Now when both 32-bit and 64-bit code inside the driver is using
fwspec it is possible to replace the utlb handling with fwspec ids
that get populated from ->of_xlate().
Suggested-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Consolidate the 32-bit and 64-bit code to make use of fwspec instead
of archdata for the 32-bit ARM case.
This is a simplified version of the fwspec handling code from Robin
posted as [PATCH] iommu/ipmmu-vmsa: Convert to iommu_fwspec
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
The 32-bit ARM code gets updated to make use of ->of_xlate() and the
code is shared between 64-bit and 32-bit ARM. The of_device_is_available()
check gets dropped since it is included in of_iommu_xlate().
Suggested-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Extend the driver to make use of iommu_device_register()/unregister()
functions together with iommu_device_set_ops() and iommu_set_fwnode().
These used to be part of the earlier posted 64-bit ARM (r8a7795) series but
it turns out that these days they are required on 32-bit ARM as well.
Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
When an administratively set MAC was previously set and should now be
switched back to 00:00:00:00:00:00 the pf_set_mac flag did not get
toggled back to false.
As a result VFs were still treated as if an administratively set MAC was
present.
Signed-off-by: Stefan Assmann <sassmann@kpanic.de>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
i40e hardware descriptor fields are in little-endian format. Driver
must use le32_to_cpu while evaluating these fields otherwise on
big-endian arch we end up evaluating incorrect values, cause errors
like:
i40evf 0000:03:0a.0: Expected response 24 from PF, received 402653184
i40evf 0000:03:0a.1: Expected response 7 from PF, received 117440512
Signed-off-by: Tushar Dave <tushar.n.dave@oracle.com>
Reviewed-by: Shannon Nelson <shannon.nelson@oracle.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Fill the XDP prog_id with the id just like we do in other XDP enabled
drivers such as ixgbe. This is needed so that on dump we can retrieve
the attached program based on the id, and dump BPF insns, opcodes, etc
back to user space. Only XDP driver missing this is currently i40e.
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
These includes were all being used in the driver, but weren't
being directly included.
Since the current advised method is to directly include anything
that you need, this implements that.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The i40e driver attempts to display the UDP tunnel name by doing a check
against the type, where for non-zero types we use "vxlan" and for zero
type we use "geneve". This is not future proof, because if new tunnel
types get added, we'll incorrectly label them. It also depends on the
value of UDP_TUNNEL_TYPE_GENEVE == 0, which is brittle.
Instead, replace this with a function that can return a constant string
depending on the type. For now we'll use "unknown" for types we don't
know about, and we can expand this in the future if new types get added.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
The compiler warned on an oddly indented bit of code, and when
investigating that, noted that the functions themselves had
an odd flow. The if condition was checked, and would exclude
a call to AQ, but then the aq_ret would be checked unconditionally
which just looks really weird, and is likely to cause objections.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
As it turns out there was only a small set of errors
on 32 bit, and we just needed to be using the right calls
for dealing with timespec64 variables.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
There are some rare cases where the release resource call will return an
admin Q timeout. In these cases the code needs to try to release the
resource again until it succeeds or it times out.
Signed-off-by: Paul M Stillwell Jr <paul.m.stillwell.jr@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
During certain events such as a CORER, multiple devices will run a work
task to handle some cleanup. This can cause issues due to
a single-threaded workqueue which can mean that a device doesn't cleanup
in time. Prevent this by removing the single-threaded restriction on the
module workqueue. This avoids the need to add more complex yielding
logic in our service task routine. This is also similar to what other
drivers such as fm10k do.
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
This patch fixes a problem found in systems when entering
S4 state. This patch fixes the problem by ensuring that
the misc vector's IRQ is disabled as well. Without this
patch a stack trace can be seen upon entering S4 state.
Signed-off-by: Carolyn Wyborny <carolyn.wyborny@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Fix incorrect variable assignment.
Based on line 1511: aq_ret = I40_ERR_PARAM; the correct variable to be
used in this instance is aq_ret instead of ret. Also, variable ret is
updated at line 1602 just before return, so assigning a value to this
variable in this code block is useless.
Addresses-Coverity-ID: 1397693
Signed-off-by: Gustavo A R Silva <garsilva@embeddedor.com>
Acked-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
drop VPFE_CMD_S_CCDC_RAW_PARAMS ioctl from dm355/dm644x following reasons:
- This ioctl was never in public api and was only defined in kernel header.
- The function set_params constantly mixes up pointers and phys_addr_t
numbers.
- This is part of a 'VPFE_CMD_S_CCDC_RAW_PARAMS' ioctl command that is
described as an 'experimental ioctl that will change in future kernels'.
- The code to allocate the table never gets called after we copy_from_user
the user input over the kernel settings, and then compare them
for inequality.
- We then go on to use an address provided by user space as both the
__user pointer for input and pass it through phys_to_virt to come up
with a kernel pointer to copy the data to. This looks like a trivially
exploitable root hole.
Signed-off-by: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
this patch makes sure VPFE_CMD_S_CCDC_RAW_PARAMS ioctl no longer works
for vpfe_capture driver with a minimal patch suitable for backporting.
- This ioctl was never in public api and was only defined in kernel header.
- The function set_params constantly mixes up pointers and phys_addr_t
numbers.
- This is part of a 'VPFE_CMD_S_CCDC_RAW_PARAMS' ioctl command that is
described as an 'experimental ioctl that will change in future kernels'.
- The code to allocate the table never gets called after we copy_from_user
the user input over the kernel settings, and then compare them
for inequality.
- We then go on to use an address provided by user space as both the
__user pointer for input and pass it through phys_to_virt to come up
with a kernel pointer to copy the data to. This looks like a trivially
exploitable root hole.
Due to these reasons we make sure this ioctl now returns -EINVAL and backport
this patch as far as possible.
Fixes: 5f15fbb68f ("V4L/DVB (12251): v4l: dm644x ccdc module for vpfe capture driver")
Signed-off-by: Lad, Prabhakar <prabhakar.csengg@gmail.com>
Cc: <stable@vger.kernel.org> # for v3.7 and up
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
In venus_boot(), we pass a pointer to a phys_addr_t
into dmam_alloc_coherent, which the compiler warns about:
platform/qcom/venus/firmware.c: In function 'venus_boot':
platform/qcom/venus/firmware.c:63:49: error: passing argument 3 of 'dmam_alloc_coherent' from incompatible pointer type [-Werror=incompatible-pointer-types]
To avoid the error refactor venus_boot function by discard
dma_alloc_coherent invocation because we don't want to map the
memory for the device. Something more, the usage of
DMA mapping API is actually wrong and the current
implementation relies on several bugs in DMA mapping code.
When these bugs are fixed that will break firmware loading,
so fix this now to avoid future troubles.
The meaning of venus_boot is to copy the content of the
firmware buffer into reserved (and memblock removed)
block of memory and pass that physical address to the
trusted zone for authentication and mapping through iommu
form the secure world. After iommu mapping is done the iova
is passed as ane entry point to the remote processor.
After this change memory-region property is parsed manually
and the physical address is memremap to CPU, call mdt_load to
load firmware segments into proper places and unmap
reserved memory.
Fixes: af2c3834c8 ("[media] media: venus: adding core part and helper functions")
Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Not entirely sure what triggers it, but with venus build as kernel
module and in initrd, we hit this crash:
Unable to handle kernel paging request at virtual address ffff80003c039000
pgd = ffff00000a14f000
[ffff80003c039000] *pgd=00000000bd9f7003, *pud=00000000bd9f6003, *pmd=00000000bd9f0003, *pte=0000000000000000
Internal error: Oops: 96000007 [#1] SMP
Modules linked in: qcom_wcnss_pil(E+) crc32_ce(E) qcom_common(E) venus_core(E+) remoteproc(E) snd_soc_msm8916_digital(E) virtio_ring(E) cdc_ether(E) snd_soc_lpass_apq8016(E) snd_soc_lpass_cpu(E) snd_soc_apq8016_sbc(E) snd_soc_lpass_platform(E) v4l2_mem2mem(E) virtio(E) snd_soc_core(E) ac97_bus(E) snd_pcm_dmaengine(E) snd_seq(E) leds_gpio(E) videobuf2_v4l2(E) videobuf2_core(E) snd_seq_device(E) snd_pcm(E) videodev(E) media(E) nvmem_qfprom(E) msm(E) snd_timer(E) snd(E) soundcore(E) spi_qup(E) mdt_loader(E) qcom_tsens(E) qcom_spmi_temp_alarm(E) nvmem_core(E) msm_rng(E) uas(E) usb_storage(E) dm9601(E) usbnet(E) mii(E) mmc_block(E) adv7511(E) drm_kms_helper(E) syscopyarea(E) sysfillrect(E) sysimgblt(E) fb_sys_fops(E) qcom_spmi_vadc(E) qcom_vadc_common(PE) industrialio(E) pinctrl_spmi_mpp(E)
pinctrl_spmi_gpio(E) rtc_pm8xxx(E) clk_smd_rpm(E) sdhci_msm(E) sdhci_pltfm(E) qcom_smd_regulator(E) drm(E) smd_rpm(E) qcom_spmi_pmic(E) regmap_spmi(E) ci_hdrc_msm(E) ci_hdrc(E) usb3503(E) extcon_usb_gpio(E) phy_msm_usb(E) udc_core(E) qcom_hwspinlock(E) extcon_core(E) ehci_msm(E) i2c_qup(E) sdhci(E) mmc_core(E) spmi_pmic_arb(E) spmi(E) qcom_smd(E) smsm(E) rpmsg_core(E) smp2p(E) smem(E) hwspinlock_core(E) gpio_keys(E)
CPU: 2 PID: 551 Comm: irq/150-venus Tainted: P E 4.12.0+ #1625
Hardware name: qualcomm dragonboard410c/dragonboard410c, BIOS 2017.07-rc2-00144-ga97bdbdf72-dirty 07/08/2017
task: ffff800037338000 task.stack: ffff800038e00000
PC is at hfi_sys_init_done+0x64/0x140 [venus_core]
LR is at hfi_process_msg_packet+0xcc/0x1e8 [venus_core]
pc : [<ffff00000118b384>] lr : [<ffff00000118c11c>] pstate: 20400145
sp : ffff800038e03c60
x29: ffff800038e03c60 x28: 0000000000000000
x27: 00000000000df018 x26: ffff00000118f4d0
x25: 0000000000020003 x24: ffff80003a8d3010
x23: ffff00000118f760 x22: ffff800037b40028
x21: ffff8000382981f0 x20: ffff800037b40028
x19: ffff80003c039000 x18: 0000000000000020
x17: 0000000000000000 x16: ffff800037338000
x15: ffffffffffffffff x14: 0000001000000014
x13: 0000000100001007 x12: 0000000100000020
x11: 0000100e00000000 x10: 0000000000000001
x9 : 0000000200000000 x8 : 0000001400000001
x7 : 0000000000001010 x6 : 0000000000000148
x5 : 0000000000001009 x4 : ffff80003c039000
x3 : 00000000cd770abb x2 : 0000000000000042
x1 : 0000000000000788 x0 : 0000000000000002
Process irq/150-venus (pid: 551, stack limit = 0xffff800038e00000)
Call trace:
[<ffff00000118b384>] hfi_sys_init_done+0x64/0x140 [venus_core]
[<ffff00000118c11c>] hfi_process_msg_packet+0xcc/0x1e8 [venus_core]
[<ffff00000118a2b4>] venus_isr_thread+0x1b4/0x208 [venus_core]
[<ffff00000118e750>] hfi_isr_thread+0x28/0x38 [venus_core]
[<ffff000008161550>] irq_thread_fn+0x30/0x70
[<ffff0000081617fc>] irq_thread+0x14c/0x1c8
[<ffff000008105e68>] kthread+0x138/0x140
[<ffff000008083590>] ret_from_fork+0x10/0x40
Code: 52820125 52820207 7a431820 54000249 (b9400263)
---[ end trace c963460f20a984b6 ]---
The problem is that in the error case, we've incremented the data ptr
but not decremented rem_bytes, and keep reading (presumably garbage)
until eventually we go beyond the end of the buffer.
Instead, on first error, we should probably just bail out. Other
option is to increment read_bytes by sizeof(u32) before the switch,
rather than only accounting for the ptype header in the non-error
case. Note that in this case it is HFI_ERR_SYS_INVALID_PARAMETER,
ie. an unrecognized/unsupported parameter, so interpreting the next
word as a property type would be bogus. The other error cases are
due to truncated buffer, so there isn't likely to be anything valid
to interpret in the remainder of the buffer. So just bailing seems
like a reasonable solution.
Signed-off-by: Rob Clark <robdclark@gmail.com>
Reviewed-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
If QCOM_MDT_LOADER is enabled, but ARCH_QCOM is not, we run into
a build error:
ERROR: "qcom_mdt_load" [drivers/media/platform/qcom/venus/venus-core.ko] undefined!
ERROR: "qcom_mdt_get_size" [drivers/media/platform/qcom/venus/venus-core.ko] undefined!
This changes the 'select' statement again, so we only try to enable
those symbols when the drivers will actually get built, and explicitly
test for QCOM_MDT_LOADER to be enabled before calling into it.
Fixes: 76724b30f2 ("[media] media: venus: enable building with COMPILE_TEST")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Without PM support, gcc warns about two unused functions:
platform/qcom/venus/core.c:146:13: error: 'venus_clks_disable' defined but not used [-Werror=unused-function]
platform/qcom/venus/core.c:126:12: error: 'venus_clks_enable' defined but not used [-Werror=unused-function]
The problem as usual are incorrect #ifdefs, so the easiest fix
is to do away with the #ifdef completely and mark the suspend/resume
handlers as __maybe_unused, which they are.
Fixes: af2c3834c8 ("[media] media: venus: adding core part and helper functions")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Wrong Kconfig option was used when adding warning for untested
big-endian capabilities. There is no CONFIG_BIG_ENDIAN option.
Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Allow calling cec_notifier_set_phys_addr and
cec_notifier_set_phys_addr_from_edid with a NULL notifier, in which
case these functions do nothing.
Add a cec_notifier_phys_addr_invalidate helper function (the notifier
equivalent of cec_phys_addr_invalidate).
These changes simplify drm CEC driver support.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The persistent_config option is used to make the CEC settings persistent by using
the eeprom inside the device to store this information. This was on by default, which
caused confusion since this device now behaves differently from other CEC devices
which all come up unconfigured.
Another reason for doing this now is that I hope a more standard way of selecting
persistent configuration will be created in the future. And for that to work all
CEC drivers should behave the same and come up unconfigured by default.
None of the open source CEC applications are using this CEC framework at the moment
so change this behavior before it is too late.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Cc: <stable@vger.kernel.org> # for v4.10 and up
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The switch in cec_transmit_attempt_done() should ignore the
CEC_TX_STATUS_MAX_RETRIES status bit.
Calling this function with e.g. CEC_TX_STATUS_NACK | CEC_TX_STATUS_MAX_RETRIES
is perfectly legal and should not trigger the WARN(1).
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
I noticed an array underflow in ov5693_enum_frame_size(). The code
looks like this:
int index = fse->index;
if (index >= N_RES)
retur -EINVAL;
fse->index is a u32 that comes from the user. We want negative values
to be counted as -EINVAL but they aren't. There are several ways to fix
this but I feel like the best fix for future proofing is to change the
type of N_RES from int to unsigned long to make it the same as if we
were comparing against ARRAY_SIZE().
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Since commit e8f4818895 ("[media] lirc: advertise
LIRC_CAN_GET_REC_RESOLUTION and improve") lircd uses the ioctl
LIRC_GET_REC_RESOLUTION to determine the shortest pulse or space that
the hardware can detect. This breaks decoding in lirc because lircd
expects the answer in microseconds, but nanoseconds is returned.
Cc: <stable@vger.kernel.org> # v2.6.36+
Reported-by: Derek <user.vdr@gmail.com>
Tested-by: Derek <user.vdr@gmail.com>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Sending patches with SVG files via e-mail has a drawback: line
size could be bigger than 998, with is the limit given by
RFC 5322[1]. So, we need to enforce a lower limit, in order to
allow those patches to be properly reviewed.
[1] https://tools.ietf.org/html/rfc5322#section-2.1.1
So, use this small Perl script to limit columns size to ~900.
use Text::Wrap;
$Text::Wrap::columns = 900;
$t.=$_ while (<>);
print wrap("","",$t);
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Debian's ImageMagick is currently unable to decode those
images. Use scour to simplify the SVG, and provide only
one font type, in order to make it more palatable.
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
This file is too big, with cause it to require a lot
of memory when parsed by texlive.
Optimize it, in order to avoid the need of touching at
main_memory at texmf.cnf.
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
The vimc platform drivers define a platform device ID table but these
are not set to the .id_table field in the platform driver structure.
So the platform device ID table is only used to fill the aliases in
the module but are not used for matching (works because the platform
subsystem fallbacks to the driver's name if no .id_table is set).
But this also means that the platform device ID table isn't used if
the driver is built-in, which leads to the following build warning:
This causes the following build warnings when the driver is built-in:
drivers/media/platform/vimc//vimc-capture.c:528:40: warning: ‘vimc_cap_driver_ids’ defined but not used [-Wunused-const-variable=]
static const struct platform_device_id vimc_cap_driver_ids[] = {
^~~~~~~~~~~~~~~~~~~
drivers/media/platform/vimc//vimc-debayer.c:588:40: warning: ‘vimc_deb_driver_ids’ defined but not used [-Wunused-const-variable=]
static const struct platform_device_id vimc_deb_driver_ids[] = {
^~~~~~~~~~~~~~~~~~~
drivers/media/platform/vimc//vimc-scaler.c:442:40: warning: ‘vimc_sca_driver_ids’ defined but not used [-Wunused-const-variable=]
static const struct platform_device_id vimc_sca_driver_ids[] = {
^~~~~~~~~~~~~~~~~~~
drivers/media/platform/vimc//vimc-sensor.c:376:40: warning: ‘vimc_sen_driver_ids’ defined but not used [-Wunused-const-variable=]
static const struct platform_device_id vimc_sen_driver_ids[] = {
^~~~~~~~~~~~~~~~~~~
Reported-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Suggested-by: Sakari Ailus <sakari.ailus@iki.fi>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Reviewed-by: Helen Koike <helen.koike@collabora.com>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Commit 7d3002cc8c ("iommu/core: split mapping to page sizes as supported
by the hardware") replaced 'int gfp_order' with a 'size_t size' of
iommu_map / iommu_unmap function arguments, but missed the function
prototypes for the disabled CONFIG_IOMMU_API case, let's correct them
for consistency.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
When adding a large scatterlist entry that covers more than the L3
superpage size (1GB) but has an alignment such that we must use L2
superpages (2MB) , we give dma_pte_free_level() a range that causes it
to free the L3 pagetable we're about to populate. We fix this by telling
dma_pte_free_pagetable() about the pagetable level we're about to populate
to prevent freeing it.
For example, mapping a scatterlist with entry lengths 854MB and 1194MB
at IOVA 0xffff80000000 would, when processing the 2MB-aligned second
entry, cause pfn_to_dma_pte() to create a L3 directory to hold L2
superpages for the mapping at IOVA 0xffffc0000000. We would previously
call dma_pte_free_pagetable(domain, 0xffffc0000, 0xfffffffff), which
would free the L3 directory pfn_to_dma_pte() just created for IO PFN
0xffffc0000. Telling dma_pte_free_pagetable() to retain the L3
directories while using L2 superpages avoids the erroneous free.
Signed-off-by: David Dillow <dillow@google.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
When a PCI device has DMA quirks, we need to ensure that an upstream
IOMMU knows about all possible aliases, since the presence of a DMA
quirk does not preclude the device still also emitting transactions
(e.g. MSIs) on its 'real' RID. Similarly, the rules for bridge aliasing
are relatively complex, and some bridges may only take ownership of
transactions under particular transient circumstances, leading again to
multiple RIDs potentially being seen at the IOMMU for the given device.
Take all this into account in the OF code by translating every RID
produced by the alias walk, not just whichever one comes out last.
Happily, this also makes things tidy enough that we can reduce the
number of both total lines of code, and confusing levels of indirection,
by pulling the "iommus"/"iommu-map" parsing helpers back in-line again.
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
write_sysreg() may misparse the value argument because it is used
without parentheses to protect it.
This patch adds the ( ) in order to avoid any surprises.
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
[will: same change to write_sysreg_s]
Signed-off-by: Will Deacon <will.deacon@arm.com>
The check for column exclusion did not verify that the event being
checked was an L2 event, and not a software event.
Software events should not be checked for column exclusion.
This resulted in a group with both software and L2 events sometimes
incorrectly rejecting the L2 event for column exclusion and
not counting it.
Add a check for PMU type before applying column exclusion logic.
Fixes: 21bdbb7102 ("perf: add qcom l2 cache perf events driver")
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Neil Leeder <nleeder@codeaurora.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
In commit efe0160cfd ("powerpc/64: Linker on-demand sfpr functions
for modules"), we added an ld version check early in the powerpc
top-level Makefile.
Because the Makefile runs before the kernel config is setup, the
checks for CONFIG_CPU_LITTLE_ENDIAN etc. all take the default case. So
we end up configuring ld for 32-bit big endian.
That would be OK, except that for historical (or perhaps no) reason,
we use 'override LD' to add the endian flags to the LD variable
itself, rather than the normal approach of adding them to LDFLAGS.
The end result is that when we check the ld version we run it as:
$(CROSS_COMPILE)ld -EB -m elf32ppc --version
This often works, unless you are using a 64-bit only and/or little
endian only, toolchain. In which case you see something like:
$ make defconfig
powerpc64le-linux-ld: unrecognised emulation mode: elf32ppc
Supported emulations: elf64lppc elf32lppc elf32lppclinux elf32lppcsim
/bin/sh: 1: [: -ge: unexpected operator
The proper fix is to stop using 'override LD', but that will require a
fair bit of testing. Instead we can fix it for now just by reordering
the Makefile to do the version check earlier.
Fixes: efe0160cfd ("powerpc/64: Linker on-demand sfpr functions for modules")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
As for commit 68baf692c4 ("powerpc/pseries: Fix of_node_put()
underflow during DLPAR remove"), the call to of_node_put() must be
removed from pSeries_reconfig_remove_node().
dlpar_detach_node() and pSeries_reconfig_remove_node() both call
of_detach_node(), and thus the node should not be released in both
cases.
Fixes: 0829f6d1f6 ("of: device_node kobject lifecycle fixes")
Cc: stable@vger.kernel.org # v3.15+
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
There's a somewhat architectural issue with Radix MMU and KVM.
When coming out of a guest with AIL (Alternate Interrupt Location, ie,
MMU enabled), we start executing hypervisor code with the PID register
still containing whatever the guest has been using.
The problem is that the CPU can (and will) then start prefetching or
speculatively load from whatever host context has that same PID (if
any), thus bringing translations for that context into the TLB, which
Linux doesn't know about.
This can cause stale translations and subsequent crashes.
Fixing this in a way that is neither racy nor a huge performance
impact is difficult. We could just make the host invalidations always
use broadcast forms but that would hurt single threaded programs for
example.
We chose to fix it instead by partitioning the PID space between guest
and host. This is possible because today Linux only use 19 out of the
20 bits of PID space, so existing guests will work if we make the host
use the top half of the 20 bits space.
We additionally add support for a property to indicate to Linux the
size of the PID register which will be useful if we eventually have
processors with a larger PID space available.
There is still an issue with malicious guests purposefully setting the
PID register to a value in the hosts PID range. Hopefully future HW
can prevent that, but in the meantime, we handle it with a pair of
kludges:
- On the way out of a guest, before we clear the current VCPU in the
PACA, we check the PID and if it's outside of the permitted range
we flush the TLB for that PID.
- When context switching, if the mm is "new" on that CPU (the
corresponding bit was set for the first time in the mm cpumask), we
check if any sibling thread is in KVM (has a non-NULL VCPU pointer
in the PACA). If that is the case, we also flush the PID for that
CPU (core).
This second part is needed to handle the case where a process is
migrated (or starts a new pthread) on a sibling thread of the CPU
coming out of KVM, as there's a window where stale translations can
exist before we detect it and flush them out.
A future optimization could be added by keeping track of whether the
PID has ever been used and avoid doing that for completely fresh PIDs.
We could similarily mark PIDs that have been the subject of a global
invalidation as "fresh". But for now this will do.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[mpe: Rework the asm to build with CONFIG_PPC_RADIX_MMU=n, drop
unneeded include of kvm_book3s_asm.h]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The z14 machine introduces new mode of the next-instruction-access-intent
NIAI instruction. With NIAI-8 it is possible to pin a cache-line on a
CPU for a small amount of time, NIAI-7 releases the cache-line again.
Finally NIAI-4 can be used to prevent the CPU to speculatively access
memory beyond the compare-and-swap instruction to get the lock.
Use these instruction in the spinlock code.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
attribute_group are not supposed to change at runtime. All functions
working with attribute_group provided by <linux/sysfs.h> work
with const attribute_group. So mark the non-const structs as const.
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
attribute_group are not supposed to change at runtime. All functions
working with attribute_group provided by <linux/sysfs.h> work
with const attribute_group. So mark the non-const structs as const.
File size before:
text data bss dec hex filename
11511 656 16 12183 2f97 drivers/s390/char/tape_core.o
File size After adding 'const':
text data bss dec hex filename
11575 592 16 12183 2f97 drivers/s390/char/tape_core.o
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
attribute_group are not supposed to change at runtime. All functions
working with attribute_group provided by <linux/sysfs.h> work
with const attribute_group. So mark the non-const structs as const.
File size before:
text data bss dec hex filename
8069 816 16 8901 22c5 drivers/s390/char/raw3270.o
File size After adding 'const':
text data bss dec hex filename
8133 752 16 8901 22c5 drivers/s390/char/raw3270.o
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
attribute_group are not supposed to change at runtime. All functions
working with attribute_group provided by <linux/sysfs.h> work
with const attribute_group. So mark the non-const structs as const.
File size before:
text data bss dec hex filename
6763 1216 0 7979 1f2b drivers/s390/net/qeth_l3_sys.o
File size After adding 'const':
text data bss dec hex filename
7019 960 0 7971 1f2b drivers/s390/net/qeth_l3_sys.o
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
