ata: ahci_brcm: Add back regulators management

[ Upstream commit 10340f8d7b6dd54e616339c8ccb2f397133ebea0 ]

While reworking the resources management and departing from using
ahci_platform_enable_resources() which did not allow a proper step
separation like we need, we unfortunately lost the ability to control
AHCI regulators. This broke some Broadcom STB systems that do expect
regulators to be turned on to link up with attached hard drives.

Fixes: c0cdf2ac4b5b ("ata: ahci_brcm: Fix AHCI resources management")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mm: proc: Invalidate TLB after clearing soft-dirty page state

[ Upstream commit 912efa17e5121693dfbadae29768f4144a3f9e62 ]

Since commit 0758cd830494 ("asm-generic/tlb: avoid potential double
flush"), TLB invalidation is elided in tlb_finish_mmu() if no entries
were batched via the tlb_remove_*() functions. Consequently, the
page-table modifications performed by clear_refs_write() in response to
a write to /proc/<pid>/clear_refs do not perform TLB invalidation.
Although this is fine when simply aging the ptes, in the case of
clearing the "soft-dirty" state we can end up with entries where
pte_write() is false, yet a writable mapping remains in the TLB.

Fix this by avoiding the mmu_gather API altogether: managing both the
'tlb_flush_pending' flag on the 'mm_struct' and explicit TLB
invalidation for the sort-dirty path, much like mprotect() does already.

Fixes: 0758cd830494 ("asm-generic/tlb: avoid potential double flush”)
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Yu Zhao <yuzhao@google.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Link: https://lkml.kernel.org/r/20210127235347.1402-2-will@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/nouveau: bail out of nouveau_channel_new if channel init fails

[ Upstream commit eaba3b28401f50e22d64351caa8afe8d29509f27 ]

Unprivileged user can crash kernel by using DRM_IOCTL_NOUVEAU_CHANNEL_ALLOC
ioctl. This was reported by trinity[1] fuzzer.

[   71.073906] nouveau 0000:01:00.0: crashme[1329]: channel failed to initialise, -17
[   71.081730] BUG: kernel NULL pointer dereference, address: 00000000000000a0
[   71.088928] #PF: supervisor read access in kernel mode
[   71.094059] #PF: error_code(0x0000) - not-present page
[   71.099189] PGD 119590067 P4D 119590067 PUD 1054f5067 PMD 0
[   71.104842] Oops: 0000 [#1] SMP NOPTI
[   71.108498] CPU: 2 PID: 1329 Comm: crashme Not tainted 5.8.0-rc6+ #2
[   71.114993] Hardware name: AMD Pike/Pike, BIOS RPK1506A 09/03/2014
[   71.121213] RIP: 0010:nouveau_abi16_ioctl_channel_alloc+0x108/0x380 [nouveau]
[   71.128339] Code: 48 89 9d f0 00 00 00 41 8b 4c 24 04 41 8b 14 24 45 31 c0 4c 8d 4b 10 48 89 ee 4c 89 f7 e8 10 11 00 00 85 c0 75 78 48 8b 43 10 <8b> 90 a0 00 00 00 41 89 54 24 08 80 7d 3d 05 0f 86 bb 01 00 00 41
[   71.147074] RSP: 0018:ffffb4a1809cfd38 EFLAGS: 00010246
[   71.152526] RAX: 0000000000000000 RBX: ffff98cedbaa1d20 RCX: 00000000000003bf
[   71.159651] RDX: 00000000000003be RSI: 0000000000000000 RDI: 0000000000030160
[   71.166774] RBP: ffff98cee776de00 R08: ffffdc0144198a08 R09: ffff98ceeefd4000
[   71.173901] R10: ffff98cee7e81780 R11: 0000000000000001 R12: ffffb4a1809cfe08
[   71.181214] R13: ffff98cee776d000 R14: ffff98cec519e000 R15: ffff98cee776def0
[   71.188339] FS:  00007fd926250500(0000) GS:ffff98ceeac80000(0000) knlGS:0000000000000000
[   71.196418] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   71.202155] CR2: 00000000000000a0 CR3: 0000000106622000 CR4: 00000000000406e0
[   71.209297] Call Trace:
[   71.211777]  ? nouveau_abi16_ioctl_getparam+0x1f0/0x1f0 [nouveau]
[   71.218053]  drm_ioctl_kernel+0xac/0xf0 [drm]
[   71.222421]  drm_ioctl+0x211/0x3c0 [drm]
[   71.226379]  ? nouveau_abi16_ioctl_getparam+0x1f0/0x1f0 [nouveau]
[   71.232500]  nouveau_drm_ioctl+0x57/0xb0 [nouveau]
[   71.237285]  ksys_ioctl+0x86/0xc0
[   71.240595]  __x64_sys_ioctl+0x16/0x20
[   71.244340]  do_syscall_64+0x4c/0x90
[   71.248110]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[   71.253162] RIP: 0033:0x7fd925d4b88b
[   71.256731] Code: Bad RIP value.
[   71.259955] RSP: 002b:00007ffc743592d8 EFLAGS: 00000206 ORIG_RAX: 0000000000000010
[   71.267514] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007fd925d4b88b
[   71.274637] RDX: 0000000000601080 RSI: 00000000c0586442 RDI: 0000000000000003
[   71.281986] RBP: 00007ffc74359340 R08: 00007fd926016ce0 R09: 00007fd926016ce0
[   71.289111] R10: 0000000000000003 R11: 0000000000000206 R12: 0000000000400620
[   71.296235] R13: 00007ffc74359420 R14: 0000000000000000 R15: 0000000000000000
[   71.303361] Modules linked in: rfkill sunrpc snd_hda_codec_realtek snd_hda_codec_generic ledtrig_audio snd_hda_intel snd_intel_dspcfg snd_hda_codec snd_hda_core edac_mce_amd snd_hwdep kvm_amd snd_seq ccp snd_seq_device snd_pcm kvm snd_timer snd irqbypass soundcore sp5100_tco pcspkr crct10dif_pclmul crc32_pclmul ghash_clmulni_intel wmi_bmof joydev i2c_piix4 fam15h_power k10temp acpi_cpufreq ip_tables xfs libcrc32c sd_mod t10_pi sg nouveau video mxm_wmi i2c_algo_bit drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops ttm broadcom bcm_phy_lib ata_generic ahci drm e1000 crc32c_intel libahci serio_raw tg3 libata firewire_ohci firewire_core wmi crc_itu_t dm_mirror dm_region_hash dm_log dm_mod
[   71.365269] CR2: 00000000000000a0

simplified reproducer
---------------------------------8<----------------------------------------
/*
 * gcc -o crashme crashme.c
 * ./crashme /dev/dri/renderD128
 */

struct drm_nouveau_channel_alloc {
uint32_t     fb_ctxdma_handle;
uint32_t     tt_ctxdma_handle;

int          channel;
uint32_t     pushbuf_domains;

/* Notifier memory */
uint32_t     notifier_handle;

/* DRM-enforced subchannel assignments */
struct {
uint32_t handle;
uint32_t grclass;
} subchan[8];
uint32_t nr_subchan;
};

static struct drm_nouveau_channel_alloc channel;

int main(int argc, char *argv[]) {
int fd;
int rv;

if (argc != 2)
die("usage: %s <dev>", 0, argv[0]);

if ((fd = open(argv[1], O_RDONLY)) == -1)
die("open %s", errno, argv[1]);

if (ioctl(fd, DRM_IOCTL_NOUVEAU_CHANNEL_ALLOC, &channel) == -1 &&
errno == EACCES)
die("ioctl %s", errno, argv[1]);

close(fd);

printf("PASS\n");

return 0;
}
---------------------------------8<----------------------------------------

[1] https://github.com/kernelslacker/trinity

Fixes: eeaf06ac1a55 ("drm/nouveau/svm: initial support for shared virtual memory")
Signed-off-by: Frantisek Hrbata <frantisek@hrbata.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

crypto: talitos - Fix ctr(aes) on SEC1

[ Upstream commit 43a942d27eaaf33bca560121cbe42f3637e92880 ]

While ctr(aes) requires the use of a special descriptor on SEC2 (see
commit 70d355ccea89 ("crypto: talitos - fix ctr-aes-talitos")), that
special descriptor doesn't work on SEC1, see commit e738c5f15562
("powerpc/8xx: Add DT node for using the SEC engine of the MPC885").

However, the common nonsnoop descriptor works properly on SEC1 for
ctr(aes).

Add a second template for ctr(aes) that will be registered
only on SEC1.

Fixes: 70d355ccea89 ("crypto: talitos - fix ctr-aes-talitos")
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Sasha Levin <sashal@kernel.org>

crypto: talitos - Work around SEC6 ERRATA (AES-CTR mode data size error)

[ Upstream commit 416b846757bcea20006a9197e67ba3a8b5b2a680 ]

Talitos Security Engine AESU considers any input
data size that is not a multiple of 16 bytes to be an error.
This is not a problem in general, except for Counter mode
that is a stream cipher and can have an input of any size.

Test Manager for ctr(aes) fails on 4th test vector which has
a length of 499 while all previous vectors which have a 16 bytes
multiple length succeed.

As suggested by Freescale, round up the input data length to the
nearest 16 bytes.

Fixes: 5e75ae1b3cef ("crypto: talitos - add new crypto modes")
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mtd: parser: imagetag: fix error codes in bcm963xx_parse_imagetag_partitions()

[ Upstream commit 12ba8f8ce29fdd277f3100052eddc1afd2f5ea3f ]

If the kstrtouint() calls fail, then this should return a negative
error code but it currently returns success.

Fixes: dd84cb022b31 ("mtd: bcm63xxpart: move imagetag parsing to its own parser")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Link: https://lore.kernel.org/linux-mtd/YBKFtNaFHGYBj+u4@mwanda
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf/arm-cmn: Move IRQs when migrating context

[ Upstream commit 1c8147ea89c883d1f4e20f1b1d9c879291430102 ]

If we migrate the PMU context to another CPU, we need to remember to
retarget the IRQs as well.

Fixes: 0ba64770a2f2 ("perf: Add Arm CMN-600 PMU driver")
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/e080640aea4ed8dfa870b8549dfb31221803eb6b.1611839564.git.robin.murphy@arm.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

perf/arm-cmn: Fix PMU instance naming

[ Upstream commit 79d7c3dca99fa96033695ddf5d495b775a3a137b ]

Although it's neat to avoid the suffix for the typical case of a
single PMU, it means systems with multiple CMN instances end up with
inconsistent naming. I think it also breaks perf tool's "uncore alias"
logic if the common instance prefix is also the full name of one.

Avoid any surprises by not trying to be clever and simply numbering
every instance, even when it might technically prove redundant.

Fixes: 0ba64770a2f2 ("perf: Add Arm CMN-600 PMU driver")
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Link: https://lore.kernel.org/r/649a2281233f193d59240b13ed91b57337c77b32.1611839564.git.robin.murphy@arm.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: SOF: Intel: hda: cancel D0i3 work during runtime suspend

[ Upstream commit 0084364d9678e9d722ee620ed916f2f9954abdbf ]

Cancel the D0i3 work during runtime suspend as no streams are
active at this point anyway.

Fixes: 63e51fd33fef ("ASoC: SOF: Intel: cnl: Implement feature to support DSP D0i3 in S0")
Signed-off-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Signed-off-by: Kai Vehmanen <kai.vehmanen@linux.intel.com>
Link: https://lore.kernel.org/r/20210128092345.1033085-1-kai.vehmanen@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: qcom: lpass-cpu: Remove bit clock state check

[ Upstream commit 6c28377b7114d04cf82eedffe9dcc8fa66ecec48 ]

No need of BCLK state maintenance from driver side as
clock_enable and clk_disable API's maintaing state counter.

One of the major issue was spotted when Headset jack inserted
while playback continues, due to same PCM device node opens twice
for playaback/capture and closes once for capture and playback continues.

It can resolve the errors in such scenarios.

Fixes: b1824968221c ("ASoC: qcom: Fix enabling BCLK and LRCLK in LPAIF invalid state")

Signed-off-by: Srinivasa Rao Mandadapu <srivasam@codeaurora.org>
Reviewed-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Link: https://lore.kernel.org/r/20210127151824.8929-1-srivasam@codeaurora.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

f2fs: compress: fix potential deadlock

[ Upstream commit 3afae09ffea5e08f523823be99a784675995d6bb ]

generic/269 reports a hangtask issue, the root cause is ABBA deadlock
described as below:

Thread A Thread B
- down_write(&sbi->gc_lock) -- A
- f2fs_write_data_pages
 - lock all pages in cluster -- B
 - f2fs_write_multi_pages
  - f2fs_write_raw_pages
   - f2fs_write_single_data_page
    - f2fs_balance_fs
     - down_write(&sbi->gc_lock) -- A
- f2fs_gc
 - do_garbage_collect
  - ra_data_block
   - pagecache_get_page -- B

To fix this, it needs to avoid calling f2fs_balance_fs() if there is
still cluster pages been locked in context of cluster writeback, so
instead, let's call f2fs_balance_fs() in the end of
f2fs_write_raw_pages() when all cluster pages were unlocked.

Fixes: 4c8ff7095bef ("f2fs: support data compression")
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched/eas: Don't update misfit status if the task is pinned

[ Upstream commit 0ae78eec8aa64e645866e75005162603a77a0f49 ]

If the task is pinned to a cpu, setting the misfit status means that
we'll unnecessarily continuously attempt to migrate the task but fail.

This continuous failure will cause the balance_interval to increase to
a high value, and eventually cause unnecessary significant delays in
balancing the system when real imbalance happens.

Caught while testing uclamp where rt-app calibration loop was pinned to
cpu 0, shortly after which we spawn another task with high util_clamp
value. The task was failing to migrate after over 40ms of runtime due to
balance_interval unnecessary expanded to a very high value from the
calibration loop.

Not done here, but it could be useful to extend the check for pinning to
verify that the affinity of the task has a cpu that fits. We could end
up in a similar situation otherwise.

Fixes: 3b1baa6496e6 ("sched/fair: Add 'group_misfit_task' load-balance type")
Signed-off-by: Qais Yousef <qais.yousef@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Quentin Perret <qperret@google.com>
Acked-by: Valentin Schneider <valentin.schneider@arm.com>
Link: https://lkml.kernel.org/r/20210119120755.2425264-1-qais.yousef@arm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: uvcvideo: Accept invalid bFormatIndex and bFrameIndex values

[ Upstream commit dc9455ffae02d7b7fb51ba1e007fffcb9dc5d890 ]

The Renkforce RF AC4K 300 Action Cam 4K reports invalid bFormatIndex and
bFrameIndex values when negotiating the video probe and commit controls.
The UVC descriptors report a single supported format and frame size,
with bFormatIndex and bFrameIndex both equal to 2, but the video probe
and commit controls report bFormatIndex and bFrameIndex set to 1.

The device otherwise operates correctly, but the driver rejects the
values and fails the format try operation. Fix it by ignoring the
invalid indices, and assuming that the format and frame requested by the
driver are accepted by the device.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=210767
Fixes: 8a652a17e3c0 ("media: uvcvideo: Ensure all probed info is returned to v4l2")
Reported-by: Till Dörges <doerges@pre-sense.de>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: pxa_camera: declare variable when DEBUG is defined

[ Upstream commit 031b9212eeee365443aaef013360ea6cded7b2c4 ]

When DEBUG is defined this error occurs

drivers/media/platform/pxa_camera.c:1410:7: error:
  ‘i’ undeclared (first use in this function)
  for (i = 0; i < vb->num_planes; i++)
       ^
The variable 'i' is missing, so declare it.

Fixes: 6f28435d1c15 ("[media] media: platform: pxa_camera: trivial move of functions")
Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: mtk-vcodec: fix argument used when DEBUG is defined

[ Upstream commit a04e187d231086a1313fd635ac42bdbc997137ad ]

When DEBUG is defined this error occurs

drivers/media/platform/mtk-vcodec/mtk_vcodec_enc_drv.c:306:41:
  error: ‘i’ undeclared (first use in this function)
  mtk_v4l2_debug(2, "reg[%d] base=0x%p", i, dev->reg_base[VENC_SYS]);

Reviewing the old line

mtk_v4l2_debug(2, "reg[%d] base=0x%p", i, dev->reg_base[i]);

All the i's need to be changed to VENC_SYS.
Fix a similar error for VENC_LT_SYS.

Fixes: 0dc4b3286125 ("media: mtk-vcodec: venc: support SCP firmware")
Signed-off-by: Tom Rix <trix@redhat.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: cx25821: Fix a bug when reallocating some dma memory

[ Upstream commit b2de3643c5024fc4fd128ba7767c7fb8b714bea7 ]

This function looks like a realloc.

However, if 'risc->cpu != NULL', the memory will be freed, but never
reallocated with the bigger 'size'.
Explicitly set 'risc->cpu' to NULL, so that the reallocation is
correctly performed a few lines below.

[hverkuil: NULL != risc->cpu -> risc->cpu]

Fixes: 5ede94c70553 ("[media] cx25821: remove bogus btcx_risc dependency)
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: qm1d1c0042: fix error return code in qm1d1c0042_init()

[ Upstream commit fcf8d018bdca0453b8d6359062e6bc1512d04c38 ]

Fix to return a negative error code from the error handling case
instead of 0 in function qm1d1c0042_init(), as done elsewhere
in this function.

Fixes: ab4d14528fdf ("[media] em28xx: add support for PLEX PX-BCUD (ISDB-S)")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Luo Meng <luomeng12@huawei.com>
Acked-by: Akihiro Tsukada <tskd08@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: atomisp: Fix a buffer overflow in debug code

[ Upstream commit 625993166b551d633917ca35d4afb7b46d7451b4 ]

The "pad" variable is a user controlled string and we haven't properly
clamped it at this point so the debug code could print from beyond the
of the array.

Fixes: a49d25364dfb ("staging/atomisp: Add support for the Intel IPU v2")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: vidtv: psi: fix missing crc for PMT

[ Upstream commit 0a933a7f73d6c545dcbecb4f7a92d272aef4417b ]

The PMT write function was refactored and this broke the CRC computation.

Fix it.

Fixes: db9569f67e2e ("media: vidtv: cleanup PMT write table function")
Signed-off-by: Daniel W. S. Almeida <dwlsalmeida@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: lmedm04: Fix misuse of comma

[ Upstream commit 59a3e78f8cc33901fe39035c1ab681374bba95ad ]

There's a comma used instead of a semicolon that causes multiple
statements to be executed after an if instead of just the intended
single statement.

Replace the comma with a semicolon.

Fixes: 15e1ce33182d ("[media] lmedm04: Fix usb_submit_urb BOGUS urb xfer, pipe 1 != type 3 in interrupt urb")
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: software_node: Fix refcounts in software_node_get_next_child()

[ Upstream commit fb5ec981adf08b94e6ce27ca16b7765c94f4513c ]

The software_node_get_next_child() function currently does not hold
references to the child software_node that it finds or put the ref that
is held against the old child - fix that.

Fixes: 59abd83672f7 ("drivers: base: Introducing software nodes to the firmware node framework")
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Daniel Scally <djrscally@gmail.com>
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Fix HDMI deep color output for DCE 6-11.

[ Upstream commit efa18405baa55a864c61d2f3cc6fe4d363818eb3 ]

This fixes corrupted display output in HDMI deep color
10/12 bpc mode at least as observed on AMD Mullins, DCE-8.3.

It will hopefully also provide fixes for other DCE's up to
DCE-11, assuming those will need similar fixes, but i could
not test that for HDMI due to lack of suitable hw, so viewer
discretion is advised.

dce110_stream_encoder_hdmi_set_stream_attribute() is used for
HDMI setup on all DCE's and is missing color_depth assignment.

dce110_program_pix_clk() is used for pixel clock setup on HDMI
for DCE 6-11, and is missing color_depth assignment.

Additionally some of the underlying Atombios specific encoder
and pixelclock setup functions are missing code which is in
the classic amdgpu kms modesetting path and the in the radeon
kms driver for DCE6/DCE8.

encoder_control_digx_v3() - Was missing setup code wrt. amdgpu
and radeon kms classic drivers. Added here, but untested due to
lack of suitable test hw.

encoder_control_digx_v4() - Added missing setup code.
Successfully tested on AMD mullins / DCE-8.3 with HDMI deep color
output at 10 bpc and 12 bpc.

Note that encoder_control_digx_v5() has proper setup code in place
and is used, e.g., by DCE-11.2, but this code wasn't used for deep
color setup due to the missing cntl.color_depth setup in the calling
function for HDMI.

set_pixel_clock_v5() - Missing setup code wrt. classic amdgpu/radeon
kms. Added here, but untested due to lack of hw.

set_pixel_clock_v6() - Missing setup code added. Successfully tested
on AMD mullins DCE-8.3. This fixes corrupted display output at HDMI
deep color output with 10 bpc or 12 bpc.

Fixes: 4562236b3bc0 ("drm/amd/dc: Add dc display driver (v2)")

Reviewed-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd/display: Fix 10/12 bpc setup in DCE output bit depth reduction.

[ Upstream commit 1916866dfa4aaceba1a70db83fde569387649d93 ]

In set_clamp(), the comments and definitions for the COLOR_DEPTH_101010
and COLOR_DEPTH_121212 cases directly contradict the code comment which
explains how this should work, whereas the COLOR_DEPTH_888 case
is consistent with the code comments. Comment says the bitmask should
be chosen to align to the top-most 10 or 12 MSB's on a 14 bit bus, but
the implementation contradicts that: 10 bit case sets a mask for 12 bpc
clamping, whereas 12 bit case sets a mask for 14 bpc clamping.

Note that during my limited testing on DCE-8.3 (HDMI deep color)
and DCE-11.2 (DP deep color), this didn't have any obvious ill
effects, neither did fixing it change anything obvious for the
better, so this fix may be inconsequential on DCE, and just
reduce the confusion of innocent bystanders when reading the code
and trying to investigate problems with 10 bpc+ output.

Fixes: 4562236b3bc0 ("drm/amd/dc: Add dc display driver (v2)")

Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Cc: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

macintosh/adb-iop: Use big-endian autopoll mask

[ Upstream commit c396dd2ec5bbd1cb80eafe32a72ab6bd6b17cb5a ]

As usual, the available documentation is inadequate and leaves endianness
unspecified for message data. However, testing shows that this patch does
improve correctness. The mistake should have been detected earlier but it
was obscured by other bugs. In testing, this patch reinstated pre-v5.9
behaviour. The old driver bugs remain and ADB input devices may stop
working. But that appears to be unrelated.

Cc: Joshua Thompson <funaho@jurai.org>
Fixes: c66da95a39ec ("macintosh/adb-iop: Implement SRQ autopolling")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Link: https://lore.kernel.org/r/20210125074524.3027452-1-geert@linux-m68k.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

bsg: free the request before return error code

[ Upstream commit 0f7b4bc6bb1e57c48ef14f1818df947c1612b206 ]

Free the request rq before returning error code.

Fixes: 972248e9111e ("scsi: bsg-lib: handle bidi requests without block layer help")
Signed-off-by: Pan Bian <bianpan2016@163.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amdgpu: toggle on DF Cstate after finishing xgmi injection

[ Upstream commit fe2d9f5abf19f2b3688b3b8da4e42f8d07886847 ]

Fixes: 5c23e9e05e42 ("drm/amdgpu: Update RAS XGMI error inject sequence")
Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/tegra: Fix reference leak when pm_runtime_get_sync() fails

[ Upstream commit dcdfe2712b68f1e9dbf4f1a96ad59b80e5cc0ef7 ]

The PM reference count is not expected to be incremented on return in
these Tegra functions.

However, pm_runtime_get_sync() will increment the PM reference count
even on failure. Forgetting to put the reference again will result in
a leak.

Replace it with pm_runtime_resume_and_get() to keep the usage counter
balanced.

Fixes: fd67e9c6ed5a ("drm/tegra: Do not implement runtime PM")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Qinglang Miao <miaoqinglang@huawei.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

MIPS: Compare __SYNC_loongson3_war against 0

[ Upstream commit 8790ccf8daf1a8c53b6cb8ce0c9a109274bd3fa8 ]

When building with clang when CONFIG_CPU_LOONGSON3_WORKAROUNDS is
enabled:

 In file included from lib/errseq.c:4:
 In file included from ./include/linux/atomic.h:7:
 ./arch/mips/include/asm/atomic.h:52:1: warning: converting the result of
 '<<' to a boolean always evaluates to true
 [-Wtautological-constant-compare]
 ATOMIC_OPS(atomic64, s64)
 ^
 ./arch/mips/include/asm/atomic.h:40:9: note: expanded from macro
 'ATOMIC_OPS'
         return cmpxchg(&v->counter, o, n);
                ^
 ./arch/mips/include/asm/cmpxchg.h:194:7: note: expanded from macro
 'cmpxchg'
         if (!__SYNC_loongson3_war)
              ^
 ./arch/mips/include/asm/sync.h:147:34: note: expanded from macro
 '__SYNC_loongson3_war'
 # define __SYNC_loongson3_war   (1 << 31)
                                    ^

While it is not wrong that the result of this shift is always true in a
boolean context, it is not a problem here. Regardless, the warning is
really noisy so rather than making the shift a boolean implicitly, use
it in an equality comparison so the shift is used as an integer value.

Fixes: 4d1dbfe6cbec ("MIPS: atomic: Emit Loongson3 sync workarounds within asm")
Fixes: a91f2a1dba44 ("MIPS: cmpxchg: Omit redundant barriers for Loongson3")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Acked-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

MIPS: properly stop .eh_frame generation

[ Upstream commit 894ef530012fb5078466efdfb2c15d8b2f1565cd ]

Commit 866b6a89c6d1 ("MIPS: Add DWARF unwinding to assembly") added
-fno-asynchronous-unwind-tables to KBUILD_CFLAGS to prevent compiler
from emitting .eh_frame symbols.
However, as MIPS heavily uses CFI, that's not enough. Use the
approach taken for x86 (as it also uses CFI) and explicitly put CFI
symbols into the .debug_frame section (except for VDSO).
This allows us to drop .eh_frame from DISCARDS as it's no longer
being generated.

Fixes: 866b6a89c6d1 ("MIPS: Add DWARF unwinding to assembly")
Suggested-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Alexander Lobakin <alobakin@pm.me>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: ti-vpe: cal: fix write to unallocated memory

[ Upstream commit 5a402af5e19f215689e8bf3cc244c21d94eba3c4 ]

The asd allocated with v4l2_async_notifier_add_fwnode_subdev() must be
of size cal_v4l2_async_subdev, otherwise access to
cal_v4l2_async_subdev->phy will go to unallocated memory.

Fixes: 8fcb7576ad19 ("media: ti-vpe: cal: Allow multiple contexts per subdev notifier")
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: imx7: csi: Fix pad link validation

[ Upstream commit f5ffb81f51376eb5a12e8c4cb4871426c65bb2af ]

We can not make the assumption that the bound subdev is always a CSI
mux, in i.MX6UL/i.MX6ULL that is not the case. So, just get the entity
selected by source directly upstream from the CSI.

Fixes: 86e02d07871c ("media: imx5/6/7: csi: Mark a bound video mux as a CSI mux")
Reported-by: Fabio Estevam <festevam@gmail.com>
Signed-off-by: Rui Miguel Silva <rmfrfs@gmail.com>
Tested-by: Fabio Estevam <festevam@gmail.com>
Tested-by: Sébastien Szymanski <sebastien.szymanski@armadeus.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: imx7: csi: Fix regression for parallel cameras on i.MX6UL

[ Upstream commit 9bac67214fbf4b5f23463f7742ccf69bfe684cbd ]

Commit 86e02d07871c ("media: imx5/6/7: csi: Mark a bound video mux as
a CSI mux") made an incorrect assumption that for imx7-media-csi,
the bound subdev must always be a CSI mux. On i.MX6UL/i.MX6ULL there
is no CSI mux at all, so do not return an error when the entity is not a
video mux and assign the IMX_MEDIA_GRP_ID_CSI_MUX group id only when
appropriate.

This is the same approach as done in imx-media-csi.c and it fixes the
csi probe regression on i.MX6UL.

Tested on a imx6ull-evk board.

Fixes: 86e02d07871c ("media: imx5/6/7: csi: Mark a bound video mux as a CSI mux")
Signed-off-by: Fabio Estevam <festevam@gmail.com>
Signed-off-by: Rui Miguel Silva <rmfrfs@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/sun4i: tcon: fix inverted DCLK polarity

[ Upstream commit 67f4aeb2b41a0629abde3794d463547f60b0cbdd ]

During commit 88bc4178568b ("drm: Use new
DRM_BUS_FLAG_*_(DRIVE|SAMPLE)_(POS|NEG)EDGE flags") DRM_BUS_FLAG_*
macros have been changed to avoid ambiguity but just because of this
ambiguity previous DRM_BUS_FLAG_PIXDATA_(POS/NEG)EDGE were used meaning
_SAMPLE_ not _DRIVE_. This leads to DLCK inversion and need to fix but
instead of swapping phase values, let's adopt an easier approach Maxime
suggested:
It turned out that bit 26 of SUN4I_TCON0_IO_POL_REG is dedicated to
invert DCLK polarity and this makes things really easier than before. So
let's handle DCLK polarity by adding SUN4I_TCON0_IO_POL_DCLK_DRIVE_NEGEDGE
as bit 26 and activating according to bus_flags the same way it is done
for all the other signals polarity.

Fixes: 88bc4178568b ("drm: Use new DRM_BUS_FLAG_*_(DRIVE|SAMPLE)_(POS|NEG)EDGE flags")
Suggested-by: Maxime Ripard <maxime@cerno.tech>
Signed-off-by: Giulio Benetti <giulio.benetti@micronovasrl.com>
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20210114081732.9386-1-giulio.benetti@benettiengineering.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched/fair: Avoid stale CPU util_est value for schedutil in task dequeue

[ Upstream commit 8c1f560c1ea3f19e22ba356f62680d9d449c9ec2 ]

CPU (root cfs_rq) estimated utilization (util_est) is currently used in
dequeue_task_fair() to drive frequency selection before it is updated.

with:

CPU_util        : rq->cfs.avg.util_avg
CPU_util_est    : rq->cfs.avg.util_est
CPU_utilization : max(CPU_util, CPU_util_est)
task_util       : p->se.avg.util_avg
task_util_est   : p->se.avg.util_est

dequeue_task_fair():

    /* (1) CPU_util and task_util update + inform schedutil about
           CPU_utilization changes */
    for_each_sched_entity() /* 2 loops */
        (dequeue_entity() ->) update_load_avg() -> cfs_rq_util_change()
         -> cpufreq_update_util() ->...-> sugov_update_[shared\|single]
         -> sugov_get_util() -> cpu_util_cfs()

    /* (2) CPU_util_est and task_util_est update */
    util_est_dequeue()

cpu_util_cfs() uses CPU_utilization which could lead to a false (too
high) utilization value for schedutil in task ramp-down or ramp-up
scenarios during task dequeue.

To mitigate the issue split the util_est update (2) into:

 (A) CPU_util_est update in util_est_dequeue()
 (B) task_util_est update in util_est_update()

Place (A) before (1) and keep (B) where (2) is. The latter is necessary
since (B) relies on task_util update in (1).

Fixes: 7f65ea42eb00 ("sched/fair: Add util_est on top of PELT")
Signed-off-by: Xuewen Yan <xuewen.yan@unisoc.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Reviewed-by: Vincent Guittot <vincent.guittot@linaro.org>
Link: https://lkml.kernel.org/r/1608283672-18240-1-git-send-email-xuewen.yan94@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

crypto: bcm - Rename struct device_private to bcm_device_private

[ Upstream commit f7f2b43eaf6b4cfe54c75100709be31d5c4b52c8 ]

Renaming 'struct device_private' to 'struct bcm_device_private',
because it clashes with 'struct device_private' from
'drivers/base/base.h'.

While it's not a functional problem, it's causing two distinct
type hierarchies in BTF data. It also breaks build with options:
  CONFIG_DEBUG_INFO_BTF=y
  CONFIG_CRYPTO_DEV_BCM_SPU=y

as reported by Qais Yousef [1].

[1] https://lore.kernel.org/lkml/20201229151352.6hzmjvu3qh6p2qgg@e107158-lin/

Fixes: 9d12ba86f818 ("crypto: brcm - Add Broadcom SPU driver")
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Qais Yousef <qais.yousef@arm.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Sasha Levin <sashal@kernel.org>

evm: Fix memleak in init_desc

[ Upstream commit ccf11dbaa07b328fa469415c362d33459c140a37 ]

tmp_tfm is allocated, but not freed on subsequent kmalloc failure, which
leads to a memory leak.  Free tmp_tfm.

Fixes: d46eb3699502b ("evm: crypto hash replaced by shash")
Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
[zohar@linux.ibm.com: formatted/reworded patch description]
Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: qcom: qdsp6: Move frontend AIFs to q6asm-dai

[ Upstream commit 6fd8d2d275f74baa7ac17b2656da1235f56dab99 ]

At the moment it is necessary to set up the DAPM routes between
front-end AIF<->DAI explicitly in the device tree, e.g. using

audio-routing =
"MM_DL1", "MultiMedia1 Playback",
"MM_DL3", "MultiMedia3 Playback",
"MM_DL4", "MultiMedia4 Playback",
"MultiMedia2 Capture", "MM_UL2";

This is prone to mistakes and (sadly) there is no clear error if one
of these routes is missing. :(

Actually, this should not be necessary because the ASoC core normally
automatically links AIF<->DAI within snd_soc_dapm_link_dai_widgets().
This is done using the "stname" parameter of SND_SOC_DAPM_AIF_IN/OUT.

For SND_SOC_DAPM_AIF_IN("MM_DL1", "MultiMedia1 Playback", 0, 0, 0, 0),
it should create the route from above: MM_DL1 <-> MultiMedia1 Playback.

This does not work at the moment because the AIF widget (MM_DL1)
and the DAI widget (MultiMedia1 Playback) belong to different
DAPM contexts (q6routing / q6asm-dai).

Fix this by declaring the AIF widgets in the same driver as the DAIs
(q6asm-dai). Now the routes above are created automatically
and no longer need to be specified in the device tree.

This is also more consistent with the back-end AIFs which are already
declared in q6afe-dais instead of q6routing. q6routing should only link
the components together using mixers.

Cc: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Fixes: 2a9e92d371db ("ASoC: qdsp6: q6asm: Add q6asm dai driver")
Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
Link: https://lore.kernel.org/r/20201211203255.148246-1-stephan@gerhold.net
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ASoC: cs42l56: fix up error handling in probe

[ Upstream commit 856fe64da84c95a1d415564b981ae3908eea2a76 ]

There are two issues with this code.  The first error path forgot to set
the error code and instead returns success.  The second error path
doesn't clean up.

Fixes: 272b5edd3b8f ("ASoC: Add support for CS42L56 CODEC")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/X9NE/9nK9/TuxuL+@mwanda
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: aspeed: fix error return code in aspeed_video_setup_video()

[ Upstream commit d497fcdab02996a4510d5dd0d743447c737c317a ]

Fix to return a negative error code from the error handling
case instead of 0, as done elsewhere in this function.

Fixes: d2b4387f3bdf ("media: platform: Add Aspeed Video Engine driver")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zhang Changzhong <zhangchangzhong@huawei.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: tm6000: Fix memleak in tm6000_start_stream

[ Upstream commit 76aaf8a96771c16365b8510f1fb97738dc88026e ]

When usb_clear_halt() fails, dvb->bulk_urb->transfer_buffer
and dvb->bulk_urb should be freed just like when
usb_submit_urb() fails.

Fixes: 3169c9b26fffa ("V4L/DVB (12788): tm6000: Add initial DVB-T support")
Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: media/pci: Fix memleak in empress_init

[ Upstream commit 15d0c52241ecb1c9d802506bff6f5c3f7872c0df ]

When vb2_queue_init() fails, dev->empress_dev
should be released just like other error handling
paths.

Fixes: 2ada815fc48bb ("[media] saa7134: convert to vb2")
Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: em28xx: Fix use-after-free in em28xx_alloc_urbs

[ Upstream commit a26efd1961a18b91ae4cd2e433adbcf865b40fa3 ]

When kzalloc() fails, em28xx_uninit_usb_xfer() will free
usb_bufs->buf and set it to NULL. Thus the later access
to usb_bufs->buf[i] will lead to null pointer dereference.
Also the kfree(usb_bufs->buf) after that is redundant.

Fixes: d571b592c6206 ("media: em28xx: don't use coherent buffer for DMA transfers")
Signed-off-by: Dinghao Liu <dinghao.liu@zju.edu.cn>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: vsp1: Fix an error handling path in the probe function

[ Upstream commit 7113469dafc2d545fa4fa9bc649c31dc27db492e ]

A previous 'rcar_fcp_get()' call must be undone in the error handling path,
as already done in the remove function.

Fixes: 94fcdf829793 ("[media] v4l: vsp1: Add FCP support")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: camss: missing error code in msm_video_register()

[ Upstream commit 9c67ed2ab299123872be14a3dc2ea44ce7e4538b ]

This error path returns success but it should return -EINVAL.

Fixes: cba3819d1e93 ("media: camss: Format configuration per hardware version")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Robert Foss <robert.foss@linaro.org>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: mtk-vcodec: fix error return code in vdec_vp9_decode()

[ Upstream commit 4397efebf039be58e98c81a194a26100eba597bb ]

Fix to return a negative error code from the error handling
case instead of 0, as done elsewhere in this function.

Fixes: dea42fb79f4f ("media: mtk-vcodec: reset segment data then trig decoder")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zhang Changzhong <zhangchangzhong@huawei.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: imx: Fix csc/scaler unregister

[ Upstream commit 89b14485caa4b7b2eaf70be0064f0978e68ebeee ]

The csc/scaler device private struct is released by
ipu_csc_scaler_video_device_release(), which can be called
by video_unregister_device() if there are no users
of the underlying struct video device.

Therefore, the mutex can't be held when calling
video_unregister_device() as its memory may be freed
by it, leading to a kernel oops.

Fortunately, the fix is quite simple as no locking
is needed when calling video_unregister_device(): v4l2-core
already has its own internal locking, and the structures
are also properly refcounted.

Fixes: a8ef0488cc59 ("media: imx: add csc/scaler mem2mem device")
Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: imx: Unregister csc/scaler only if registered

[ Upstream commit bb2216548a2b13cf2942a058b475438a7a6bb028 ]

The csc/scaler device pointer (imxmd->m2m_vdev) is assigned
after the imx media device v4l2-async probe completes,
therefore we need to check if the device is non-NULL
before trying to unregister it.

This can be the case if the non-completed imx media device
is unbinded (or the driver is removed), leading to a kernel oops.

Fixes: a8ef0488cc59 ("media: imx: add csc/scaler mem2mem device")
Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: i2c: ov5670: Fix PIXEL_RATE minimum value

[ Upstream commit dc1eb7c9c290cba52937c9a224b22a400bb0ffd7 ]

The driver currently reports a single supported value for
V4L2_CID_PIXEL_RATE and initializes the control's minimum value to 0,
which is very risky, as userspace might accidentally use it as divider
when calculating the time duration of a line.

Fix this by using as minimum the only supported value when registering
the control.

Fixes: 5de35c9b8dcd1 ("media: i2c: Add Omnivision OV5670 5M sensor support")
Signed-off-by: Jacopo Mondi <jacopo@jmondi.org>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: ipu3-cio2: Build only for x86

[ Upstream commit 3ef5e42d281ea108f4ccdca186de4ce20a346326 ]

According to the original code in the driver it was never assumed to work
with big page sizes: unsigned short type followed by PAGE_SHIFT and
PAGE_MASK which may be different on non-x86 architectures.

Recently LKP found an issue on non-x86 architectures due to above
mentioned limitations. Since Sakari acknowledges that it's not really
useful to be able to compile this elsewhere, mark it x86 only.

Fixes: a31d19f88932 ("media: ipu3: allow building it with COMPILE_TEST on non-x86 archs")
Reported-by: kernel test robot <lkp@intel.com>
Suggested-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/fourcc: fix Amlogic format modifier masks

[ Upstream commit cc3283f8f41f741fbaef63d0503d8fb4a7919100 ]

The comment says the layout and options use 8 bits, and the shift
uses 8 bits. However the mask is 0xf, ie. 0b00001111 (4 bits).

This could be surprising when introducing new layouts or options
that take more than 4 bits, as this would silently drop the high
bits.

Make the masks consistent with the comment and the shift.

Found when writing a drm_info patch [1].

[1]: https://github.com/ascent12/drm_info/pull/67

Signed-off-by: Simon Ser <contact@emersion.fr>
Fixes: d6528ec88309 ("drm/fourcc: Add modifier definitions for describing Amlogic Video Framebuffer Compression")
Cc: Neil Armstrong <narmstrong@baylibre.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Kevin Hilman <khilman@baylibre.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Neil Armstrong <narmstrong@baylibre.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210110125103.15447-1-contact@emersion.fr
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/virtio: make sure context is created in gem open

[ Upstream commit 8aeef9d4f48917ce85710949b079548974b4a638 ]

The context might still be missing when DRM_IOCTL_PRIME_FD_TO_HANDLE is
the first ioctl on the drm_file.

Fixes: 72b48ae800da ("drm/virtio: enqueue virtio_gpu_create_context after the first 3D ioctl")
Cc: Gurchetan Singh <gurchetansingh@chromium.org>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20210107210726.269584-1-olvaffe@gmail.com
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

MIPS: lantiq: Explicitly compare LTQ_EBU_PCC_ISTAT against 0

[ Upstream commit c6f2a9e17b9bef7677caddb1626c2402f3e9d2bd ]

When building xway_defconfig with clang:

arch/mips/lantiq/irq.c:305:48: error: use of logical '&&' with constant
operand [-Werror,-Wconstant-logical-operand]
        if ((irq == LTQ_ICU_EBU_IRQ) && (module == 0) && LTQ_EBU_PCC_ISTAT)
                                                      ^ ~~~~~~~~~~~~~~~~~
arch/mips/lantiq/irq.c:305:48: note: use '&' for a bitwise operation
        if ((irq == LTQ_ICU_EBU_IRQ) && (module == 0) && LTQ_EBU_PCC_ISTAT)
                                                      ^~
                                                      &
arch/mips/lantiq/irq.c:305:48: note: remove constant to silence this
warning
        if ((irq == LTQ_ICU_EBU_IRQ) && (module == 0) && LTQ_EBU_PCC_ISTAT)
                                                     ~^~~~~~~~~~~~~~~~~~~~
1 error generated.

Explicitly compare the constant LTQ_EBU_PCC_ISTAT against 0 to fix the
warning. Additionally, remove the unnecessary parentheses as this is a
simple conditional statement and shorthand '== 0' to '!'.

Fixes: 3645da0276ae ("OF: MIPS: lantiq: implement irq_domain support")
Link: https://github.com/ClangBuiltLinux/linux/issues/807
Reported-by: Dmitry Golovin <dima@golovin.in>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

MIPS: c-r4k: Fix section mismatch for loongson2_sc_init

[ Upstream commit c58734eee6a2151ba033c0dcb31902c89e310374 ]

When building with clang, the following section mismatch warning occurs:

WARNING: modpost: vmlinux.o(.text+0x24490): Section mismatch in
reference from the function r4k_cache_init() to the function
.init.text:loongson2_sc_init()

This should have been fixed with commit ad4fddef5f23 ("mips: fix Section
mismatch in reference") but it was missed. Remove the improper __init
annotation like that commit did.

Fixes: 078a55fc824c ("MIPS: Delete __cpuinit/__CPUINIT usage from MIPS code")
Link: https://github.com/ClangBuiltLinux/linux/issues/787
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Reviewed-by: Huacai Chen <chenhuacai@kernel.org>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amdgpu: Fix macro name _AMDGPU_TRACE_H_ in preprocessor if condition

[ Upstream commit 956e20eb0fbb206e5e795539db5469db099715c8 ]

Add an underscore in amdgpu_trace.h line 24 "_AMDGPU_TRACE_H".

Fixes: d38ceaf99ed0 ("drm/amdgpu: add core driver (v4)")
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
Signed-off-by: Chenyang Li <lichenyang@loongson.cn>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm: rcar-du: Fix the return check of of_parse_phandle and of_find_device_by_node

[ Upstream commit 8d7d33f6be06f929ac2c5e8ea2323fec272790d4 ]

of_parse_phandle and of_find_device_by_node may return NULL
which cannot be checked by IS_ERR.

Fixes: 8de707aeb452 ("drm: rcar-du: kms: Initialize CMM instances")
Signed-off-by: Wang Xiaojun <wangxiaojun11@huawei.com>
Reported-by: Hulk Robot <hulkci@huawei.com>
Acked-by: Jacopo Mondi <jacopo+renesas@jmondi.org>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
[Replace -ENODEV with -EINVAL]

Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm: rcar-du: Fix crash when using LVDS1 clock for CRTC

[ Upstream commit 53ced169373aab52d3b5da0fee6a342002d1876d ]

On D3 and E3 platforms, the LVDS encoder includes a PLL that can
generate a clock for the corresponding CRTC, used even when the CRTC
output to a non-LVDS port. This mechanism is supported by the driver,
but the implementation is broken in dual-link LVDS mode. In that case,
the LVDS1 drm_encoder is skipped, which causes a crash when trying to
access its bridge later on.

Fix this by storing bridge pointers internally instead of retrieving
them from the encoder. The rcar_du_device encoders field isn't used
anymore and can be dropped.

Fixes: 8e8fddab0d0a ("drm: rcar-du: Skip LVDS1 output on Gen3 when using dual-link LVDS mode")
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo+renesas@jmondi.org>
Reviewed-by: Kieran Bingham <kieran.bingham+renesas@ideasonboard.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm: rcar-du: Fix PM reference leak in rcar_cmm_enable()

[ Upstream commit 136ce7684bc1ff4a088812f600c63daca50b32c2 ]

pm_runtime_get_sync will increment pm usage counter even it failed.
Forgetting to putting operation will result in a reference leak here.

A new function pm_runtime_resume_and_get is introduced in [0] to keep
usage counter balanced. So We fix the reference leak by replacing it
with new funtion.

[0] dd8088d5a896 ("PM: runtime: Add  pm_runtime_resume_and_get to deal with usage counter")

Fixes: e08e934d6c28 ("drm: rcar-du: Add support for CMM")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Qinglang Miao <miaoqinglang@huawei.com>
Acked-by: Jacopo Mondi <jacopo@jmondi.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

kcsan: Rewrite kcsan_prandom_u32_max() without prandom_u32_state()

[ Upstream commit 71a076f4a61a6c779794ad286f356b39725edc3b ]

Rewrite kcsan_prandom_u32_max() to not depend on code that might be
instrumented, removing any dependency on lib/random32.c. The rewrite
implements a simple linear congruential generator, that is sufficient
for our purposes (for udelay() and skip_watch counter randomness).

The initial motivation for this was to allow enabling KCSAN for
kernel/sched (remove KCSAN_SANITIZE := n from kernel/sched/Makefile),
with CONFIG_DEBUG_PREEMPT=y. Without this change, we could observe
recursion:

check_access() [via instrumentation]
  kcsan_setup_watchpoint()
    reset_kcsan_skip()
      kcsan_prandom_u32_max()
        get_cpu_var()
  preempt_disable()
    preempt_count_add() [in kernel/sched/core.c]
      check_access() [via instrumentation]

Note, while this currently does not affect an unmodified kernel, it'd be
good to keep a KCSAN kernel working when KCSAN_SANITIZE := n is removed
from kernel/sched/Makefile to permit testing scheduler code with KCSAN
if desired.

Fixes: cd290ec24633 ("kcsan: Use tracing-safe version of prandom")
Signed-off-by: Marco Elver <elver@google.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

media: allegro: Fix use after free on error

[ Upstream commit ce814ad4bb52bfc7c0472e6da0aa742ab88f4361 ]

The "channel" is added to the "dev->channels" but then if
v4l2_m2m_ctx_init() fails then we free "channel" but it's still on the
list so it could lead to a use after free.  Let's not add it to the
list until after v4l2_m2m_ctx_init() succeeds.

Fixes: cc62c74749a3 ("media: allegro: add missed checks in allegro_open()")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Michael Tretter <m.tretter@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

hwrng: ingenic - Fix a resource leak in an error handling path

[ Upstream commit c4ff41b93d1f10d1b8be258c31a0436c5769fc00 ]

In case of error, we should call 'clk_disable_unprepare()' to undo a
previous 'clk_prepare_enable()' call, as already done in the remove
function.

Fixes: 406346d22278 ("hwrng: ingenic - Add hardware TRNG for Ingenic X1830")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Tested-by: 周琰杰 (Zhou Yanjie) <zhouyanjie@wanyeetech.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Sasha Levin <sashal@kernel.org>

crypto: arm64/aes-ce - really hide slower algos when faster ones are enabled

[ Upstream commit 15deb4333cd6d4e1e3216582e4c531ec40a6b060 ]

Commit 69b6f2e817e5b ("crypto: arm64/aes-neon - limit exposed routines if
faster driver is enabled") intended to hide modes from the plain NEON
driver that are also implemented by the faster bit sliced NEON one if
both are enabled. However, the defined() CPP function does not detect
if the bit sliced NEON driver is enabled as a module. So instead, let's
use IS_ENABLED() here.

Fixes: 69b6f2e817e5b ("crypto: arm64/aes-neon - limit exposed routines if ...")
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Sasha Levin <sashal@kernel.org>

crypto: sun4i-ss - fix kmap usage

[ Upstream commit 9bc3dd24e7dccd50757db743a3635ad5b0497e6e ]

With the recent kmap change, some tests which were conditional on
CONFIG_DEBUG_HIGHMEM now are enabled by default.
This permit to detect a problem in sun4i-ss usage of kmap.

sun4i-ss uses two kmap via sg_miter (one for input, one for output), but
using two kmap at the same time is hard:
"the ordering has to be correct and with sg_miter that's probably hard to get
right." (quoting Tlgx)

So the easiest solution is to never have two sg_miter/kmap open at the same time.
After each use of sg_miter, I store the current index, for being able to
resume sg_miter to the right place.

Fixes: 6298e948215f ("crypto: sunxi-ss - Add Allwinner Security System crypto accelerator")
Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Sasha Levin <sashal@kernel.org>

crypto: sun4i-ss - linearize buffers content must be kept

[ Upstream commit 583513510a7acd2306787865bcd19ebb2f629d42 ]

When running the non-optimized cipher function, SS produce partial random
output.
This is due to linearize buffers being reseted after each loop.

For preserving stack, instead of moving them back to start of function,
I move them in sun4i_ss_ctx.

Fixes: 8d3bcb9900ca ("crypto: sun4i-ss - reduce stack usage")
Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/vc4: hdmi: Take into account the clock doubling flag in atomic_check

[ Upstream commit 320e84dc6111ecc1c957e2b186d4d2bafee6bde2 ]

Commit 63495f6b4aed ("drm/vc4: hdmi: Make sure our clock rate is within
limits") was intended to compute the pixel rate to make sure we remain
within the boundaries of what the hardware can provide.

However, unlike what mode_valid was checking for, we forgot to take
into account the clock doubling flag that can be set for modes. Let's
honor that flag if it's there.

Acked-by: Thomas Zimmermann <tzimmermann@suse.de>
Reported-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Fixes: 63495f6b4aed ("drm/vc4: hdmi: Make sure our clock rate is within limits")
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
Link: https://patchwork.freedesktop.org/patch/msgid/20201215154243.540115-4-maxime@cerno.tech
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/panel: mantix: Tweak init sequence

[ Upstream commit dd396dbc4d7811c1567cc43faa4b9ad68094c44d ]

We've seen some (non permanent) burn in and bad white balance
on some of the panels. Adding this bit from a vendor supplied
sequence fixes it.

Fixes: 72967d5616d3 ("drm/panel: Add panel driver for the Mantix MLAF057WE51-X DSI panel")
Signed-off-by: Guido Günther <agx@sigxcpu.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Sam Ravnborg <sam@ravnborg.org>
Link: https://patchwork.freedesktop.org/patch/msgid/8451831b60d5ecb73a156613d98218a31bd55680.1605688147.git.agx@sigxcpu.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/fb-helper: Add missed unlocks in setcmap_legacy()

[ Upstream commit 0a260e731d6c4c17547ac275a2cde888a9eb4a3d ]

setcmap_legacy() does not call drm_modeset_unlock_all() in some exits,
add the missed unlocks with goto to fix it.

Fixes: 964c60063bff ("drm/fb-helper: separate the fb_setcmap helper into atomic and legacy paths")
Signed-off-by: Chuhong Yuan <hslester96@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20201203144248.418281-1-hslester96@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

gma500: clean up error handling in init

[ Upstream commit 15ccc39b3aab667c6fa131206f01f31bfbccdf6a ]

The main problem with this error handling was that it didn't clean up if
i2c_add_numbered_adapter() failed.  This code is pretty old, and doesn't
match with today's checkpatch.pl standards so I took the opportunity to
tidy it up a bit.  I changed the NULL comparison, and removed the
WARNING message if kzalloc() fails and updated the label names.

Fixes: 1b082ccf5901 ("gma500: Add Oaktrail support")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/X8ikkAqZfnDO2lu6@mwanda
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/gma500: Fix error return code in psb_driver_load()

[ Upstream commit 6926872ae24452d4f2176a3ba2dee659497de2c4 ]

Fix to return a negative error code from the error handling
case instead of 0, as done elsewhere in this function.

Fixes: 5c49fd3aa0ab ("gma500: Add the core DRM files and headers")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Jialin Zhang <zhangjialin11@huawei.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20201130020216.1906141-1-zhangjialin11@huawei.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

fbdev: aty: SPARC64 requires FB_ATY_CT

[ Upstream commit c6c90c70db4d9a0989111d6b994d545659410f7a ]

It looks like SPARC64 requires FB_ATY_CT to build without errors,
so have FB_ATY select FB_ATY_CT if both SPARC64 and PCI are enabled
instead of using "default y if SPARC64 && PCI", which is not strong
enough to prevent build errors.

As it currently is, FB_ATY_CT can be disabled, resulting in build
errors:

ERROR: modpost: "aty_postdividers" [drivers/video/fbdev/aty/atyfb.ko] undefined!
ERROR: modpost: "aty_ld_pll_ct" [drivers/video/fbdev/aty/atyfb.ko] undefined!

Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Fixes: f7018c213502 ("video: move fbdev to drivers/video/fbdev")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: sparclinux@vger.kernel.org
Cc: Tomi Valkeinen <tomi.valkeinen@ti.com>
Cc: dri-devel@lists.freedesktop.org
Cc: linux-fbdev@vger.kernel.org
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: David Airlie <airlied@linux.ie>
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20201127031752.10371-1-rdunlap@infradead.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

tty: implement read_iter

[ Upstream commit dd78b0c483e33225e0e0782b0ed887129b00f956 ]

Now that the ldisc read() function takes kernel pointers, it's fairly
straightforward to make the tty file operations use .read_iter() instead
of .read().

That automatically gives us vread() and friends, and also makes it
possible to do .splice_read() on ttys again.

Fixes: 36e2c7421f02 ("fs: don't allow splice read/write without explicit ops")
Reported-by: Oliver Giles <ohw.giles@gmail.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tty: convert tty_ldisc_ops 'read()' function to take a kernel pointer

[ Upstream commit 3b830a9c34d5897be07176ce4e6f2d75e2c8cfd7 ]

The tty line discipline .read() function was passed the final user
pointer destination as an argument, which doesn't match the 'write()'
function, and makes it very inconvenient to do a splice method for
ttys.

This is a conversion to use a kernel buffer instead.

NOTE! It does this by passing the tty line discipline ->read() function
an additional "cookie" to fill in, and an offset into the cookie data.

The line discipline can fill in the cookie data with its own private
information, and then the reader will repeat the read until either the
cookie is cleared or it runs out of data.

The only real user of this is N_HDLC, which can use this to handle big
packets, even if the kernel buffer is smaller than the whole packet.

Cc: Christoph Hellwig <hch@lst.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: enetc: fix destroyed phylink dereference during unbind

[ Upstream commit 3af409ca278d4a8d50e91f9f7c4c33b175645cf3 ]

The following call path suggests that calling unregister_netdev on an
interface that is up will first bring it down.

enetc_pf_remove
-> unregister_netdev
   -> unregister_netdevice_queue
      -> unregister_netdevice_many
         -> dev_close_many
            -> __dev_close_many
               -> enetc_close
                  -> enetc_stop
                     -> phylink_stop

However, enetc first destroys the phylink instance, then calls
unregister_netdev. This is already dissimilar to the setup (and error
path teardown path) from enetc_pf_probe, but more than that, it is buggy
because it is invalid to call phylink_stop after phylink_destroy.

So let's first unregister the netdev (and let the .ndo_stop events
consume themselves), then destroy the phylink instance, then free the
netdev.

Fixes: 71b77a7a27a3 ("enetc: Migrate to PHYLINK and PCS_LYNX")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: mvneta: Remove per-cpu queue mapping for Armada 3700

[ Upstream commit cf9bf871280d9e0a8869d98c2602d29caf69dfa3 ]

According to Errata #23 "The per-CPU GbE interrupt is limited to Core
0", we can't use the per-cpu interrupt mechanism on the Armada 3700
familly.

This is correctly checked for RSS configuration, but the initial queue
mapping is still done by having the queues spread across all the CPUs in
the system, both in the init path and in the cpu_hotplug path.

Fixes: 2636ac3cc2b4 ("net: mvneta: Add network support for Armada 3700 SoC")
Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: amd-xgbe: Fix network fluctuations when using 1G BELFUSE SFP

[ Upstream commit 9eab3fdb419916f66a72d1572f68d82cd9b3f963 ]

Frequent link up/down events can happen when a Bel Fuse SFP part is
connected to the amd-xgbe device. Try to avoid the frequent link
issues by resetting the PHY as documented in Bel Fuse SFP datasheets.

Fixes: e722ec82374b ("amd-xgbe: Update the BelFuse quirk to support SGMII")
Co-developed-by: Sudheesh Mavila <sudheesh.mavila@amd.com>
Signed-off-by: Sudheesh Mavila <sudheesh.mavila@amd.com>
Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Acked-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: amd-xgbe: Reset link when the link never comes back

[ Upstream commit 84fe68eb67f9499309cffd97c1ba269de125ff14 ]

Normally, auto negotiation and reconnect should be automatically done by
the hardware. But there seems to be an issue where auto negotiation has
to be restarted manually. This happens because of link training and so
even though still connected to the partner the link never "comes back".
This needs an auto-negotiation restart.

Also, a change in xgbe-mdio is needed to get ethtool to recognize the
link down and get the link change message. This change is only
required in a backplane connection mode.

Fixes: abf0a1c2b26a ("amd-xgbe: Add support for SFP+ modules")
Co-developed-by: Sudheesh Mavila <sudheesh.mavila@amd.com>
Signed-off-by: Sudheesh Mavila <sudheesh.mavila@amd.com>
Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Acked-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: amd-xgbe: Fix NETDEV WATCHDOG transmit queue timeout warning

[ Upstream commit 186edbb510bd60e748f93975989ccba25ee99c50 ]

The current driver calls netif_carrier_off() late in the link tear down
which can result in a netdev watchdog timeout.

Calling netif_carrier_off() immediately after netif_tx_stop_all_queues()
avoids the warning.

 ------------[ cut here ]------------
 NETDEV WATCHDOG: enp3s0f2 (amd-xgbe): transmit queue 0 timed out
 WARNING: CPU: 3 PID: 0 at net/sched/sch_generic.c:461 dev_watchdog+0x20d/0x220
 Modules linked in: amd_xgbe(E)  amd-xgbe 0000:03:00.2 enp3s0f2: Link is Down
 CPU: 3 PID: 0 Comm: swapper/3 Tainted: G            E
 Hardware name: AMD Bilby-RV2/Bilby-RV2, BIOS RBB1202A 10/18/2019
 RIP: 0010:dev_watchdog+0x20d/0x220
 Code: 00 49 63 4e e0 eb 92 4c 89 e7 c6 05 c6 e2 c1 00 01 e8 e7 ce fc ff 89 d9 48
 RSP: 0018:ffff90cfc28c3e88 EFLAGS: 00010286
 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000006
 RDX: 0000000000000007 RSI: 0000000000000086 RDI: ffff90cfc28d63c0
 RBP: ffff90cfb977845c R08: 0000000000000050 R09: 0000000000196018
 R10: ffff90cfc28c3ef8 R11: 0000000000000000 R12: ffff90cfb9778000
 R13: 0000000000000003 R14: ffff90cfb9778480 R15: 0000000000000010
 FS:  0000000000000000(0000) GS:ffff90cfc28c0000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00007f240ff2d9d0 CR3: 00000001e3e0a000 CR4: 00000000003406e0
 Call Trace:
  <IRQ>
  ? pfifo_fast_reset+0x100/0x100
  call_timer_fn+0x2b/0x130
  run_timer_softirq+0x3e8/0x440
  ? enqueue_hrtimer+0x39/0x90

Fixes: e722ec82374b ("amd-xgbe: Update the BelFuse quirk to support SGMII")
Co-developed-by: Sudheesh Mavila <sudheesh.mavila@amd.com>
Signed-off-by: Sudheesh Mavila <sudheesh.mavila@amd.com>
Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Acked-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: amd-xgbe: Reset the PHY rx data path when mailbox command timeout

[ Upstream commit 30b7edc82ec82578f4f5e6706766f0a9535617d3 ]

Sometimes mailbox commands timeout when the RX data path becomes
unresponsive. This prevents the submission of new mailbox commands to DXIO.
This patch identifies the timeout and resets the RX data path so that the
next message can be submitted properly.

Fixes: 549b32af9f7c ("amd-xgbe: Simplify mailbox interface rate change code")
Co-developed-by: Sudheesh Mavila <sudheesh.mavila@amd.com>
Signed-off-by: Sudheesh Mavila <sudheesh.mavila@amd.com>
Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Acked-by: Tom Lendacky <thomas.lendacky@amd.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: phy: mscc: adding LCPLL reset to VSC8514

[ Upstream commit 3cc2c646be0b22037f31c958e96c0544a073d108 ]

At Power-On Reset, transients may cause the LCPLL to lock onto a
clock that is momentarily unstable. This is normally seen in QSGMII
setups where the higher speed 6G SerDes is being used.
This patch adds an initial LCPLL Reset to the PHY (first instance)
to avoid this issue.

Fixes: e4f9ba642f0b ("net: phy: mscc: add support for VSC8514 PHY.")
Signed-off-by: Steen Hegelund <steen.hegelund@microchip.com>
Signed-off-by: Bjarni Jonasson <bjarni.jonasson@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: felix: don't deinitialize unused ports

[ Upstream commit 42b5adbbac03bdb396192316c015fa3e64ffd5a1 ]

ocelot_init_port is called only if dsa_is_unused_port == false, however
ocelot_deinit_port is called unconditionally. This causes a warning in
the skb_queue_purge inside ocelot_deinit_port saying that the spin lock
protecting ocelot_port->tx_skbs was not initialized.

Fixes: e5fb512d81d0 ("net: mscc: ocelot: deinitialize only initialized ports")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: felix: perform teardown in reverse order of setup

[ Upstream commit d19741b0f54487cf3a11307900f8633935cd2849 ]

In general it is desirable that cleanup is the reverse process of setup.
In this case I am not seeing any particular issue, but with the
introduction of devlink-sb for felix, a non-obvious decision had to be
made as to where to put its cleanup method. When there's a convention in
place, that decision becomes obvious.

Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ibmvnic: skip send_request_unmap for timeout reset

[ Upstream commit 7d3a7b9ea59ddb223aec59b45fa1713c633aaed4 ]

Timeout reset will trigger the VIOS to unmap it automatically,
similarly as FAILVOER and MOBILITY events. If we unmap it
in the linux side, we will see errors like
"30000003: Error 4 in REQUEST_UNMAP_RSP".
So, don't call send_request_unmap for timeout reset.

Fixes: ed651a10875f ("ibmvnic: Updated reset handling")
Signed-off-by: Lijun Pan <ljp@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ibmvnic: add memory barrier to protect long term buffer

[ Upstream commit 42557dab78edc8235aba5b441f2eb35f725a0ede ]

dma_rmb() barrier is added to load the long term buffer before copying
it to socket buffer; and dma_wmb() barrier is added to update the
long term buffer before it being accessed by VIOS (virtual i/o server).

Fixes: 032c5e82847a ("Driver for IBM System i/p VNIC protocol")
Signed-off-by: Lijun Pan <ljp@linux.ibm.com>
Acked-by: Thomas Falcon <tlfalcon@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf: Clear subreg_def for global function return values

[ Upstream commit 45159b27637b0fef6d5ddb86fc7c46b13c77960f ]

test_global_func4 fails on s390 as reported by Yauheni in [1].

The immediate problem is that the zext code includes the instruction,
whose result needs to be zero-extended, into the zero-extension
patchlet, and if this instruction happens to be a branch, then its
delta is not adjusted. As a result, the verifier rejects the program
later.

However, according to [2], as far as the verifier's algorithm is
concerned and as specified by the insn_no_def() function, branching
insns do not define anything. This includes call insns, even though
one might argue that they define %r0.

This means that the real problem is that zero extension kicks in at
all. This happens because clear_caller_saved_regs() sets BPF_REG_0's
subreg_def after global function calls. This can be fixed in many
ways; this patch mimics what helper function call handling already
does.

  [1] https://lore.kernel.org/bpf/20200903140542.156624-1-yauheni.kaliuta@redhat.com/
  [2] https://lore.kernel.org/bpf/CAADnVQ+2RPKcftZw8d+B1UwB35cpBhpF5u3OocNh90D9pETPwg@mail.gmail.com/

Fixes: 51c39bb1d5d1 ("bpf: Introduce function-by-function verification")
Reported-by: Yauheni Kaliuta <yauheni.kaliuta@redhat.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20210212040408.90109-1-iii@linux.ibm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

b43: N-PHY: Fix the update of coef for the PHY revision >= 3case

[ Upstream commit 4773acf3d4b50768bf08e9e97a204819e9ea0895 ]

The documentation for the PHY update [1] states:

Loop 4 times with index i

    If PHY Revision >= 3
        Copy table[i] to coef[i]
    Otherwise
        Set coef[i] to 0

the copy of the table to coef is currently implemented the wrong way
around, table is being updated from uninitialized values in coeff.
Fix this by swapping the assignment around.

[1] https://bcm-v4.sipsolutions.net/802.11/PHY/N/RestoreCal/

Fixes: 2f258b74d13c ("b43: N-PHY: implement restoring general configuration")
Addresses-Coverity: ("Uninitialized scalar variable")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

cxgb4/chtls/cxgbit: Keeping the max ofld immediate data size same in cxgb4 and ulds

[ Upstream commit 2355a6773a2cb0d2dce13432dde78497f1d6617b ]

The Max imm data size in cxgb4 is not similar to the max imm data size
in the chtls. This caused an mismatch in output of is_ofld_imm() of
cxgb4 and chtls. So fixed this by keeping the max wreq size of imm data
same in both chtls and cxgb4 as MAX_IMM_OFLD_TX_DATA_WR_LEN.

As cxgb4's max imm. data value for ofld packets is changed to
MAX_IMM_OFLD_TX_DATA_WR_LEN. Using the same in cxgbit also.

Fixes: 36bedb3f2e5b8 ("crypto: chtls - Inline TLS record Tx")
Signed-off-by: Ayush Sawal <ayush.sawal@chelsio.com>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: axienet: Handle deferred probe on clock properly

[ Upstream commit 57baf8cc70ea4cf5503c9d42f31f6a86d7f5ff1a ]

This driver is set up to use a clock mapping in the device tree if it is
present, but still work without one for backward compatibility. However,
if getting the clock returns -EPROBE_DEFER, then we need to abort and
return that error from our driver initialization so that the probe can
be retried later after the clock is set up.

Move clock initialization to earlier in the process so we do not waste as
much effort if the clock is not yet available. Switch to use
devm_clk_get_optional and abort initialization on any error reported.
Also enable the clock regardless of whether the controller is using an MDIO
bus, as the clock is required in any case.

Fixes: 09a0354cadec267be7f ("net: axienet: Use clock framework to get device clock rate")
Signed-off-by: Robert Hancock <robert.hancock@calian.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tcp: fix SO_RCVLOWAT related hangs under mem pressure

[ Upstream commit f969dc5a885736842c3511ecdea240fbb02d25d9 ]

While commit 24adbc1676af ("tcp: fix SO_RCVLOWAT hangs with fat skbs")
fixed an issue vs too small sk_rcvbuf for given sk_rcvlowat constraint,
it missed to address issue caused by memory pressure.

1) If we are under memory pressure and socket receive queue is empty.
First incoming packet is allowed to be queued, after commit
76dfa6082032 ("tcp: allow one skb to be received per socket under memory pressure")

But we do not send EPOLLIN yet, in case tcp_data_ready() sees sk_rcvlowat
is bigger than skb length.

2) Then, when next packet comes, it is dropped, and we directly
call sk->sk_data_ready().

3) If application is using poll(), tcp_poll() will then use
tcp_stream_is_readable() and decide the socket receive queue is
not yet filled, so nothing will happen.

Even when sender retransmits packets, phases 2) & 3) repeat
and flow is effectively frozen, until memory pressure is off.

Fix is to consider tcp_under_memory_pressure() to take care
of global memory pressure or memcg pressure.

Fixes: 24adbc1676af ("tcp: fix SO_RCVLOWAT hangs with fat skbs")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Arjun Roy <arjunroy@google.com>
Suggested-by: Wei Wang <weiwan@google.com>
Reviewed-by: Wei Wang <weiwan@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: mptcp: fix ACKRX debug message

[ Upstream commit f384221a381751508f390b36d0e51bd5a7beb627 ]

Info from received MPCapable SYN were printed instead of the ones from
received MPCapable 3rd ACK.

Fixes: fed61c4b584c ("selftests: mptcp: make 2nd net namespace use tcp syn cookies unconditionally")
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf: Fix bpf_fib_lookup helper MTU check for SKB ctx

[ Upstream commit 2c0a10af688c02adcf127aad29e923e0056c6b69 ]

BPF end-user on Cilium slack-channel (Carlo Carraro) wants to use
bpf_fib_lookup for doing MTU-check, but *prior* to extending packet size,
by adjusting fib_params 'tot_len' with the packet length plus the expected
encap size. (Just like the bpf_check_mtu helper supports). He discovered
that for SKB ctx the param->tot_len was not used, instead skb->len was used
(via MTU check in is_skb_forwardable() that checks against netdev MTU).

Fix this by using fib_params 'tot_len' for MTU check. If not provided (e.g.
zero) then keep existing TC behaviour intact. Notice that 'tot_len' for MTU
check is done like XDP code-path, which checks against FIB-dst MTU.

V16:
- Revert V13 optimization, 2nd lookup is against egress/resulting netdev

V13:
- Only do ifindex lookup one time, calling dev_get_by_index_rcu().

V10:
- Use same method as XDP for 'tot_len' MTU check

Fixes: 4c79579b44b1 ("bpf: Change bpf_fib_lookup to return lookup status")
Reported-by: Carlo Carraro <colrack@gmail.com>
Signed-off-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/161287789444.790810.15247494756551413508.stgit@firesoul
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf, devmap: Use GFP_KERNEL for xdp bulk queue allocation

[ Upstream commit 7d4553b69fb335496c597c31590e982485ebe071 ]

The devmap bulk queue is allocated with GFP_ATOMIC and the allocation
may fail if there is no available space in existing percpu pool.

Since commit 75ccae62cb8d42 ("xdp: Move devmap bulk queue into struct net_device")
moved the bulk queue allocation to NETDEV_REGISTER callback, whose context
is allowed to sleep, use GFP_KERNEL instead of GFP_ATOMIC to let percpu
allocator extend the pool when needed and avoid possible failure of netdev
registration.

As the required alignment is natural, we can simply use alloc_percpu().

Fixes: 75ccae62cb8d42 ("xdp: Move devmap bulk queue into struct net_device")
Signed-off-by: Jun'ichi Nomura <junichi.nomura@nec.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Cc: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://lore.kernel.org/bpf/20210209082451.GA44021@jeru.linux.bs1.fc.nec.co.jp
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf: Fix an unitialized value in bpf_iter

[ Upstream commit 17d8beda277a36203585943e70c7909b60775fd5 ]

Commit 15d83c4d7cef ("bpf: Allow loading of a bpf_iter program")
cached btf_id in struct bpf_iter_target_info so later on
if it can be checked cheaply compared to checking registered names.

syzbot found a bug that uninitialized value may occur to
bpf_iter_target_info->btf_id. This is because we allocated
bpf_iter_target_info structure with kmalloc and never initialized
field btf_id afterwards. This uninitialized btf_id is typically
compared to a u32 bpf program func proto btf_id, and the chance
of being equal is extremely slim.

This patch fixed the issue by using kzalloc which will also
prevent future likely instances due to adding new fields.

Fixes: 15d83c4d7cef ("bpf: Allow loading of a bpf_iter program")
Reported-by: syzbot+580f4f2a272e452d55cb@syzkaller.appspotmail.com
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20210212005926.2875002-1-yhs@fb.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

libbpf: Ignore non function pointer member in struct_ops

[ Upstream commit d2836dddc95d5dd82c7cb23726c97d8c9147f050 ]

When libbpf initializes the kernel's struct_ops in
"bpf_map__init_kern_struct_ops()", it enforces all
pointer types must be a function pointer and rejects
others.  It turns out to be too strict.  For example,
when directly using "struct tcp_congestion_ops" from vmlinux.h,
it has a "struct module *owner" member and it is set to NULL
in a bpf_tcp_cc.o.

Instead, it only needs to ensure the member is a function
pointer if it has been set (relocated) to a bpf-prog.
This patch moves the "btf_is_func_proto(kern_mtype)" check
after the existing "if (!prog) { continue; }".  The original debug
message in "if (!prog) { continue; }" is also removed since it is
no longer valid.  Beside, there is a later debug message to tell
which function pointer is set.

The "btf_is_func_proto(mtype)" has already been guaranteed
in "bpf_object__collect_st_ops_relos()" which has been run
before "bpf_map__init_kern_struct_ops()".  Thus, this check
is removed.

v2:
- Remove outdated debug message (Andrii)
  Remove because there is a later debug message to tell
  which function pointer is set.
- Following mtype->type is no longer needed. Remove:
  "skip_mods_and_typedefs(btf, mtype->type, &mtype_id)"
- Do "if (!prog)" test before skip_mods_and_typedefs.

Fixes: 590a00888250 ("bpf: libbpf: Add STRUCT_OPS support")
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: Andrii Nakryiko <andrii@kernel.org>
Link: https://lore.kernel.org/bpf/20210212021030.266932-1-kafai@fb.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

mac80211: fix potential overflow when multiplying to u32 integers

[ Upstream commit 6194f7e6473be78acdc5d03edd116944bdbb2c4e ]

The multiplication of the u32 variables tx_time and estimated_retx is
performed using a 32 bit multiplication and the result is stored in
a u64 result. This has a potential u32 overflow issue, so avoid this
by casting tx_time to a u64 to force a 64 bit multiply.

Addresses-Coverity: ("Unintentional integer overflow")
Fixes: 050ac52cbe1f ("mac80211: code for on-demand Hybrid Wireless Mesh Protocol")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/r/20210205175352.208841-1-colin.king@canonical.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: Check tunnel offload is required before setting SWP

[ Upstream commit e1c3940c6003d820c787473c65711b49c2d1bc42 ]

Check that tunnel offload is required before setting Software Parser
offsets to get Geneve HW offload. In case of Geneve packet we check HW
offload support of SWP in mlx5e_tunnel_features_check() and set features
accordingly, this should be reflected in skb offload requested by the
kernel and we should add the Software Parser offsets only if requested.
Otherwise, in case HW doesn't support SWP for Geneve, data path will
mistakenly try to offload Geneve SKBs with skb->encapsulation set,
regardless of whether offload was requested or not on this specific SKB.

Fixes: e3cfc7e6b7bd ("net/mlx5e: TX, Add geneve tunnel stateless offload support")
Signed-off-by: Moshe Shemesh <moshe@nvidia.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: CT: manage the lifetime of the ct entry object

[ Upstream commit a2173131526dc845eb1968a15bc192b3fc2ff000 ]

The ct entry object is accessed by the ct add, del, stats and restore
methods. In addition, it is referenced from several hash tables.

The lifetime of the ct entry object was not managed which triggered race
conditions as in the following kasan dump:
[ 3374.973945] ==================================================================
[ 3374.988552] BUG: KASAN: use-after-free in memcmp+0x4c/0x98
[ 3374.999590] Read of size 1 at addr ffff00036129ea55 by task ksoftirqd/1/15
[ 3375.016415] CPU: 1 PID: 15 Comm: ksoftirqd/1 Tainted: G           O      5.4.31+ #1
[ 3375.055301] Call trace:
[ 3375.060214]  dump_backtrace+0x0/0x238
[ 3375.067580]  show_stack+0x24/0x30
[ 3375.074244]  dump_stack+0xe0/0x118
[ 3375.081085]  print_address_description.isra.9+0x74/0x3d0
[ 3375.091771]  __kasan_report+0x198/0x1e8
[ 3375.099486]  kasan_report+0xc/0x18
[ 3375.106324]  __asan_load1+0x60/0x68
[ 3375.113338]  memcmp+0x4c/0x98
[ 3375.119409]  mlx5e_tc_ct_restore_flow+0x3a4/0x6f8 [mlx5_core]
[ 3375.131073]  mlx5e_rep_tc_update_skb+0x1d4/0x2f0 [mlx5_core]
[ 3375.142553]  mlx5e_handle_rx_cqe_rep+0x198/0x308 [mlx5_core]
[ 3375.154034]  mlx5e_poll_rx_cq+0x2a0/0x1060 [mlx5_core]
[ 3375.164459]  mlx5e_napi_poll+0x1d4/0xa78 [mlx5_core]
[ 3375.174453]  net_rx_action+0x28c/0x7a8
[ 3375.182004]  __do_softirq+0x1b4/0x5d0

Manage the lifetime of the ct entry object by using synchornization
mechanisms for concurrent access.

Fixes: ac991b48d43c ("net/mlx5e: CT: Offload established flows")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Oz Shlomo <ozsh@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5: Disable devlink reload for lag devices

[ Upstream commit edac23c2b3d3ac64cfcd351087295893671adbf5 ]

Devlink reload can't be allowed on lag devices since reloading one lag
device will cause traffic on the bond to get stucked.
Users who wish to reload a lag device, need to remove the device from
the bond, and only then reload it.

Fixes: 4383cfcc65e7 ("net/mlx5: Add devlink reload")
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5: Disallow RoCE on lag device

[ Upstream commit 7ab91f2b03367f9d25dd807ebdfb0d67295e0e41 ]

In lag mode, setting roce enabled/disable of lag device have no effect.
e.g.: bond device (roce/vf_lag) roce status remain unchanged.
Therefore disable it and add an error message.

Fixes: cc9defcbb8fa ("net/mlx5: Handle "enable_roce" devlink param")
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5: Disallow RoCE on multi port slave device

[ Upstream commit c70f8597fcc1399ef6d5b5ce648a31d887d5dba2 ]

In dual port mode, setting roce enabled/disable for the slave device
have no effect. e.g.: the slave device roce status remain unchanged.
Therefore disable it and add an error message.
Enable or disable roce of the master device affect both master and slave
devices.

Fixes: cc9defcbb8fa ("net/mlx5: Handle "enable_roce" devlink param")
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5: Disable devlink reload for multi port slave device

[ Upstream commit d89ddaae1766f8fe571ea6eb63ec098ff556f1dd ]

Devlink reload can't be allowed on a multi port slave device, because
reload of slave device doesn't take effect.

The right flow is to disable devlink reload for multi port slave
device. Hence, disabling it in mlx5_core probing.

Fixes: 4383cfcc65e7 ("net/mlx5: Add devlink reload")
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/mlx5e: kTLS, Use refcounts to free kTLS RX priv context

[ Upstream commit b850bbff965129c34f50962638c0a66c82563536 ]

wait_for_resync is unreliable - if it timeouts, priv_rx will be freed
anyway. However, mlx5e_ktls_handle_get_psv_completion will be called
sooner or later, leading to use-after-free. For example, it can happen
if a CQ error happened, and ICOSQ stopped, but later on the queues are
destroyed, and ICOSQ is flushed with mlx5e_free_icosq_descs.

This patch converts the lifecycle of priv_rx to fully refcount-based, so
that the struct won't be freed before the refcount goes to zero.

Fixes: 0419d8c9d8f8 ("net/mlx5e: kTLS, Add kTLS RX resync support")
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Reviewed-by: Tariq Toukan <tariqt@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>