);
spider_net_enable_rxdmac(card);
return 0;
error:
spider_net_free_rx_chain_contents(card);
return -ENOMEM;
}
/**
* spider_net_get_multicast_hash - generates hash for multicast filter table
* @addr: multicast address
*
* returns the hash value.
*
* spider_net_get_multicast_hash calculates a hash value for a given multicast
* address, that is used to set the multicast filter tables
*/
static u8
spider_net_get_multicast_hash(struct net_device *netdev, __u8 *addr)
{
u32 crc;
u8 hash;
char addr_for_crc[ETH_ALEN] = { 0, };
int i, bit;
for (i = 0; i < ETH_ALEN * 8; i++) {
bit = (addr[i / 8] >> (i % 8)) & 1;
addr_for_crc[ETH_ALEN - 1 - i / 8] += bit << (7 - (i % 8));
}
crc = crc32_be(~0, addr_for_crc, netdev->addr_len);
hash = (crc >> 27);
hash <<= 3;
hash |= crc & 7;
hash &= 0xff;
return hash;
}
/**
* spider_net_set_multi - sets multicast addresses and promisc flags
* @netdev: interface device structure
*
* spider_net_set_multi configures multicast addresses as needed for the
* netdev interface. It also sets up multicast, allmulti and promisc
* flags appropriately
*/
static void
spider_net_set_multi(struct net_device *netdev)
{
struct netdev_hw_addr *ha;
u8 hash;
int i;
u32 reg;
struct spider_net_card *card = netdev_priv(netdev);
unsigned long bitmask[SPIDER_NET_MULTICAST_HASHES / BITS_PER_LONG] =
{0, };
spider_net_set_promisc(card);
if (netdev->flags & IFF_ALLMULTI) {
for (i = 0; i < SPIDER_NET_MULTICAST_HASHES; i++) {
set_bit(i, bitmask);
}
goto write_hash;
}
/* well, we know, what the broadcast hash value is: it's xfd
hash = spider_net_get_multicast_hash(netdev, netdev->broadcast); */
set_bit(0xfd, bitmask);
netdev_for_each_mc_addr(ha, netdev) {
hash = spider_net_get_multicast_hash(netdev, ha->addr);
set_bit(hash, bitmask);
}
write_hash:
for (i = 0; i < SPIDER_NET_MULTICAST_HASHES / 4; i++) {
reg = 0;
if (test_bit(i * 4, bitmask))
reg += 0x08;
reg <<= 8;
if (test_bit(i * 4 + 1, bitmask))
reg += 0x08;
reg <<= 8;
if (test_bit(i * 4 + 2, bitmask))
reg += 0x08;
reg <<= 8;
if (test_bit(i * 4 + 3, bitmask))
reg += 0x08;
spider_net_write_reg(card, SPIDER_NET_GMRMHFILnR + i * 4, reg);
}
}
/**
* spider_net_prepare_tx_descr - fill tx descriptor with skb data
* @card: card structure
* @skb: packet to use
*
* returns 0 on success, <0 on failure.
*
* fills out the descriptor structure with skb data and len. Copies data,
* if needed (32bit DMA!)
*/
static int
spider_net_prepare_tx_descr(struct spider_net_card *card,
struct sk_buff *skb)
{
struct spider_net_descr_chain *chain = &card->tx_chain;
struct spider_net_descr *descr;
struct spider_net_hw_descr *hwdescr;
dma_addr_t buf;
unsigned long flags;
buf = pci_map_single(card->pdev, skb->data, skb->len, PCI_DMA_TODEVICE);
dma-mapping: add the device argument to dma_mapping_error() Add per-device dma_mapping_ops support for CONFIG_X86_64 as POWER architecture does: This enables us to cleanly fix the Calgary IOMMU issue that some devices are not behind the IOMMU (http://lkml.org/lkml/2008/5/8/423). I think that per-device dma_mapping_ops support would be also helpful for KVM people to support PCI passthrough but Andi thinks that this makes it difficult to support the PCI passthrough (see the above thread). So I CC'ed this to KVM camp. Comments are appreciated. A pointer to dma_mapping_ops to struct dev_archdata is added. If the pointer is non NULL, DMA operations in asm/dma-mapping.h use it. If it's NULL, the system-wide dma_ops pointer is used as before. If it's useful for KVM people, I plan to implement a mechanism to register a hook called when a new pci (or dma capable) device is created (it works with hot plugging). It enables IOMMUs to set up an appropriate dma_mapping_ops per device. The major obstacle is that dma_mapping_error doesn't take a pointer to the device unlike other DMA operations. So x86 can't have dma_mapping_ops per device. Note all the POWER IOMMUs use the same dma_mapping_error function so this is not a problem for POWER but x86 IOMMUs use different dma_mapping_error functions. The first patch adds the device argument to dma_mapping_error. The patch is trivial but large since it touches lots of drivers and dma-mapping.h in all the architecture. This patch: dma_mapping_error() doesn't take a pointer to the device unlike other DMA operations. So we can't have dma_mapping_ops per device. Note that POWER already has dma_mapping_ops per device but all the POWER IOMMUs use the same dma_mapping_error function. x86 IOMMUs use device argument. [akpm@linux-foundation.org: fix sge] [akpm@linux-foundation.org: fix svc_rdma] [akpm@linux-foundation.org: build fix] [akpm@linux-foundation.org: fix bnx2x] [akpm@linux-foundation.org: fix s2io] [akpm@linux-foundation.org: fix pasemi_mac] [akpm@linux-foundation.org: fix sdhci] [akpm@linux-foundation.org: build fix] [akpm@linux-foundation.org: fix sparc] [akpm@linux-foundation.org: fix ibmvscsi] Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@elte.hu> Cc: Avi Kivity <avi@qumranet.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-26 02:44:49 +00:00
if (pci_dma_mapping_error(card->pdev, buf)) {
if (netif_msg_tx_err(card) && net_ratelimit())
dev_err(&card->netdev->dev, "could not iommu-map packet (%p, %i). "
"Dropping packet\n", skb->data, skb->len);
card->spider_stats.tx_iommu_map_error++;
return -ENOMEM;
}
spin_lock_irqsave(&chain->lock, flags);
descr = card->tx_chain.head;
if (descr->next == chain->tail->prev) {
spin_unlock_irqrestore(&chain->lock, flags);
pci_unmap_single(card->pdev, buf, skb->len, PCI_DMA_TODEVICE);
return -ENOMEM;
}
hwdescr = descr->hwdescr;
chain->head = descr->next;
descr->skb = skb;
hwdescr->buf_addr = buf;
hwdescr->buf_size = skb->len;
hwdescr->next_descr_addr = 0;
hwdescr->data_status = 0;
hwdescr->dmac_cmd_status =
SPIDER_NET_DESCR_CARDOWNED | SPIDER_NET_DMAC_TXFRMTL;
spin_unlock_irqrestore(&chain->lock, flags);
if (skb->ip_summed == CHECKSUM_PARTIAL)
switch (ip_hdr(skb)->protocol) {
case IPPROTO_TCP:
hwdescr->dmac_cmd_status |= SPIDER_NET_DMAC_TCP;
break;
case IPPROTO_UDP:
hwdescr->dmac_cmd_status |= SPIDER_NET_DMAC_UDP;
break;
}
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 21:11:33 +00:00
/* Chain the bus address, so that the DMA engine finds this descr. */
wmb();
descr->prev->hwdescr->next_descr_addr = descr->bus_addr;
card->netdev->trans_start = jiffies; /* set netdev watchdog timer */
return 0;
}
static int
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 21:11:33 +00:00
spider_net_set_low_watermark(struct spider_net_card *card)
{
struct spider_net_descr *descr = card->tx_chain.tail;
struct spider_net_hw_descr *hwdescr;
unsigned long flags;
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 21:11:33 +00:00
int status;
int cnt=0;
int i;
/* Measure the length of the queue. Measurement does not
* need to be precise -- does not need a lock. */
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 21:11:33 +00:00
while (descr != card->tx_chain.head) {
status = descr->hwdescr->dmac_cmd_status & SPIDER_NET_DESCR_NOT_IN_USE;
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 21:11:33 +00:00
if (status == SPIDER_NET_DESCR_NOT_IN_USE)
break;
descr = descr->next;
cnt++;
}
/* If TX queue is short, don't even bother with interrupts */
if (cnt < card->tx_chain.num_desc/4)
return cnt;
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 21:11:33 +00:00
/* Set low-watermark 3/4th's of the way into the queue. */
descr = card->tx_chain.tail;
cnt = (cnt*3)/4;
for (i=0;i<cnt; i++)
descr = descr->next;
/* Set the new watermark, clear the old watermark */
spin_lock_irqsave(&card->tx_chain.lock, flags);
descr->hwdescr->dmac_cmd_status |= SPIDER_NET_DESCR_TXDESFLG;
if (card->low_watermark && card->low_watermark != descr) {
hwdescr = card->low_watermark->hwdescr;
hwdescr->dmac_cmd_status =
hwdescr->dmac_cmd_status & ~SPIDER_NET_DESCR_TXDESFLG;
}
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 21:11:33 +00:00
card->low_watermark = descr;
spin_unlock_irqrestore(&card->tx_chain.lock, flags);
return cnt;
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 21:11:33 +00:00
}
/**
* spider_net_release_tx_chain - processes sent tx descriptors
* @card: adapter structure
* @brutal: if set, don't care about whether descriptor seems to be in use
*
* returns 0 if the tx ring is empty, otherwise 1.
*
* spider_net_release_tx_chain releases the tx descriptors that spider has
* finished with (if non-brutal) or simply release tx descriptors (if brutal).
* If some other context is calling this function, we return 1 so that we're
* scheduled again (if we were scheduled) and will not lose initiative.
*/
static int
spider_net_release_tx_chain(struct spider_net_card *card, int brutal)
{
struct net_device *dev = card->netdev;
struct spider_net_descr_chain *chain = &card->tx_chain;
struct spider_net_descr *descr;
struct spider_net_hw_descr *hwdescr;
struct sk_buff *skb;
u32 buf_addr;
unsigned long flags;
int status;
while (1) {
spin_lock_irqsave(&chain->lock, flags);
if (chain->tail == chain->head) {
spin_unlock_irqrestore(&chain->lock, flags);
return 0;
}
descr = chain->tail;
hwdescr = descr->hwdescr;
status = spider_net_get_descr_status(hwdescr);
switch (status) {
case SPIDER_NET_DESCR_COMPLETE:
dev->stats.tx_packets++;
dev->stats.tx_bytes += descr->skb->len;
break;
case SPIDER_NET_DESCR_CARDOWNED:
if (!brutal) {
spin_unlock_irqrestore(&chain->lock, flags);
return 1;
}
/* fallthrough, if we release the descriptors
* brutally (then we don't care about
* SPIDER_NET_DESCR_CARDOWNED) */
case SPIDER_NET_DESCR_RESPONSE_ERROR:
case SPIDER_NET_DESCR_PROTECTION_ERROR:
case SPIDER_NET_DESCR_FORCE_END:
if (netif_msg_tx_err(card))
dev_err(&card->netdev->dev, "forcing end of tx descriptor "
"with status x%02x\n", status);
dev->stats.tx_errors++;
break;
default:
dev->stats.tx_dropped++;
if (!brutal) {
spin_unlock_irqrestore(&chain->lock, flags);
return 1;
}
}
chain->tail = descr->next;
hwdescr->dmac_cmd_status |= SPIDER_NET_DESCR_NOT_IN_USE;
skb = descr->skb;
descr->skb = NULL;
buf_addr = hwdescr->buf_addr;
spin_unlock_irqrestore(&chain->lock, flags);
/* unmap the skb */
if (skb) {
pci_unmap_single(card->pdev, buf_addr, skb->len,
PCI_DMA_TODEVICE);
dev_kfree_skb(skb);
}
}
return 0;
}
/**
* spider_net_kick_tx_dma - enables TX DMA processing
* @card: card structure
*
* This routine will start the transmit DMA running if
* it is not already running. This routine ned only be
* called when queueing a new packet to an empty tx queue.
* Writes the current tx chain head as start address
* of the tx descriptor chain and enables the transmission
* DMA engine.
*/
static inline void
spider_net_kick_tx_dma(struct spider_net_card *card)
{
struct spider_net_descr *descr;
if (spider_net_read_reg(card, SPIDER_NET_GDTDMACCNTR) &
SPIDER_NET_TX_DMA_EN)
goto out;
descr = card->tx_chain.tail;
for (;;) {
if (spider_net_get_descr_status(descr->hwdescr) ==
SPIDER_NET_DESCR_CARDOWNED) {
spider_net_write_reg(card, SPIDER_NET_GDTDCHA,
descr->bus_addr);
spider_net_write_reg(card, SPIDER_NET_GDTDMACCNTR,
SPIDER_NET_DMA_TX_VALUE);
break;
}
if (descr == card->tx_chain.head)
break;
descr = descr->next;
}
out:
mod_timer(&card->tx_timer, jiffies + SPIDER_NET_TX_TIMER);
}
/**
* spider_net_xmit - transmits a frame over the device
* @skb: packet to send out
* @netdev: interface device structure
*
* returns 0 on success, !0 on failure
*/
static int
spider_net_xmit(struct sk_buff *skb, struct net_device *netdev)
{
int cnt;
struct spider_net_card *card = netdev_priv(netdev);
spider_net_release_tx_chain(card, 0);
if (spider_net_prepare_tx_descr(card, skb) != 0) {
netdev->stats.tx_dropped++;
netif_stop_queue(netdev);
return NETDEV_TX_BUSY;
}
cnt = spider_net_set_low_watermark(card);
if (cnt < 5)
spider_net_kick_tx_dma(card);
return NETDEV_TX_OK;
}
/**
* spider_net_cleanup_tx_ring - cleans up the TX ring
* @card: card structure
*
* spider_net_cleanup_tx_ring is called by either the tx_timer
* or from the NAPI polling routine.
* This routine releases resources associted with transmitted
* packets, including updating the queue tail pointer.
*/
static void
spider_net_cleanup_tx_ring(struct spider_net_card *card)
{
if ((spider_net_release_tx_chain(card, 0) != 0) &&
(card->netdev->flags & IFF_UP)) {
spider_net_kick_tx_dma(card);
netif_wake_queue(card->netdev);
}
}
/**
* spider_net_do_ioctl - called for device ioctls
* @netdev: interface device structure
* @ifr: request parameter structure for ioctl
* @cmd: command code for ioctl
*
* returns 0 on success, <0 on failure. Currently, we have no special ioctls.
* -EOPNOTSUPP is returned, if an unknown ioctl was requested
*/
static int
spider_net_do_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
{
switch (cmd) {
default:
return -EOPNOTSUPP;
}
}
/**
* spider_net_pass_skb_up - takes an skb from a descriptor and passes it on
* @descr: descriptor to process
* @card: card structure
*
* Fills out skb structure and passes the data to the stack.
* The descriptor state is not changed.
*/
static void
spider_net_pass_skb_up(struct spider_net_descr *descr,
struct spider_net_card *card)
{
struct spider_net_hw_descr *hwdescr = descr->hwdescr;
struct sk_buff *skb = descr->skb;
struct net_device *netdev = card->netdev;
u32 data_status = hwdescr->data_status;
u32 data_error = hwdescr->data_error;
skb_put(skb, hwdescr->valid_size);
/* the card seems to add 2 bytes of junk in front
* of the ethernet frame */
#define SPIDER_MISALIGN 2
skb_pull(skb, SPIDER_MISALIGN);
skb->protocol = eth_type_trans(skb, netdev);
/* checksum offload */
skb_checksum_none_assert(skb);
if (netdev->features & NETIF_F_RXCSUM) {
if ( ( (data_status & SPIDER_NET_DATA_STATUS_CKSUM_MASK) ==
SPIDER_NET_DATA_STATUS_CKSUM_MASK) &&
!(data_error & SPIDER_NET_DATA_ERR_CKSUM_MASK))
skb->ip_summed = CHECKSUM_UNNECESSARY;
}
if (data_status & SPIDER_NET_VLAN_PACKET) {
/* further enhancements: HW-accel VLAN */
}
/* update netdevice statistics */
netdev->stats.rx_packets++;
netdev->stats.rx_bytes += skb->len;
/* pass skb up to stack */
netif_receive_skb(skb);
}
static void show_rx_chain(struct spider_net_card *card)
{
struct spider_net_descr_chain *chain = &card->rx_chain;
struct spider_net_descr *start= chain->tail;
struct spider_net_descr *descr= start;
struct spider_net_hw_descr *hwd = start->hwdescr;
struct device *dev = &card->netdev->dev;
u32 curr_desc, next_desc;
int status;
int tot = 0;
int cnt = 0;
int off = start - chain->ring;
int cstat = hwd->dmac_cmd_status;
dev_info(dev, "Total number of descrs=%d\n",
chain->num_desc);
dev_info(dev, "Chain tail located at descr=%d, status=0x%x\n",
off, cstat);
curr_desc = spider_net_read_reg(card, SPIDER_NET_GDACTDPA);
next_desc = spider_net_read_reg(card, SPIDER_NET_GDACNEXTDA);
status = cstat;
do
{
hwd = descr->hwdescr;
off = descr - chain->ring;
status = hwd->dmac_cmd_status;
if (descr == chain->head)
dev_info(dev, "Chain head is at %d, head status=0x%x\n",
off, status);
if (curr_desc == descr->bus_addr)
dev_info(dev, "HW curr desc (GDACTDPA) is at %d, status=0x%x\n",
off, status);
if (next_desc == descr->bus_addr)
dev_info(dev, "HW next desc (GDACNEXTDA) is at %d, status=0x%x\n",
off, status);
if (hwd->next_descr_addr == 0)
dev_info(dev, "chain is cut at %d\n", off);
if (cstat != status) {
int from = (chain->num_desc + off - cnt) % chain->num_desc;
int to = (chain->num_desc + off - 1) % chain->num_desc;
dev_info(dev, "Have %d (from %d to %d) descrs "
"with stat=0x%08x\n", cnt, from, to, cstat);
cstat = status;
cnt = 0;
}
cnt ++;
tot ++;
descr = descr->next;
} while (descr != start);
dev_info(dev, "Last %d descrs with stat=0x%08x "
"for a total of %d descrs\n", cnt, cstat, tot);
#ifdef DEBUG
/* Now dump the whole ring */
descr = start;
do
{
struct spider_net_hw_descr *hwd = descr->hwdescr;
status = spider_net_get_descr_status(hwd);
cnt = descr - chain->ring;
dev_info(dev, "Descr %d stat=0x%08x skb=%p\n",
cnt, status, descr->skb);
dev_info(dev, "bus addr=%08x buf addr=%08x sz=%d\n",
descr->bus_addr, hwd->buf_addr, hwd->buf_size);
dev_info(dev, "next=%08x result sz=%d valid sz=%d\n",
hwd->next_descr_addr, hwd->result_size,
hwd->valid_size);
dev_info(dev, "dmac=%08x data stat=%08x data err=%08x\n",
hwd->dmac_cmd_status, hwd->data_status,
hwd->data_error);
dev_info(dev, "\n");
descr = descr->next;
} while (descr != start);
#endif
}
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 18:21:13 +00:00
/**
* spider_net_resync_head_ptr - Advance head ptr past empty descrs
*
* If the driver fails to keep up and empty the queue, then the
* hardware wil run out of room to put incoming packets. This
* will cause the hardware to skip descrs that are full (instead
* of halting/retrying). Thus, once the driver runs, it wil need
* to "catch up" to where the hardware chain pointer is at.
*/
static void spider_net_resync_head_ptr(struct spider_net_card *card)
{
unsigned long flags;
struct spider_net_descr_chain *chain = &card->rx_chain;
struct spider_net_descr *descr;
int i, status;
/* Advance head pointer past any empty descrs */
descr = chain->head;
status = spider_net_get_descr_status(descr->hwdescr);
if (status == SPIDER_NET_DESCR_NOT_IN_USE)
return;
spin_lock_irqsave(&chain->lock, flags);
descr = chain->head;
status = spider_net_get_descr_status(descr->hwdescr);
for (i=0; i<chain->num_desc; i++) {
if (status != SPIDER_NET_DESCR_CARDOWNED) break;
descr = descr->next;
status = spider_net_get_descr_status(descr->hwdescr);
}
chain->head = descr;
spin_unlock_irqrestore(&chain->lock, flags);
}
static int spider_net_resync_tail_ptr(struct spider_net_card *card)
{
struct spider_net_descr_chain *chain = &card->rx_chain;
struct spider_net_descr *descr;
int i, status;
/* Advance tail pointer past any empty and reaped descrs */
descr = chain->tail;
status = spider_net_get_descr_status(descr->hwdescr);
for (i=0; i<chain->num_desc; i++) {
if ((status != SPIDER_NET_DESCR_CARDOWNED) &&
(status != SPIDER_NET_DESCR_NOT_IN_USE)) break;
descr = descr->next;
status = spider_net_get_descr_status(descr->hwdescr);
}
chain->tail = descr;
if ((i == chain->num_desc) || (i == 0))
return 1;
return 0;
}
/**
* spider_net_decode_one_descr - processes an RX descriptor
* @card: card structure
*
* Returns 1 if a packet has been sent to the stack, otherwise 0.
*
* Processes an RX descriptor by iommu-unmapping the data buffer
* and passing the packet up to the stack. This function is called
* in softirq context, e.g. either bottom half from interrupt or
* NAPI polling context.
*/
static int
spider_net_decode_one_descr(struct spider_net_card *card)
{
struct net_device *dev = card->netdev;
struct spider_net_descr_chain *chain = &card->rx_chain;
struct spider_net_descr *descr = chain->tail;
struct spider_net_hw_descr *hwdescr = descr->hwdescr;
u32 hw_buf_addr;
int status;
status = spider_net_get_descr_status(hwdescr);
/* Nothing in the descriptor, or ring must be empty */
if ((status == SPIDER_NET_DESCR_CARDOWNED) ||
(status == SPIDER_NET_DESCR_NOT_IN_USE))
return 0;
/* descriptor definitively used -- move on tail */
chain->tail = descr->next;
/* unmap descriptor */
hw_buf_addr = hwdescr->buf_addr;
hwdescr->buf_addr = 0xffffffff;
pci_unmap_single(card->pdev, hw_buf_addr,
SPIDER_NET_MAX_FRAME, PCI_DMA_FROMDEVICE);
if ( (status == SPIDER_NET_DESCR_RESPONSE_ERROR) ||
(status == SPIDER_NET_DESCR_PROTECTION_ERROR) ||
(status == SPIDER_NET_DESCR_FORCE_END) ) {
if (netif_msg_rx_err(card))
dev_err(&dev->dev,
"dropping RX descriptor with state %d\n", status);
dev->stats.rx_dropped++;
goto bad_desc;
}
if ( (status != SPIDER_NET_DESCR_COMPLETE) &&
(status != SPIDER_NET_DESCR_FRAME_END) ) {
if (netif_msg_rx_err(card))
dev_err(&card->netdev->dev,
"RX descriptor with unknown state %d\n", status);
card->spider_stats.rx_desc_unk_state++;
goto bad_desc;
}
/* The cases we'll throw away the packet immediately */
if (hwdescr->data_error & SPIDER_NET_DESTROY_RX_FLAGS) {
if (netif_msg_rx_err(card))
dev_err(&card->netdev->dev,
"error in received descriptor found, "
"data_status=x%08x, data_error=x%08x\n",
hwdescr->data_status, hwdescr->data_error);
goto bad_desc;
}
if (hwdescr->dmac_cmd_status & SPIDER_NET_DESCR_BAD_STATUS) {
dev_err(&card->netdev->dev, "bad status, cmd_status=x%08x\n",
hwdescr->dmac_cmd_status);
pr_err("buf_addr=x%08x\n", hw_buf_addr);
pr_err("buf_size=x%08x\n", hwdescr->buf_size);
pr_err("next_descr_addr=x%08x\n", hwdescr->next_descr_addr);
pr_err("result_size=x%08x\n", hwdescr->result_size);
pr_err("valid_size=x%08x\n", hwdescr->valid_size);
pr_err("data_status=x%08x\n", hwdescr->data_status);
pr_err("data_error=x%08x\n", hwdescr->data_error);
pr_err("which=%ld\n", descr - card->rx_chain.ring);
card->spider_stats.rx_desc_error++;
goto bad_desc;
}
/* Ok, we've got a packet in descr */
spider_net_pass_skb_up(descr, card);
descr->skb = NULL;
hwdescr->dmac_cmd_status = SPIDER_NET_DESCR_NOT_IN_USE;
return 1;
bad_desc:
if (netif_msg_rx_err(card))
show_rx_chain(card);
dev_kfree_skb_irq(descr->skb);
descr->skb = NULL;
hwdescr->dmac_cmd_status = SPIDER_NET_DESCR_NOT_IN_USE;
return 0;
}
/**
* spider_net_poll - NAPI poll function called by the stack to return packets
* @netdev: interface device structure
* @budget: number of packets we can pass to the stack at most
*
* returns 0 if no more packets available to the driver/stack. Returns 1,
* if the quota is exceeded, but the driver has still packets.
*
* spider_net_poll returns all packets from the rx descriptors to the stack
* (using netif_receive_skb). If all/enough packets are up, the driver
* reenables interrupts and returns 0. If not, 1 is returned.
*/
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 23:41:36 +00:00
static int spider_net_poll(struct napi_struct *napi, int budget)
{
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 23:41:36 +00:00
struct spider_net_card *card = container_of(napi, struct spider_net_card, napi);
int packets_done = 0;
while (packets_done < budget) {
if (!spider_net_decode_one_descr(card))
break;
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 23:41:36 +00:00
packets_done++;
}
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 18:21:13 +00:00
if ((packets_done == 0) && (card->num_rx_ints != 0)) {
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 23:41:36 +00:00
if (!spider_net_resync_tail_ptr(card))
packets_done = budget;
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 18:21:13 +00:00
spider_net_resync_head_ptr(card);
}
card->num_rx_ints = 0;
spider_net_refill_rx_chain(card);
spider_net_enable_rxdmac(card);
spider_net_cleanup_tx_ring(card);
/* if all packets are in the stack, enable interrupts and return 0 */
/* if not, return 1 */
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 23:41:36 +00:00
if (packets_done < budget) {
napi_complete(napi);
spider_net_rx_irq_on(card);
card->ignore_rx_ramfull = 0;
}
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 23:41:36 +00:00
return packets_done;
}
/**
* spider_net_change_mtu - changes the MTU of an interface
* @netdev: interface device structure
* @new_mtu: new MTU value
*
* returns 0 on success, <0 on failure
*/
static int
spider_net_change_mtu(struct net_device *netdev, int new_mtu)
{
/* no need to re-alloc skbs or so -- the max mtu is about 2.3k
* and mtu is outbound only anyway */
if ( (new_mtu < SPIDER_NET_MIN_MTU ) ||
(new_mtu > SPIDER_NET_MAX_MTU) )
return -EINVAL;
netdev->mtu = new_mtu;
return 0;
}
/**
* spider_net_set_mac - sets the MAC of an interface
* @netdev: interface device structure
* @ptr: pointer to new MAC address
*
* Returns 0 on success, <0 on failure. Currently, we don't support this
* and will always return EOPNOTSUPP.
*/
static int
spider_net_set_mac(struct net_device *netdev, void *p)
{
struct spider_net_card *card = netdev_priv(netdev);
u32 macl, macu, regvalue;
struct sockaddr *addr = p;
if (!is_valid_ether_addr(addr->sa_data))
return -EADDRNOTAVAIL;
/* switch off GMACTPE and GMACRPE */
regvalue = spider_net_read_reg(card, SPIDER_NET_GMACOPEMD);
regvalue &= ~((1 << 5) | (1 << 6));
spider_net_write_reg(card, SPIDER_NET_GMACOPEMD, regvalue);
/* write mac */
macu = (addr->sa_data[0]<<24) + (addr->sa_data[1]<<16) +
(addr->sa_data[2]<<8) + (addr->sa_data[3]);
macl = (addr->sa_data[4]<<8) + (addr->sa_data[5]);
spider_net_write_reg(card, SPIDER_NET_GMACUNIMACU, macu);
spider_net_write_reg(card, SPIDER_NET_GMACUNIMACL, macl);
/* switch GMACTPE and GMACRPE back on */
regvalue = spider_net_read_reg(card, SPIDER_NET_GMACOPEMD);
regvalue |= ((1 << 5) | (1 << 6));
spider_net_write_reg(card, SPIDER_NET_GMACOPEMD, regvalue);
spider_net_set_promisc(card);
/* look up, whether we have been successful */
if (spider_net_get_mac_address(netdev))
return -EADDRNOTAVAIL;
if (memcmp(netdev->dev_addr,addr->sa_data,netdev->addr_len))
return -EADDRNOTAVAIL;
return 0;
}
/**
* spider_net_link_reset
* @netdev: net device structure
*
* This is called when the PHY_LINK signal is asserted. For the blade this is
* not connected so we should never get here.
*
*/
static void
spider_net_link_reset(struct net_device *netdev)
{
struct spider_net_card *card = netdev_priv(netdev);
del_timer_sync(&card->aneg_timer);
/* clear interrupt, block further interrupts */
spider_net_write_reg(card, SPIDER_NET_GMACST,
spider_net_read_reg(card, SPIDER_NET_GMACST));
spider_net_write_reg(card, SPIDER_NET_GMACINTEN, 0);
/* reset phy and setup aneg */
card->aneg_count = 0;
card->medium = BCM54XX_COPPER;
spider_net_setup_aneg(card);
mod_timer(&card->aneg_timer, jiffies + SPIDER_NET_ANEG_TIMER);
}
/**
* spider_net_handle_error_irq - handles errors raised by an interrupt
* @card: card structure
* @status_reg: interrupt status register 0 (GHIINT0STS)
*
* spider_net_handle_error_irq treats or ignores all error conditions
* found when an interrupt is presented
*/
static void
spider_net_handle_error_irq(struct spider_net_card *card, u32 status_reg,
u32 error_reg1, u32 error_reg2)
{
u32 i;
int show_error = 1;
/* check GHIINT0STS ************************************/
if (status_reg)
for (i = 0; i < 32; i++)
if (status_reg & (1<<i))
switch (i)
{
/* let error_reg1 and error_reg2 evaluation decide, what to do
case SPIDER_NET_PHYINT:
case SPIDER_NET_GMAC2INT:
case SPIDER_NET_GMAC1INT:
case SPIDER_NET_GFIFOINT:
case SPIDER_NET_DMACINT:
case SPIDER_NET_GSYSINT:
break; */
case SPIDER_NET_GIPSINT:
show_error = 0;
break;
case SPIDER_NET_GPWOPCMPINT:
/* PHY write operation completed */
show_error = 0;
break;
case SPIDER_NET_GPROPCMPINT:
/* PHY read operation completed */
/* we don't use semaphores, as we poll for the completion
* of the read operation in spider_net_read_phy. Should take
* about 50 us */
show_error = 0;
break;
case SPIDER_NET_GPWFFINT:
/* PHY command queue full */
if (netif_msg_intr(card))
dev_err(&card->netdev->dev, "PHY write queue full\n");
show_error = 0;
break;
/* case SPIDER_NET_GRMDADRINT: not used. print a message */
/* case SPIDER_NET_GRMARPINT: not used. print a message */
/* case SPIDER_NET_GRMMPINT: not used. print a message */
case SPIDER_NET_GDTDEN0INT:
/* someone has set TX_DMA_EN to 0 */
show_error = 0;
break;
case SPIDER_NET_GDDDEN0INT: /* fallthrough */
case SPIDER_NET_GDCDEN0INT: /* fallthrough */
case SPIDER_NET_GDBDEN0INT: /* fallthrough */
case SPIDER_NET_GDADEN0INT:
/* someone has set RX_DMA_EN to 0 */
show_error = 0;
break;
/* RX interrupts */
case SPIDER_NET_GDDFDCINT:
case SPIDER_NET_GDCFDCINT:
case SPIDER_NET_GDBFDCINT:
case SPIDER_NET_GDAFDCINT:
/* case SPIDER_NET_GDNMINT: not used. print a message */
/* case SPIDER_NET_GCNMINT: not used. print a message */
/* case SPIDER_NET_GBNMINT: not used. print a message */
/* case SPIDER_NET_GANMINT: not used. print a message */
/* case SPIDER_NET_GRFNMINT: not used. print a message */
show_error = 0;
break;
/* TX interrupts */
case SPIDER_NET_GDTFDCINT:
show_error = 0;
break;
case SPIDER_NET_GTTEDINT:
show_error = 0;
break;
case SPIDER_NET_GDTDCEINT:
/* chain end. If a descriptor should be sent, kick off
* tx dma
if (card->tx_chain.tail != card->tx_chain.head)
spider_net_kick_tx_dma(card);
*/
show_error = 0;
break;
/* case SPIDER_NET_G1TMCNTINT: not used. print a message */
/* case SPIDER_NET_GFREECNTINT: not used. print a message */
}
/* check GHIINT1STS ************************************/
if (error_reg1)
for (i = 0; i < 32; i++)
if (error_reg1 & (1<<i))
switch (i)
{
case SPIDER_NET_GTMFLLINT:
/* TX RAM full may happen on a usual case.
* Logging is not needed. */
show_error = 0;
break;
case SPIDER_NET_GRFDFLLINT: /* fallthrough */
case SPIDER_NET_GRFCFLLINT: /* fallthrough */
case SPIDER_NET_GRFBFLLINT: /* fallthrough */
case SPIDER_NET_GRFAFLLINT: /* fallthrough */
case SPIDER_NET_GRMFLLINT:
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 18:21:13 +00:00
/* Could happen when rx chain is full */
if (card->ignore_rx_ramfull == 0) {
card->ignore_rx_ramfull = 1;
spider_net_resync_head_ptr(card);
spider_net_refill_rx_chain(card);
spider_net_enable_rxdmac(card);
card->num_rx_ints ++;
napi_schedule(&card->napi);
}
show_error = 0;
break;
/* case SPIDER_NET_GTMSHTINT: problem, print a message */
case SPIDER_NET_GDTINVDINT:
/* allrighty. tx from previous descr ok */
show_error = 0;
break;
/* chain end */
case SPIDER_NET_GDDDCEINT: /* fallthrough */
case SPIDER_NET_GDCDCEINT: /* fallthrough */
case SPIDER_NET_GDBDCEINT: /* fallthrough */
case SPIDER_NET_GDADCEINT:
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 18:21:13 +00:00
spider_net_resync_head_ptr(card);
spider_net_refill_rx_chain(card);
spider_net_enable_rxdmac(card);
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 18:21:13 +00:00
card->num_rx_ints ++;
napi_schedule(&card->napi);
show_error = 0;
break;
/* invalid descriptor */
case SPIDER_NET_GDDINVDINT: /* fallthrough */
case SPIDER_NET_GDCINVDINT: /* fallthrough */
case SPIDER_NET_GDBINVDINT: /* fallthrough */
case SPIDER_NET_GDAINVDINT:
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 18:21:13 +00:00
/* Could happen when rx chain is full */
spider_net_resync_head_ptr(card);
spider_net_refill_rx_chain(card);
spider_net_enable_rxdmac(card);
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 18:21:13 +00:00
card->num_rx_ints ++;
napi_schedule(&card->napi);
show_error = 0;
break;
/* case SPIDER_NET_GDTRSERINT: problem, print a message */
/* case SPIDER_NET_GDDRSERINT: problem, print a message */
/* case SPIDER_NET_GDCRSERINT: problem, print a message */
/* case SPIDER_NET_GDBRSERINT: problem, print a message */
/* case SPIDER_NET_GDARSERINT: problem, print a message */
/* case SPIDER_NET_GDSERINT: problem, print a message */
/* case SPIDER_NET_GDTPTERINT: problem, print a message */
/* case SPIDER_NET_GDDPTERINT: problem, print a message */
/* case SPIDER_NET_GDCPTERINT: problem, print a message */
/* case SPIDER_NET_GDBPTERINT: problem, print a message */
/* case SPIDER_NET_GDAPTERINT: problem, print a message */
default:
show_error = 1;
break;
}
/* check GHIINT2STS ************************************/
if (error_reg2)
for (i = 0; i < 32; i++)
if (error_reg2 & (1<<i))
switch (i)
{
/* there is nothing we can (want to) do at this time. Log a
* message, we can switch on and off the specific values later on
case SPIDER_NET_GPROPERINT:
case SPIDER_NET_GMCTCRSNGINT:
case SPIDER_NET_GMCTLCOLINT:
case SPIDER_NET_GMCTTMOTINT:
case SPIDER_NET_GMCRCAERINT:
case SPIDER_NET_GMCRCALERINT:
case SPIDER_NET_GMCRALNERINT:
case SPIDER_NET_GMCROVRINT:
case SPIDER_NET_GMCRRNTINT:
case SPIDER_NET_GMCRRXERINT:
case SPIDER_NET_GTITCSERINT:
case SPIDER_NET_GTIFMTERINT:
case SPIDER_NET_GTIPKTRVKINT:
case SPIDER_NET_GTISPINGINT:
case SPIDER_NET_GTISADNGINT:
case SPIDER_NET_GTISPDNGINT:
case SPIDER_NET_GRIFMTERINT:
case SPIDER_NET_GRIPKTRVKINT:
case SPIDER_NET_GRISPINGINT:
case SPIDER_NET_GRISADNGINT:
case SPIDER_NET_GRISPDNGINT:
break;
*/
default:
break;
}
if ((show_error) && (netif_msg_intr(card)) && net_ratelimit())
dev_err(&card->netdev->dev, "Error interrupt, GHIINT0STS = 0x%08x, "
"GHIINT1STS = 0x%08x, GHIINT2STS = 0x%08x\n",
status_reg, error_reg1, error_reg2);
/* clear interrupt sources */
spider_net_write_reg(card, SPIDER_NET_GHIINT1STS, error_reg1);
spider_net_write_reg(card, SPIDER_NET_GHIINT2STS, error_reg2);
}
/**
* spider_net_interrupt - interrupt handler for spider_net
* @irq: interrupt number
* @ptr: pointer to net_device
*
* returns IRQ_HANDLED, if interrupt was for driver, or IRQ_NONE, if no
* interrupt found raised by card.
*
* This is the interrupt handler, that turns off
* interrupts for this device and makes the stack poll the driver
*/
static irqreturn_t
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
spider_net_interrupt(int irq, void *ptr)
{
struct net_device *netdev = ptr;
struct spider_net_card *card = netdev_priv(netdev);
u32 status_reg, error_reg1, error_reg2;
status_reg = spider_net_read_reg(card, SPIDER_NET_GHIINT0STS);
error_reg1 = spider_net_read_reg(card, SPIDER_NET_GHIINT1STS);
error_reg2 = spider_net_read_reg(card, SPIDER_NET_GHIINT2STS);
if (!(status_reg & SPIDER_NET_INT0_MASK_VALUE) &&
!(error_reg1 & SPIDER_NET_INT1_MASK_VALUE) &&
!(error_reg2 & SPIDER_NET_INT2_MASK_VALUE))
return IRQ_NONE;
if (status_reg & SPIDER_NET_RXINT ) {
spider_net_rx_irq_off(card);
napi_schedule(&card->napi);
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 18:21:13 +00:00
card->num_rx_ints ++;
}
if (status_reg & SPIDER_NET_TXINT)
napi_schedule(&card->napi);
if (status_reg & SPIDER_NET_LINKINT)
spider_net_link_reset(netdev);
if (status_reg & SPIDER_NET_ERRINT )
spider_net_handle_error_irq(card, status_reg,
error_reg1, error_reg2);
/* clear interrupt sources */
spider_net_write_reg(card, SPIDER_NET_GHIINT0STS, status_reg);
return IRQ_HANDLED;
}
#ifdef CONFIG_NET_POLL_CONTROLLER
/**
* spider_net_poll_controller - artificial interrupt for netconsole etc.
* @netdev: interface device structure
*
* see Documentation/networking/netconsole.txt
*/
static void
spider_net_poll_controller(struct net_device *netdev)
{
disable_irq(netdev->irq);
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 13:55:46 +00:00
spider_net_interrupt(netdev->irq, netdev);
enable_irq(netdev->irq);
}
#endif /* CONFIG_NET_POLL_CONTROLLER */
/**
* spider_net_enable_interrupts - enable interrupts
* @card: card structure
*
* spider_net_enable_interrupt enables several interrupts
*/
static void
spider_net_enable_interrupts(struct spider_net_card *card)
{
spider_net_write_reg(card, SPIDER_NET_GHIINT0MSK,
SPIDER_NET_INT0_MASK_VALUE);
spider_net_write_reg(card, SPIDER_NET_GHIINT1MSK,
SPIDER_NET_INT1_MASK_VALUE);
spider_net_write_reg(card, SPIDER_NET_GHIINT2MSK,
SPIDER_NET_INT2_MASK_VALUE);
}
/**
* spider_net_disable_interrupts - disable interrupts
* @card: card structure
*
* spider_net_disable_interrupts disables all the interrupts
*/
static void
spider_net_disable_interrupts(struct spider_net_card *card)
{
spider_net_write_reg(card, SPIDER_NET_GHIINT0MSK, 0);
spider_net_write_reg(card, SPIDER_NET_GHIINT1MSK, 0);
spider_net_write_reg(card, SPIDER_NET_GHIINT2MSK, 0);
spider_net_write_reg(card, SPIDER_NET_GMACINTEN, 0);
}
/**
* spider_net_init_card - initializes the card
* @card: card structure
*
* spider_net_init_card initializes the card so that other registers can
* be used
*/
static void
spider_net_init_card(struct spider_net_card *card)
{
spider_net_write_reg(card, SPIDER_NET_CKRCTRL,
SPIDER_NET_CKRCTRL_STOP_VALUE);
spider_net_write_reg(card, SPIDER_NET_CKRCTRL,
SPIDER_NET_CKRCTRL_RUN_VALUE);
/* trigger ETOMOD signal */
spider_net_write_reg(card, SPIDER_NET_GMACOPEMD,
spider_net_read_reg(card, SPIDER_NET_GMACOPEMD) | 0x4);
spider_net_disable_interrupts(card);
}
/**
* spider_net_enable_card - enables the card by setting all kinds of regs
* @card: card structure
*
* spider_net_enable_card sets a lot of SMMIO registers to enable the device
*/
static void
spider_net_enable_card(struct spider_net_card *card)
{
int i;
/* the following array consists of (register),(value) pairs
* that are set in this function. A register of 0 ends the list */
u32 regs[][2] = {
{ SPIDER_NET_GRESUMINTNUM, 0 },
{ SPIDER_NET_GREINTNUM, 0 },
/* set interrupt frame number registers */
/* clear the single DMA engine registers first */
{ SPIDER_NET_GFAFRMNUM, SPIDER_NET_GFXFRAMES_VALUE },
{ SPIDER_NET_GFBFRMNUM, SPIDER_NET_GFXFRAMES_VALUE },
{ SPIDER_NET_GFCFRMNUM, SPIDER_NET_GFXFRAMES_VALUE },
{ SPIDER_NET_GFDFRMNUM, SPIDER_NET_GFXFRAMES_VALUE },
/* then set, what we really need */
{ SPIDER_NET_GFFRMNUM, SPIDER_NET_FRAMENUM_VALUE },
/* timer counter registers and stuff */
{ SPIDER_NET_GFREECNNUM, 0 },
{ SPIDER_NET_GONETIMENUM, 0 },
{ SPIDER_NET_GTOUTFRMNUM, 0 },
/* RX mode setting */
{ SPIDER_NET_GRXMDSET, SPIDER_NET_RXMODE_VALUE },
/* TX mode setting */
{ SPIDER_NET_GTXMDSET, SPIDER_NET_TXMODE_VALUE },
/* IPSEC mode setting */
{ SPIDER_NET_GIPSECINIT, SPIDER_NET_IPSECINIT_VALUE },
{ SPIDER_NET_GFTRESTRT, SPIDER_NET_RESTART_VALUE },
{ SPIDER_NET_GMRWOLCTRL, 0 },
{ SPIDER_NET_GTESTMD, 0x10000000 },
{ SPIDER_NET_GTTQMSK, 0x00400040 },
{ SPIDER_NET_GMACINTEN, 0 },
/* flow control stuff */
{ SPIDER_NET_GMACAPAUSE, SPIDER_NET_MACAPAUSE_VALUE },
{ SPIDER_NET_GMACTXPAUSE, SPIDER_NET_TXPAUSE_VALUE },
{ SPIDER_NET_GMACBSTLMT, SPIDER_NET_BURSTLMT_VALUE },
{ 0, 0}
};
i = 0;
while (regs[i][0]) {
spider_net_write_reg(card, regs[i][0], regs[i][1]);
i++;
}
/* clear unicast filter table entries 1 to 14 */
for (i = 1; i <= 14; i++) {
spider_net_write_reg(card,
SPIDER_NET_GMRUAFILnR + i * 8,
0x00080000);
spider_net_write_reg(card,
SPIDER_NET_GMRUAFILnR + i * 8 + 4,
0x00000000);
}
spider_net_write_reg(card, SPIDER_NET_GMRUA0FIL15R, 0x08080000);
spider_net_write_reg(card, SPIDER_NET_ECMODE, SPIDER_NET_ECMODE_VALUE);
/* set chain tail address for RX chains and
* enable DMA */
spider_net_enable_rxchtails(card);
spider_net_enable_rxdmac(card);
spider_net_write_reg(card, SPIDER_NET_GRXDMAEN, SPIDER_NET_WOL_VALUE);
spider_net_write_reg(card, SPIDER_NET_GMACLENLMT,
SPIDER_NET_LENLMT_VALUE);
spider_net_write_reg(card, SPIDER_NET_GMACOPEMD,
SPIDER_NET_OPMODE_VALUE);
spider_net_write_reg(card, SPIDER_NET_GDTDMACCNTR,
SPIDER_NET_GDTBSTA);
}
/**
* spider_net_download_firmware - loads firmware into the adapter
* @card: card structure
* @firmware_ptr: pointer to firmware data
*
* spider_net_download_firmware loads the firmware data into the
* adapter. It assumes the length etc. to be allright.
*/
static int
spider_net_download_firmware(struct spider_net_card *card,
const void *firmware_ptr)
{
int sequencer, i;
const u32 *fw_ptr = firmware_ptr;
/* stop sequencers */
spider_net_write_reg(card, SPIDER_NET_GSINIT,
SPIDER_NET_STOP_SEQ_VALUE);
for (sequencer = 0; sequencer < SPIDER_NET_FIRMWARE_SEQS;
sequencer++) {
spider_net_write_reg(card,
SPIDER_NET_GSnPRGADR + sequencer * 8, 0);
for (i = 0; i < SPIDER_NET_FIRMWARE_SEQWORDS; i++) {
spider_net_write_reg(card, SPIDER_NET_GSnPRGDAT +
sequencer * 8, *fw_ptr);
fw_ptr++;
}
}
if (spider_net_read_reg(card, SPIDER_NET_GSINIT))
return -EIO;
spider_net_write_reg(card, SPIDER_NET_GSINIT,
SPIDER_NET_RUN_SEQ_VALUE);
return 0;
}
/**
* spider_net_init_firmware - reads in firmware parts
* @card: card structure
*
* Returns 0 on success, <0 on failure
*
* spider_net_init_firmware opens the sequencer firmware and does some basic
* checks. This function opens and releases the firmware structure. A call
* to download the firmware is performed before the release.
*
* Firmware format
* ===============
* spider_fw.bin is expected to be a file containing 6*1024*4 bytes, 4k being
* the program for each sequencer. Use the command
* tail -q -n +2 Seq_code1_0x088.txt Seq_code2_0x090.txt \
* Seq_code3_0x098.txt Seq_code4_0x0A0.txt Seq_code5_0x0A8.txt \
* Seq_code6_0x0B0.txt | xxd -r -p -c4 > spider_fw.bin
*
* to generate spider_fw.bin, if you have sequencer programs with something
* like the following contents for each sequencer:
* <ONE LINE COMMENT>
* <FIRST 4-BYTES-WORD FOR SEQUENCER>
* <SECOND 4-BYTES-WORD FOR SEQUENCER>
* ...
* <1024th 4-BYTES-WORD FOR SEQUENCER>
*/
static int
spider_net_init_firmware(struct spider_net_card *card)
{
struct firmware *firmware = NULL;
struct device_node *dn;
const u8 *fw_prop = NULL;
int err = -ENOENT;
int fw_size;
if (request_firmware((const struct firmware **)&firmware,
SPIDER_NET_FIRMWARE_NAME, &card->pdev->dev) == 0) {
if ( (firmware->size != SPIDER_NET_FIRMWARE_LEN) &&
netif_msg_probe(card) ) {
dev_err(&card->netdev->dev,
"Incorrect size of spidernet firmware in " \
"filesystem. Looking in host firmware...\n");
goto try_host_fw;
}
err = spider_net_download_firmware(card, firmware->data);
release_firmware(firmware);
if (err)
goto try_host_fw;
goto done;
}
try_host_fw:
dn = pci_device_to_OF_node(card->pdev);
if (!dn)
goto out_err;
fw_prop = of_get_property(dn, "firmware", &fw_size);
if (!fw_prop)
goto out_err;
if ( (fw_size != SPIDER_NET_FIRMWARE_LEN) &&
netif_msg_probe(card) ) {
dev_err(&card->netdev->dev,
"Incorrect size of spidernet firmware in host firmware\n");
goto done;
}
err = spider_net_download_firmware(card, fw_prop);
done:
return err;
out_err:
if (netif_msg_probe(card))
dev_err(&card->netdev->dev,
"Couldn't find spidernet firmware in filesystem " \
"or host firmware\n");
return err;
}
/**
* spider_net_open - called upon ifonfig up
* @netdev: interface device structure
*
* returns 0 on success, <0 on failure
*
* spider_net_open allocates all the descriptors and memory needed for
* operation, sets up multicast list and enables interrupts
*/
int
spider_net_open(struct net_device *netdev)
{
struct spider_net_card *card = netdev_priv(netdev);
int result;
result = spider_net_init_firmware(card);
if (result)
goto init_firmware_failed;
/* start probing with copper */
card->aneg_count = 0;
card->medium = BCM54XX_COPPER;
spider_net_setup_aneg(card);
if (card->phy.def->phy_id)
mod_timer(&card->aneg_timer, jiffies + SPIDER_NET_ANEG_TIMER);
result = spider_net_init_chain(card, &card->tx_chain);
if (result)
goto alloc_tx_failed;
[PATCH] powerpc/cell spidernet low watermark patch. Implement basic low-watermark support for the transmit queue. Hardware low-watermarks allow a properly configured kernel to continously stream data to a device and not have to handle any interrupts at all in doing so. Correct zero-interrupt operation can be actually observed for this driver, when the socket buffer is made large enough. The basic idea of a low-watermark interrupt is as follows. The device driver queues up a bunch of packets for the hardware to transmit, and then kicks the hardware to get it started. As the hardware drains the queue of pending, untransmitted packets, the device driver will want to know when the queue is almost empty, so that it can queue some more packets. If the queue drains down to the low waterark, then an interrupt will be generated. However, if the kernel/driver continues to add enough packets to keep the queue partially filled, no interrupt will actually be generated, and the hardware can continue streaming packets indefinitely in this mode. The impelmentation is done by setting the DESCR_TXDESFLG flag in one of the packets. When the hardware sees this flag, it will interrupt the device driver. Because this flag is on a fixed packet, rather than at fixed location in the queue, the code below needs to move the flag as more packets are queued up. This implementation attempts to keep the flag at about 1/4 from "empty". Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: James K Lewis <jklewis@us.ibm.com> Acked-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2006-10-10 21:11:33 +00:00
card->low_watermark = NULL;
result = spider_net_init_chain(card, &card->rx_chain);
if (result)
goto alloc_rx_failed;
/* Allocate rx skbs */
if (spider_net_alloc_rx_skbs(card))
goto alloc_skbs_failed;
spider_net_set_multi(netdev);
/* further enhancement: setup hw vlan, if needed */
result = -EBUSY;
if (request_irq(netdev->irq, spider_net_interrupt,
IRQF_SHARED, netdev->name, netdev))
goto register_int_failed;
spider_net_enable_card(card);
netif_start_queue(netdev);
netif_carrier_on(netdev);
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 23:41:36 +00:00
napi_enable(&card->napi);
spider_net_enable_interrupts(card);
return 0;
register_int_failed:
spider_net_free_rx_chain_contents(card);
alloc_skbs_failed:
spider_net_free_chain(card, &card->rx_chain);
alloc_rx_failed:
spider_net_free_chain(card, &card->tx_chain);
alloc_tx_failed:
del_timer_sync(&card->aneg_timer);
init_firmware_failed:
return result;
}
/**
* spider_net_link_phy
* @data: used for pointer to card structure
*
*/
static void spider_net_link_phy(unsigned long data)
{
struct spider_net_card *card = (struct spider_net_card *)data;
struct mii_phy *phy = &card->phy;
/* if link didn't come up after SPIDER_NET_ANEG_TIMEOUT tries, setup phy again */
if (card->aneg_count > SPIDER_NET_ANEG_TIMEOUT) {
pr_debug("%s: link is down trying to bring it up\n",
card->netdev->name);
switch (card->medium) {
case BCM54XX_COPPER:
/* enable fiber with autonegotiation first */
if (phy->def->ops->enable_fiber)
phy->def->ops->enable_fiber(phy, 1);
card->medium = BCM54XX_FIBER;
break;
case BCM54XX_FIBER:
/* fiber didn't come up, try to disable fiber autoneg */
if (phy->def->ops->enable_fiber)
phy->def->ops->enable_fiber(phy, 0);
card->medium = BCM54XX_UNKNOWN;
break;
case BCM54XX_UNKNOWN:
/* copper, fiber with and without failed,
* retry from beginning */
spider_net_setup_aneg(card);
card->medium = BCM54XX_COPPER;
break;
}
card->aneg_count = 0;
mod_timer(&card->aneg_timer, jiffies + SPIDER_NET_ANEG_TIMER);
return;
}
/* link still not up, try again later */
if (!(phy->def->ops->poll_link(phy))) {
card->aneg_count++;
mod_timer(&card->aneg_timer, jiffies + SPIDER_NET_ANEG_TIMER);
return;
}
/* link came up, get abilities */
phy->def->ops->read_link(phy);
spider_net_write_reg(card, SPIDER_NET_GMACST,
spider_net_read_reg(card, SPIDER_NET_GMACST));
spider_net_write_reg(card, SPIDER_NET_GMACINTEN, 0x4);
if (phy->speed == 1000)
spider_net_write_reg(card, SPIDER_NET_GMACMODE, 0x00000001);
else
spider_net_write_reg(card, SPIDER_NET_GMACMODE, 0);
card->aneg_count = 0;
pr_info("%s: link up, %i Mbps, %s-duplex %sautoneg.\n",
card->netdev->name, phy->speed,
phy->duplex == 1 ? "Full" : "Half",
phy->autoneg == 1 ? "" : "no ");
}
/**
* spider_net_setup_phy - setup PHY
* @card: card structure
*
* returns 0 on success, <0 on failure
*
* spider_net_setup_phy is used as part of spider_net_probe.
**/
static int
spider_net_setup_phy(struct spider_net_card *card)
{
struct mii_phy *phy = &card->phy;
spider_net_write_reg(card, SPIDER_NET_GDTDMASEL,
SPIDER_NET_DMASEL_VALUE);
spider_net_write_reg(card, SPIDER_NET_GPCCTRL,
SPIDER_NET_PHY_CTRL_VALUE);
phy->dev = card->netdev;
phy->mdio_read = spider_net_read_phy;
phy->mdio_write = spider_net_write_phy;
for (phy->mii_id = 1; phy->mii_id <= 31; phy->mii_id++) {
unsigned short id;
id = spider_net_read_phy(card->netdev, phy->mii_id, MII_BMSR);
if (id != 0x0000 && id != 0xffff) {
if (!mii_phy_probe(phy, phy->mii_id)) {
pr_info("Found %s.\n", phy->def->name);
break;
}
}
}
return 0;
}
/**
* spider_net_workaround_rxramfull - work around firmware bug
* @card: card structure
*
* no return value
**/
static void
spider_net_workaround_rxramfull(struct spider_net_card *card)
{
int i, sequencer = 0;
/* cancel reset */
spider_net_write_reg(card, SPIDER_NET_CKRCTRL,
SPIDER_NET_CKRCTRL_RUN_VALUE);
/* empty sequencer data */
for (sequencer = 0; sequencer < SPIDER_NET_FIRMWARE_SEQS;
sequencer++) {
spider_net_write_reg(card, SPIDER_NET_GSnPRGADR +
sequencer * 8, 0x0);
for (i = 0; i < SPIDER_NET_FIRMWARE_SEQWORDS; i++) {
spider_net_write_reg(card, SPIDER_NET_GSnPRGDAT +
sequencer * 8, 0x0);
}
}
/* set sequencer operation */
spider_net_write_reg(card, SPIDER_NET_GSINIT, 0x000000fe);
/* reset */
spider_net_write_reg(card, SPIDER_NET_CKRCTRL,
SPIDER_NET_CKRCTRL_STOP_VALUE);
}
/**
* spider_net_stop - called upon ifconfig down
* @netdev: interface device structure
*
* always returns 0
*/
int
spider_net_stop(struct net_device *netdev)
{
struct spider_net_card *card = netdev_priv(netdev);
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 23:41:36 +00:00
napi_disable(&card->napi);
netif_carrier_off(netdev);
netif_stop_queue(netdev);
del_timer_sync(&card->tx_timer);
del_timer_sync(&card->aneg_timer);
spider_net_disable_interrupts(card);
free_irq(netdev->irq, netdev);
spider_net_write_reg(card, SPIDER_NET_GDTDMACCNTR,
SPIDER_NET_DMA_TX_FEND_VALUE);
/* turn off DMA, force end */
spider_net_disable_rxdmac(card);
/* release chains */
spider_net_release_tx_chain(card, 1);
spider_net_free_rx_chain_contents(card);
spider_net_free_chain(card, &card->tx_chain);
spider_net_free_chain(card, &card->rx_chain);
return 0;
}
/**
* spider_net_tx_timeout_task - task scheduled by the watchdog timeout
* function (to be called not under interrupt status)
* @data: data, is interface device structure
*
* called as task when tx hangs, resets interface (if interface is up)
*/
static void
spider_net_tx_timeout_task(struct work_struct *work)
{
struct spider_net_card *card =
container_of(work, struct spider_net_card, tx_timeout_task);
struct net_device *netdev = card->netdev;
if (!(netdev->flags & IFF_UP))
goto out;
netif_device_detach(netdev);
spider_net_stop(netdev);
spider_net_workaround_rxramfull(card);
spider_net_init_card(card);
if (spider_net_setup_phy(card))
goto out;
spider_net_open(netdev);
spider_net_kick_tx_dma(card);
netif_device_attach(netdev);
out:
atomic_dec(&card->tx_timeout_task_counter);
}
/**
* spider_net_tx_timeout - called when the tx timeout watchdog kicks in.
* @netdev: interface device structure
*
* called, if tx hangs. Schedules a task that resets the interface
*/
static void
spider_net_tx_timeout(struct net_device *netdev)
{
struct spider_net_card *card;
card = netdev_priv(netdev);
atomic_inc(&card->tx_timeout_task_counter);
if (netdev->flags & IFF_UP)
schedule_work(&card->tx_timeout_task);
else
atomic_dec(&card->tx_timeout_task_counter);
card->spider_stats.tx_timeouts++;
}
static const struct net_device_ops spider_net_ops = {
.ndo_open = spider_net_open,
.ndo_stop = spider_net_stop,
.ndo_start_xmit = spider_net_xmit,
.ndo_set_multicast_list = spider_net_set_multi,
.ndo_set_mac_address = spider_net_set_mac,
.ndo_change_mtu = spider_net_change_mtu,
.ndo_do_ioctl = spider_net_do_ioctl,
.ndo_tx_timeout = spider_net_tx_timeout,
.ndo_validate_addr = eth_validate_addr,
/* HW VLAN */
#ifdef CONFIG_NET_POLL_CONTROLLER
/* poll controller */
.ndo_poll_controller = spider_net_poll_controller,
#endif /* CONFIG_NET_POLL_CONTROLLER */
};
/**
* spider_net_setup_netdev_ops - initialization of net_device operations
* @netdev: net_device structure
*
* fills out function pointers in the net_device structure
*/
static void
spider_net_setup_netdev_ops(struct net_device *netdev)
{
netdev->netdev_ops = &spider_net_ops;
netdev->watchdog_timeo = SPIDER_NET_WATCHDOG_TIMEOUT;
/* ethtool ops */
netdev->ethtool_ops = &spider_net_ethtool_ops;
}
/**
* spider_net_setup_netdev - initialization of net_device
* @card: card structure
*
* Returns 0 on success or <0 on failure
*
* spider_net_setup_netdev initializes the net_device structure
**/
static int
spider_net_setup_netdev(struct spider_net_card *card)
{
int result;
struct net_device *netdev = card->netdev;
struct device_node *dn;
struct sockaddr addr;
const u8 *mac;
SET_NETDEV_DEV(netdev, &card->pdev->dev);
pci_set_drvdata(card->pdev, netdev);
init_timer(&card->tx_timer);
card->tx_timer.function =
(void (*)(unsigned long)) spider_net_cleanup_tx_ring;
card->tx_timer.data = (unsigned long) card;
netdev->irq = card->pdev->irq;
card->aneg_count = 0;
init_timer(&card->aneg_timer);
card->aneg_timer.function = spider_net_link_phy;
card->aneg_timer.data = (unsigned long) card;
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-03 23:41:36 +00:00
netif_napi_add(netdev, &card->napi,
spider_net_poll, SPIDER_NET_NAPI_WEIGHT);
spider_net_setup_netdev_ops(netdev);
netdev->hw_features = NETIF_F_RXCSUM | NETIF_F_IP_CSUM;
if (SPIDER_NET_RX_CSUM_DEFAULT)
netdev->features |= NETIF_F_RXCSUM;
netdev->features |= NETIF_F_IP_CSUM | NETIF_F_LLTX;
/* some time: NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX |
* NETIF_F_HW_VLAN_FILTER */
netdev->irq = card->pdev->irq;
spidernet: Cure RX ram full bug This patch fixes a rare deadlock that can occur when the kernel is not able to empty out the RX ring quickly enough. Below follows a detailed description of the bug and the fix. As long as the OS can empty out the RX buffers at a rate faster than the hardware can fill them, there is no problem. If, for some reason, the OS fails to empty the RX ring fast enough, the hardware GDACTDPA pointer will catch up to the head, notice the not-empty condition, ad stop. However, RX packets may still continue arriving on the wire. The spidernet chip can save some limited number of these in local RAM. When this local ram fills up, the spider chip will issue an interrupt indicating this (GHIINT0STS will show ERRINT, and the GRMFLLINT bit will be set in GHIINT1STS). When te RX ram full condition occurs, a certain bug/feature is triggered that has to be specially handled. This section describes the special handling for this condition. When the OS finally has a chance to run, it will empty out the RX ring. In particular, it will clear the descriptor on which the hardware had stopped. However, once the hardware has decided that a certain descriptor is invalid, it will not restart at that descriptor; instead it will restart at the next descr. This potentially will lead to a deadlock condition, as the tail pointer will be pointing at this descr, which, from the OS point of view, is empty; the OS will be waiting for this descr to be filled. However, the hardware has skipped this descr, and is filling the next descrs. Since the OS doesn't see this, there is a potential deadlock, with the OS waiting for one descr to fill, while the hardware is waiting for a differen set of descrs to become empty. A call to show_rx_chain() at this point indicates the nature of the problem. A typical print when the network is hung shows the following: net eth1: Spider RX RAM full, incoming packets might be discarded! net eth1: Total number of descrs=256 net eth1: Chain tail located at descr=255 net eth1: Chain head is at 255 net eth1: HW curr desc (GDACTDPA) is at 0 net eth1: Have 1 descrs with stat=xa0800000 net eth1: HW next desc (GDACNEXTDA) is at 1 net eth1: Have 127 descrs with stat=x40800101 net eth1: Have 1 descrs with stat=x40800001 net eth1: Have 126 descrs with stat=x40800101 net eth1: Last 1 descrs with stat=xa0800000 Both the tail and head pointers are pointing at descr 255, which is marked xa... which is "empty". Thus, from the OS point of view, there is nothing to be done. In particular, there is the implicit assumption that everything in front of the "empty" descr must surely also be empty, as explained in the last section. The OS is waiting for descr 255 to become non-empty, which, in this case, will never happen. The HW pointer is at descr 0. This descr is marked 0x4.. or "full". Since its already full, the hardware can do nothing more, and thus has halted processing. Notice that descrs 0 through 254 are all marked "full", while descr 254 and 255 are empty. (The "Last 1 descrs" is descr 254, since tail was at 255.) Thus, the system is deadlocked, and there can be no forward progress; the OS thinks there's nothing to do, and the hardware has nowhere to put incoming data. This bug/feature is worked around with the spider_net_resync_head_ptr() routine. When the driver receives RX interrupts, but an examination of the RX chain seems to show it is empty, then it is probable that the hardware has skipped a descr or two (sometimes dozens under heavy network conditions). The spider_net_resync_head_ptr() subroutine will search the ring for the next full descr, and the driver will resume operations there. Since this will leave "holes" in the ring, there is also a spider_net_resync_tail_ptr() that will skip over such holes. Signed-off-by: Linas Vepstas <linas@austin.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-06-11 18:21:13 +00:00
card->num_rx_ints = 0;
card->ignore_rx_ramfull = 0;
dn = pci_device_to_OF_node(card->pdev);
if (!dn)
return -EIO;
mac = of_get_property(dn, "local-mac-address", NULL);
if (!mac)
return -EIO;
memcpy(addr.sa_data, mac, ETH_ALEN);
result = spider_net_set_mac(netdev, &addr);
if ((result) && (netif_msg_probe(card)))
dev_err(&card->netdev->dev,
"Failed to set MAC address: %i\n", result);
result = register_netdev(netdev);
if (result) {
if (netif_msg_probe(card))
dev_err(&card->netdev->dev,
"Couldn't register net_device: %i\n", result);
return result;
}
if (netif_msg_probe(card))
pr_info("Initialized device %s.\n", netdev->name);
return 0;
}
/**
* spider_net_alloc_card - allocates net_device and card structure
*
* returns the card structure or NULL in case of errors
*
* the card and net_device structures are linked to each other
*/
static struct spider_net_card *
spider_net_alloc_card(void)
{
struct net_device *netdev;
struct spider_net_card *card;
size_t alloc_size;
alloc_size = sizeof(struct spider_net_card) +
(tx_descriptors + rx_descriptors) * sizeof(struct spider_net_descr);
netdev = alloc_etherdev(alloc_size);
if (!netdev)
return NULL;
card = netdev_priv(netdev);
card->netdev = netdev;
card->msg_enable = SPIDER_NET_DEFAULT_MSG;
INIT_WORK(&card->tx_timeout_task, spider_net_tx_timeout_task);
init_waitqueue_head(&card->waitq);
atomic_set(&card->tx_timeout_task_counter, 0);
card->rx_chain.num_desc = rx_descriptors;
card->rx_chain.ring = card->darray;
card->tx_chain.num_desc = tx_descriptors;
card->tx_chain.ring = card->darray + rx_descriptors;
return card;
}
/**
* spider_net_undo_pci_setup - releases PCI ressources
* @card: card structure
*
* spider_net_undo_pci_setup releases the mapped regions
*/
static void
spider_net_undo_pci_setup(struct spider_net_card *card)
{
iounmap(card->regs);
pci_release_regions(card->pdev);
}
/**
* spider_net_setup_pci_dev - sets up the device in terms of PCI operations
* @pdev: PCI device
*
* Returns the card structure or NULL if any errors occur
*
* spider_net_setup_pci_dev initializes pdev and together with the
* functions called in spider_net_open configures the device so that
* data can be transferred over it
* The net_device structure is attached to the card structure, if the
* function returns without error.
**/
static struct spider_net_card *
spider_net_setup_pci_dev(struct pci_dev *pdev)
{
struct spider_net_card *card;
unsigned long mmio_start, mmio_len;
if (pci_enable_device(pdev)) {
dev_err(&pdev->dev, "Couldn't enable PCI device\n");
return NULL;
}
if (!(pci_resource_flags(pdev, 0) & IORESOURCE_MEM)) {
dev_err(&pdev->dev,
"Couldn't find proper PCI device base address.\n");
goto out_disable_dev;
}
if (pci_request_regions(pdev, spider_net_driver_name)) {
dev_err(&pdev->dev,
"Couldn't obtain PCI resources, aborting.\n");
goto out_disable_dev;
}
pci_set_master(pdev);
card = spider_net_alloc_card();
if (!card) {
dev_err(&pdev->dev,
"Couldn't allocate net_device structure, aborting.\n");
goto out_release_regions;
}
card->pdev = pdev;
/* fetch base address and length of first resource */
mmio_start = pci_resource_start(pdev, 0);
mmio_len = pci_resource_len(pdev, 0);
card->netdev->mem_start = mmio_start;
card->netdev->mem_end = mmio_start + mmio_len;
card->regs = ioremap(mmio_start, mmio_len);
if (!card->regs) {
dev_err(&pdev->dev,
"Couldn't obtain PCI resources, aborting.\n");
goto out_release_regions;
}
return card;
out_release_regions:
pci_release_regions(pdev);
out_disable_dev:
pci_disable_device(pdev);
pci_set_drvdata(pdev, NULL);
return NULL;
}
/**
* spider_net_probe - initialization of a device
* @pdev: PCI device
* @ent: entry in the device id list
*
* Returns 0 on success, <0 on failure
*
* spider_net_probe initializes pdev and registers a net_device
* structure for it. After that, the device can be ifconfig'ed up
**/
static int __devinit
spider_net_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
{
int err = -EIO;
struct spider_net_card *card;
card = spider_net_setup_pci_dev(pdev);
if (!card)
goto out;
spider_net_workaround_rxramfull(card);
spider_net_init_card(card);
err = spider_net_setup_phy(card);
if (err)
goto out_undo_pci;
err = spider_net_setup_netdev(card);
if (err)
goto out_undo_pci;
return 0;
out_undo_pci:
spider_net_undo_pci_setup(card);
free_netdev(card->netdev);
out:
return err;
}
/**
* spider_net_remove - removal of a device
* @pdev: PCI device
*
* Returns 0 on success, <0 on failure
*
* spider_net_remove is called to remove the device and unregisters the
* net_device
**/
static void __devexit
spider_net_remove(struct pci_dev *pdev)
{
struct net_device *netdev;
struct spider_net_card *card;
netdev = pci_get_drvdata(pdev);
card = netdev_priv(netdev);
wait_event(card->waitq,
atomic_read(&card->tx_timeout_task_counter) == 0);
unregister_netdev(netdev);
/* switch off card */
spider_net_write_reg(card, SPIDER_NET_CKRCTRL,
SPIDER_NET_CKRCTRL_STOP_VALUE);
spider_net_write_reg(card, SPIDER_NET_CKRCTRL,
SPIDER_NET_CKRCTRL_RUN_VALUE);
spider_net_undo_pci_setup(card);
free_netdev(netdev);
}
static struct pci_driver spider_net_driver = {
.name = spider_net_driver_name,
.id_table = spider_net_pci_tbl,
.probe = spider_net_probe,
.remove = __devexit_p(spider_net_remove)
};
/**
* spider_net_init - init function when the driver is loaded
*
* spider_net_init registers the device driver
*/
static int __init spider_net_init(void)
{
printk(KERN_INFO "Spidernet version %s.\n", VERSION);
if (rx_descriptors < SPIDER_NET_RX_DESCRIPTORS_MIN) {
rx_descriptors = SPIDER_NET_RX_DESCRIPTORS_MIN;
pr_info("adjusting rx descriptors to %i.\n", rx_descriptors);
}
if (rx_descriptors > SPIDER_NET_RX_DESCRIPTORS_MAX) {
rx_descriptors = SPIDER_NET_RX_DESCRIPTORS_MAX;
pr_info("adjusting rx descriptors to %i.\n", rx_descriptors);
}
if (tx_descriptors < SPIDER_NET_TX_DESCRIPTORS_MIN) {
tx_descriptors = SPIDER_NET_TX_DESCRIPTORS_MIN;
pr_info("adjusting tx descriptors to %i.\n", tx_descriptors);
}
if (tx_descriptors > SPIDER_NET_TX_DESCRIPTORS_MAX) {
tx_descriptors = SPIDER_NET_TX_DESCRIPTORS_MAX;
pr_info("adjusting tx descriptors to %i.\n", tx_descriptors);
}
return pci_register_driver(&spider_net_driver);
}
/**
* spider_net_cleanup - exit function when driver is unloaded
*
* spider_net_cleanup unregisters the device driver
*/
static void __exit spider_net_cleanup(void)
{
pci_unregister_driver(&spider_net_driver);
}
module_init(spider_net_init);
module_exit(spider_net_cleanup);