When send a broadcast packet to a tap device, which was added to a bridge,
br_nf_local_in() is called to confirm the conntrack. If another conntrack
with the same hash value is added to the hash table, which can be
triggered by a normal packet to a non-bridge device, the below warning
may happen.
To solve the hash conflict, nf_ct_resolve_clash() try to merge the
conntracks, and update skb->_nfct. However, br_nf_local_in() still use the
old ct from local variable 'nfct' after confirm(), which leads to this
warning.
If confirm() does not insert the conntrack entry and return NF_DROP, the
warning may also occur. There is no need to reserve the WARN_ON_ONCE, just
remove it.
Link: https://lore.kernel.org/netdev/20250820043329.2902014-1-wangliang74@huawei.com/ Fixes: 62e7151ae3eb ("netfilter: bridge: confirm multicast packets before passing them up the stack") Suggested-by: Florian Westphal <fw@strlen.de> Signed-off-by: Wang Liang <wangliang74@huawei.com> Signed-off-by: Florian Westphal <fw@strlen.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
The brcmf_btcoex_detach() only shuts down the btcoex timer, if the
flag timer_on is false. However, the brcmf_btcoex_timerfunc(), which
runs as timer handler, sets timer_on to false. This creates critical
race conditions:
1.If brcmf_btcoex_detach() is called while brcmf_btcoex_timerfunc()
is executing, it may observe timer_on as false and skip the call to
timer_shutdown_sync().
2.The brcmf_btcoex_timerfunc() may then reschedule the brcmf_btcoex_info
worker after the cancel_work_sync() has been executed, resulting in
use-after-free bugs.
The use-after-free bugs occur in two distinct scenarios, depending on
the timing of when the brcmf_btcoex_info struct is freed relative to
the execution of its worker thread.
Scenario 1: Freed before the worker is scheduled
The brcmf_btcoex_info is deallocated before the worker is scheduled.
A race condition can occur when schedule_work(&bt_local->work) is
called after the target memory has been freed. The sequence of events
is detailed below:
The brcmf_btcoex_info is freed after the worker has been scheduled
but before or during its execution. In this case, statements within
the brcmf_btcoex_handler() — such as the container_of macro and
subsequent dereferences of the brcmf_btcoex_info object will cause
a use-after-free access. The following timeline illustrates this
scenario:
CPU0 | CPU1
brcmf_btcoex_detach | brcmf_btcoex_timerfunc
| bt_local->timer_on = false;
if (cfg->btcoex->timer_on) |
... |
cancel_work_sync(); |
... | schedule_work(); // Reschedule
|
kfree(cfg->btcoex); // FREE | brcmf_btcoex_handler() // Worker
/* | btci = container_of(....); // USE
The kfree() above could | ...
also occur at any point | btci-> // USE
during the worker's execution|
*/ |
To resolve the race conditions, drop the conditional check and call
timer_shutdown_sync() directly. It can deactivate the timer reliably,
regardless of its current state. Once stopped, the timer_on state is
then set to false.
Fixes: 61730d4dfffc ("brcmfmac: support critical protocol API for DHCP") Acked-by: Arend van Spriel <arend.vanspriel@broadcom.com> Signed-off-by: Duoming Zhou <duoming@zju.edu.cn> Link: https://patch.msgid.link/20250822050839.4413-1-duoming@zju.edu.cn Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Following bss_free() quirk introduced in commit 776b3580178f
("cfg80211: track hidden SSID networks properly"), adjust
cfg80211_update_known_bss() to free the last beacon frame
elements only if they're not shared via the corresponding
'hidden_beacon_bss' pointer.
During SD suspend/resume without a full card rescan (when using
non-removable SD cards for rootfs), the SD card initialization may fail
after resume. This occurs because, after a host controller reset, the
card detect logic may take time to stabilize due to debounce logic.
Without waiting for stabilization, the host may attempt powering up the
card prematurely, leading to command timeouts during resume flow.
Add sdhci_arasan_set_power_and_bus_voltage() to wait for the card detect
stable bit before power up the card. Since the stabilization time
is not fixed, a maximum timeout of one second is used to ensure
sufficient wait time for the card detect signal to stabilize.
Add hw_reset callback to support emmc hardware reset, this callback get
called from the mmc core only when "cap-mmc-hw-reset" property is
defined in the DT.
Signed-off-by: Paul Alvin <alvin.paulp@amd.com> Acked-by: Adrian Hunter <adrian.hunter@intel.com> Link: https://lore.kernel.org/r/20241007095445.19340-1-alvin.paulp@amd.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Stable-dep-of: e251709aaddb ("mmc: sdhci-of-arasan: Ensure CD logic stabilization before power-up") Signed-off-by: Sasha Levin <sashal@kernel.org>
fe00e50b2db8 ("ARM: 8858/1: vdso: use $(LD) instead of $(CC) to link VDSO") 691efbedc60d ("arm64: vdso: use $(LD) instead of $(CC) to link VDSO") 2ff906994b6c ("MIPS: VDSO: Use $(LD) instead of $(CC) to link VDSO") 2b2a25845d53 ("s390/vdso: Use $(LD) instead of $(CC) to link vDSO")
Glibc added support for .gnu.hash in 2006 and .hash has been obsoleted
far before the first LoongArch CPU was taped. Using --hash-style=sysv
might imply unaddressed issues and confuse readers.
Some architectures use an explicit --hash-style=both for vDSO here, but
DT_GNU_HASH has already been supported by Glibc and Musl and become the
de-facto standard of the distros when the first LoongArch CPU was taped.
So DT_HASH seems just wasting storage space for LoongArch.
Just drop the option and rely on the linker default, which is likely
"gnu" (Arch, Debian, Gentoo, LFS) on all LoongArch distros (confirmed on
Arch, Debian, Gentoo, and LFS; AOSC now defaults to "both" but it seems
just an oversight).
Following the logic of commit 48f6430505c0b049 ("arm64/vdso: Remove
--hash-style=sysv").
Testing by the syzbot fuzzer showed that the HID core gets a
shift-out-of-bounds exception when it tries to convert a 32-bit
quantity to a 0-bit quantity. Ideally this should never occur, but
there are buggy devices and some might have a report field with size
set to zero; we shouldn't reject the report or the device just because
of that.
Instead, harden the s32ton() routine so that it returns a reasonable
result instead of crashing when it is called with the number of bits
set to 0 -- the same as what snto32() does.
The only user of hid_snto32() is Logitech HID++ driver, which always
calls hid_snto32() with valid size (constant, either 12 or 8) and
therefore can simply use sign_extend32().
Make the switch and remove hid_snto32(). Move snto32() and s32ton() to
avoid introducing forward declaration.
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com> Link: https://patch.msgid.link/20241003144656.3786064-2-dmitry.torokhov@gmail.com
[bentiss: fix checkpatch warning] Signed-off-by: Benjamin Tissoires <bentiss@kernel.org>
Stable-dep-of: a6b87bfc2ab5 ("HID: core: Harden s32ton() against conversion to 0 bits") Signed-off-by: Sasha Levin <sashal@kernel.org>
snto32() does exactly what sign_extend32() does, but handles
potentially malformed data coming from the device. Keep the checks,
but then call sign_extend32() to perform the actual conversion.
Add missing microSD slot vqmmc-supply property, otherwise the kernel
might shut down LDO5 regulator and that would power off the microSD
card slot, possibly while it is in use. Add the property to make sure
the kernel is aware of the LDO5 regulator which supplies the microSD
slot and keeps the LDO5 enabled.
Fixes: 562d222f23f0 ("arm64: dts: imx8mp: Add support for Data Modul i.MX8M Plus eDM SBC") Signed-off-by: Marek Vasut <marek.vasut@mailbox.org> Signed-off-by: Shawn Guo <shawnguo@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Add missing microSD slot vqmmc-supply property, otherwise the kernel
might shut down LDO5 regulator and that would power off the microSD
card slot, possibly while it is in use. Add the property to make sure
the kernel is aware of the LDO5 regulator which supplies the microSD
slot and keeps the LDO5 enabled.
Fixes: 8d6712695bc8 ("arm64: dts: imx8mp: Add support for DH electronics i.MX8M Plus DHCOM and PDK2") Signed-off-by: Marek Vasut <marek.vasut@mailbox.org> Signed-off-by: Shawn Guo <shawnguo@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Fix SD card removal caused by automatic LDO5 power off after boot:
LDO5: disabling
mmc1: card 59b4 removed
EXT4-fs (mmcblk1p2): shut down requested (2)
Aborting journal on device mmcblk1p2-8.
JBD2: I/O error when updating journal superblock for mmcblk1p2-8.
To prevent this, add vqmmc regulator for USDHC, using a GPIO-controlled
regulator that is supplied by LDO5. Since this is implemented on SoM but
used on baseboards with SD-card interface, implement the functionality
on SoM part and optionally enable it on baseboards if needed.
Fixes: 418d1d840e42 ("arm64: dts: freescale: add initial device tree for TQMa8MPQL with i.MX8MP") Signed-off-by: Markus Niebel <Markus.Niebel@ew.tq-group.com> Signed-off-by: Alexander Stein <alexander.stein@ew.tq-group.com> Signed-off-by: Shawn Guo <shawnguo@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Don't call ZONE FINISH for conventional zones as this will result in I/O
errors. Instead check if the zone that needs finishing is a conventional
zone and if yes skip it.
Also factor out the actual handling of finishing a single zone into a
helper function, as do_zone_finish() is growing ever bigger and the
indentations levels are getting higher.
Reviewed-by: Naohiro Aota <naohiro.aota@wdc.com> Reviewed-by: Anand Jain <anand.jain@oracle.com> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Link: https://github.com/bluez/bluez/issues/1442 Signed-off-by: Yang Li <yang.li@amlogic.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The set subcommand's -t option is documented as being available for boost
configuration, but it was not actually functioning due to a bug
in the option handling.
Now if preemption happens between protected_save_fpu_context() and
protected_save_lbt_context(), FTOP context is lost. Because FTOP is
saved by protected_save_lbt_context() but protected_save_fpu_context()
disables TM before that. So save LBT before FPU in setup_sigcontext()
to avoid this potential risk.
At inode_logged() we do a couple lockless checks for ->logged_trans, and
these are generally safe except the second one in case we get a load or
store tearing due to a concurrent call updating ->logged_trans (either at
btrfs_log_inode() or later at inode_logged()).
In the first case it's safe to compare to the current transaction ID since
once ->logged_trans is set the current transaction, we never set it to a
lower value.
In the second case, where we check if it's greater than zero, we are prone
to load/store tearing races, since we can have a concurrent task updating
to the current transaction ID with store tearing for example, instead of
updating with a single 64 bits write, to update with two 32 bits writes or
four 16 bits writes. In that case the reading side at inode_logged() could
see a positive value that does not match the current transaction and then
return a false negative.
Fix this by doing the second check while holding the inode's spinlock, add
some comments about it too. Also add the data_race() annotation to the
first check to avoid any reports from KCSAN (or similar tools) and comment
about it.
Fixes: 0f8ce49821de ("btrfs: avoid inode logging during rename and link when possible") Reviewed-by: Boris Burkov <boris@bur.io> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
At inode_logged() if we find that the inode was not logged before we
update its ->last_dir_index_offset to (u64)-1 with the goal that the
next directory log operation will see the (u64)-1 and then figure out
it must check what was the index of the last logged dir index key and
update ->last_dir_index_offset to that key's offset (this is done in
update_last_dir_index_offset()).
This however has a possibility for a time window where a race can happen
and lead to directory logging skipping dir index keys that should be
logged. The race happens like this:
1) Task A calls inode_logged(), sees ->logged_trans as 0 and then checks
that the inode item was logged before, but before it sets the inode's
->last_dir_index_offset to (u64)-1...
2) Task B is at btrfs_log_inode() which calls inode_logged() early, and
that has set ->last_dir_index_offset to (u64)-1;
3) Task B then enters log_directory_changes() which calls
update_last_dir_index_offset(). There it sees ->last_dir_index_offset
is (u64)-1 and that the inode was logged before (ctx->logged_before is
true), and so it searches for the last logged dir index key in the log
tree and it finds that it has an offset (index) value of N, so it sets
->last_dir_index_offset to N, so that we can skip index keys that are
less than or equal to N (later at process_dir_items_leaf());
4) Task A now sets ->last_dir_index_offset to (u64)-1, undoing the update
that task B just did;
5) Task B will now skip every index key when it enters
process_dir_items_leaf(), since ->last_dir_index_offset is (u64)-1.
Fix this by making inode_logged() not touch ->last_dir_index_offset and
initializing it to 0 when an inode is loaded (at btrfs_alloc_inode()) and
then having update_last_dir_index_offset() treat a value of 0 as meaning
we must check the log tree and update with the index of the last logged
index key. This is fine since the minimum possible value for
->last_dir_index_offset is 1 (BTRFS_DIR_START_INDEX - 1 = 2 - 1 = 1).
This also simplifies the management of ->last_dir_index_offset and now
all accesses to it are done under the inode's log_mutex.
Fixes: 0f8ce49821de ("btrfs: avoid inode logging during rename and link when possible") Reviewed-by: Boris Burkov <boris@bur.io> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
There's a race between checking if an inode was logged before and logging
an inode that can cause us to mark an inode as not logged just after it
was logged by a concurrent task:
1) We have inode X which was not logged before neither in the current
transaction not in past transaction since the inode was loaded into
memory, so it's ->logged_trans value is 0;
2) We are at transaction N;
3) Task A calls inode_logged() against inode X, sees that ->logged_trans
is 0 and there is a log tree and so it proceeds to search in the log
tree for an inode item for inode X. It doesn't see any, but before
it sets ->logged_trans to N - 1...
3) Task B calls btrfs_log_inode() against inode X, logs the inode and
sets ->logged_trans to N;
4) Task A now sets ->logged_trans to N - 1;
5) At this point anyone calling inode_logged() gets 0 (inode not logged)
since ->logged_trans is greater than 0 and less than N, but our inode
was really logged. As a consequence operations like rename, unlink and
link that happen afterwards in the current transaction end up not
updating the log when they should.
Fix this by ensuring inode_logged() only updates ->logged_trans in case
the inode item is not found in the log tree if after tacking the inode's
lock (spinlock struct btrfs_inode::lock) the ->logged_trans value is still
zero, since the inode lock is what protects setting ->logged_trans at
btrfs_log_inode().
Fixes: 0f8ce49821de ("btrfs: avoid inode logging during rename and link when possible") Reviewed-by: Boris Burkov <boris@bur.io> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Lonial reported that an out-of-bounds access in cgroup local storage
can be crafted via tail calls. Given two programs each utilizing a
cgroup local storage with a different value size, and one program
doing a tail call into the other. The verifier will validate each of
the indivial programs just fine. However, in the runtime context
the bpf_cg_run_ctx holds an bpf_prog_array_item which contains the
BPF program as well as any cgroup local storage flavor the program
uses. Helpers such as bpf_get_local_storage() pick this up from the
runtime context:
For the second program which was called from the originally attached
one, this means bpf_get_local_storage() will pick up the former
program's map, not its own. With mismatching sizes, this can result
in an unintended out-of-bounds access.
To fix this issue, we need to extend bpf_map_owner with an array of
storage_cookie[] to match on i) the exact maps from the original
program if the second program was using bpf_get_local_storage(), or
ii) allow the tail call combination if the second program was not
using any of the cgroup local storage maps.
Fixes: 7d9c3427894f ("bpf: Make cgroup storages shared between programs on the same cgroup") Reported-by: Lonial Con <kongln9170@gmail.com> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Link: https://lore.kernel.org/r/20250730234733.530041-4-daniel@iogearbox.net Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Given this is only relevant for BPF tail call maps, it is adding up space
and penalizing other map types. We also need to extend this with further
objects to track / compare to. Therefore, lets move this out into a separate
structure and dynamically allocate it only for BPF tail call maps.
Add a cookie to BPF maps to uniquely identify BPF maps for the timespan
when the node is up. This is different to comparing a pointer or BPF map
id which could get rolled over and reused.
The upstream commit a40c5d727b8111b5db424a1e43e14a1dcce1e77f ("drm/dp:
Change AUX DPCD probe address from DPCD_REV to LANE0_1_STATUS") the
reverted commit backported causes a regression, on one eDP panel at
least resulting in display flickering, described in detail at the Link:
below. The issue fixed by the upstream commit will need a different
solution, revert the backport for now.
As per PCIe r6.0, sec 6.6.1, a Downstream Port that supports Link speeds
greater than 5.0 GT/s, software must wait a minimum of 100 ms after Link
training completes before sending a Configuration Request.
Add this delay in dw_pcie_wait_for_link(), after the link is reported as
up. The delay will only be performed in the success case where the link
came up.
DWC glue drivers that have a link up IRQ (drivers that set
use_linkup_irq = true) do not call dw_pcie_wait_for_link(), instead they
perform this delay in their threaded link up IRQ handler.
Signed-off-by: Niklas Cassel <cassel@kernel.org> Signed-off-by: Manivannan Sadhasivam <mani@kernel.org> Reviewed-by: Damien Le Moal <dlemoal@kernel.org> Reviewed-by: Wilfred Mallawa <wilfred.mallawa@wdc.com> Link: https://patch.msgid.link/20250625102347.1205584-14-cassel@kernel.org Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
the PPSMC_MSG_GetPptLimit msg is not valid for gfx 11.0.3 on vf mode,
so skiped to create power1_cap* hwmon sysfs node.
Signed-off-by: Yang Wang <kevinyang.wang@amd.com> Reviewed-by: Asad Kamal <asad.kamal@amd.com> Acked-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit e82a8d441038d8cb10b63047a9e705c42479d156) Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Using device_find_child() and of_find_device_by_node() to locate
devices could cause an imbalance in the device's reference count.
device_find_child() and of_find_device_by_node() both call
get_device() to increment the reference count of the found device
before returning the pointer. In mtk_drm_get_all_drm_priv(), these
references are never released through put_device(), resulting in
permanent reference count increments. Additionally, the
for_each_child_of_node() iterator fails to release node references in
all code paths. This leaks device node references when loop
termination occurs before reaching MAX_CRTC. These reference count
leaks may prevent device/node resources from being properly released
during driver unbind operations.
As comment of device_find_child() says, 'NOTE: you will need to drop
the reference with put_device() after use'.
Cc: stable@vger.kernel.org Fixes: 1ef7ed48356c ("drm/mediatek: Modify mediatek-drm for mt8195 multi mmsys support") Signed-off-by: Ma Ke <make24@iscas.ac.cn> Reviewed-by: CK Hu <ck.hu@mediatek.com> Link: https://patchwork.kernel.org/project/dri-devel/patch/20250812071932.471730-1-make24@iscas.ac.cn/ Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Function nvkm_gsp_fwsec_v2() sets 'ret' if the kmemdup() call fails, but
it never uses or returns 'ret' after that point. We always need to release
the firmware regardless, so do that and then check for error.
Fixes: 176fdcbddfd2 ("drm/nouveau/gsp/r535: add support for booting GSP-RM") Cc: stable@vger.kernel.org # v6.7+ Signed-off-by: Timur Tabi <ttabi@nvidia.com> Link: https://lore.kernel.org/r/20250813001004.2986092-1-ttabi@nvidia.com Signed-off-by: Danilo Krummrich <dakr@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On some chipsets, which block-linear modifiers are
supported is format-specific. However, linear
modifiers are always be supported. The prior
modifier filtering logic was not accounting for
the linear case.
Cc: stable@vger.kernel.org Fixes: c586f30bf74c ("drm/nouveau/kms: Add format mod prop to base/ovly/nvdisp") Signed-off-by: James Jones <jajones@nvidia.com> Link: https://lore.kernel.org/r/20250811220017.1337-3-jajones@nvidia.com Signed-off-by: Danilo Krummrich <dakr@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ENODATA (aka ENOATTR) has a very specific meaning in the xfs xattr code;
namely, that the requested attribute name could not be found.
However, a medium error from disk may also return ENODATA. At best,
this medium error may escape to userspace as "attribute not found"
when in fact it's an IO (disk) error.
At worst, we may oops in xfs_attr_leaf_get() when we do:
because an ENODATA/ENOATTR error from disk leaves us with a null bp,
and the xfs_trans_brelse will then null-deref it.
As discussed on the list, we really need to modify the lower level
IO functions to trap all disk errors and ensure that we don't let
unique errors like this leak up into higher xfs functions - many
like this should be remapped to EIO.
However, this patch directly addresses a reported bug in the xattr
code, and should be safe to backport to stable kernels. A larger-scope
patch to handle more unique errors at lower levels can follow later.
(Note, prior to 07120f1abdff we did not oops, but we did return the
wrong error code to userspace.)
Signed-off-by: Eric Sandeen <sandeen@redhat.com> Fixes: 07120f1abdff ("xfs: Add xfs_has_attr and subroutines") Cc: stable@vger.kernel.org # v5.9+ Reviewed-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Carlos Maiolino <cem@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We were returning -EOPNOTSUPP for various remap_file_range cases
but for some of these the copy_file_range_syscall() requires -EINVAL
to be returned (e.g. where source and target file ranges overlap when
source and target are the same file). This fixes xfstest generic/157
which was expecting EINVAL for that (and also e.g. for when the src
offset is beyond end of file).
Cc: stable@vger.kernel.org Acked-by: Paulo Alcantara (Red Hat) <pc@manguebit.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A possible inconsistent update of refcount was identified in `smb2_compound_op`.
Such inconsistent update could lead to possible resource leaks.
Why it is a possible bug:
1. In the comment section of the function, it clearly states that the
reference to `cfile` should be dropped after calling this function.
2. Every control flow path would check and drop the reference to
`cfile`, except the patched one.
3. Existing callers would not handle refcount update of `cfile` if
-ENOMEM is returned.
To fix the bug, an extra goto label "out" is added, to make sure that the
cleanup logic would always be respected. As the problem is caused by the
allocation failure of `vars`, the cleanup logic between label "finished"
and "out" can be safely ignored. According to the definition of function
`is_replayable_error`, the error code of "-ENOMEM" is not recoverable.
Therefore, the replay logic also gets ignored.
Signed-off-by: Shuhao Fu <sfual@cse.ust.hk> Acked-by: Paulo Alcantara (Red Hat) <pc@manguebit.org> Cc: stable@vger.kernel.org Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When CONFIG_DMA_DIRECT_REMAP is enabled, atomic pool pages are
remapped via dma_common_contiguous_remap() using the supplied
pgprot. Currently, the mapping uses
pgprot_dmacoherent(PAGE_KERNEL), which leaves the memory encrypted
on systems with memory encryption enabled (e.g., ARM CCA Realms).
This can cause the DMA layer to fail or crash when accessing the
memory, as the underlying physical pages are not configured as
expected.
Fix this by requesting a decrypted mapping in the vmap() call:
pgprot_decrypted(pgprot_dmacoherent(PAGE_KERNEL))
This ensures that atomic pool memory is consistently mapped
unencrypted.
If preparing a write bio fails then blk_zone_wplug_bio_work() calls
bio_endio() with zwplug->lock held. If a device mapper driver is stacked
on top of the zoned block device then this results in nested locking of
zwplug->lock. The resulting lockdep complaint is a false positive
because this is nested locking and not recursive locking. Suppress this
false positive by calling blk_zone_wplug_bio_io_error() without holding
zwplug->lock. This is safe because no code in
blk_zone_wplug_bio_io_error() depends on zwplug->lock being held. This
patch suppresses the following lockdep complaint:
WARNING: possible recursive locking detected
--------------------------------------------
kworker/3:0H/46 is trying to acquire lock: ffffff882968b830 (&zwplug->lock){-...}-{2:2}, at: blk_zone_write_plug_bio_endio+0x64/0x1f0
but task is already holding lock: ffffff88315bc230 (&zwplug->lock){-...}-{2:2}, at: blk_zone_wplug_bio_work+0x8c/0x48c
other info that might help us debug this:
Possible unsafe locking scenario:
in ntrig_report_version(), hdev parameter passed from hid_probe().
sending descriptor to /dev/uhid can make hdev->dev.parent->parent to null
if hdev->dev.parent->parent is null, usb_dev has
invalid address(0xffffffffffffff58) that hid_to_usb_dev(hdev) returned
when usb_rcvctrlpipe() use usb_dev,it trigger
page fault error for address(0xffffffffffffff58)
add null check logic to ntrig_report_version()
before calling hid_to_usb_dev()
Adds support for the G PRO 2 LIGHTSPEED Wireless via it's nano receiver
or directly. This nano receiver appears to work identically to the 1_1
receiver for the case I've verified, which is the battery status through
lg-hidpp.
The same appears to be the case wired, sharing much with the Pro X
Superlight 2; differences seemed to lie in userland configuration rather
than in interfaces used by hid_logitech_hidpp on the kernel side.
I verified the sysfs interface for battery charge/discharge status, and
capacity read to be working on my 910-007290 device (white).
The Legion Go features detachable controllers which support a dual
dinput mode. In this mode, the controllers appear under a single HID
device with two applications.
Currently, both controllers appear under the same event device, causing
their controls to be mixed up. This patch separates the two so that
they can be used independently.
In addition, the latest firmware update for the Legion Go swaps the IDs
to the ones used by the Legion Go 2, so add those IDs as well.
A malicious HID device can trigger a slab out-of-bounds during
mt_report_fixup() by passing in report descriptor smaller than
607 bytes. mt_report_fixup() attempts to patch byte offset 607
of the descriptor with 0x25 by first checking if byte offset
607 is 0x15 however it lacks bounds checks to verify if the
descriptor is big enough before conducting this check. Fix
this bug by ensuring the descriptor size is at least 608
bytes before accessing it.
Below is the KASAN splat after the out of bounds access happens:
After hid_hw_start() is called hidinput_connect() will eventually be
called to set up the device with the input layer since the
HID_CONNECT_DEFAULT connect mask is used. During hidinput_connect()
all input and output reports are processed and corresponding hid_inputs
are allocated and configured via hidinput_configure_usages(). This
process involves slot tagging report fields and configuring usages
by setting relevant bits in the capability bitmaps. However it is possible
that the capability bitmaps are not set at all leading to the subsequent
hidinput_has_been_populated() check to fail leading to the freeing of the
hid_input and the underlying input device.
This becomes problematic because a malicious HID device like a
ASUS ROG N-Key keyboard can trigger the above scenario via a
specially crafted descriptor which then leads to a user-after-free
when the name of the freed input device is written to later on after
hid_hw_start(). Below, report 93 intentionally utilises the
HID_UP_UNDEFINED Usage Page which is skipped during usage
configuration, leading to the frees.
Prior to the topology parsing rewrite and the switchover to the new parsing
logic for AMD processors in
c749ce393b8f ("x86/cpu: Use common topology code for AMD"),
the initial_apicid on these platforms was:
- First initialized to the LocalApicId from CPUID leaf 0x1 EBX[31:24].
- Then overwritten by the ExtendedLocalApicId in CPUID leaf 0xb
EDX[31:0] on processors that supported topoext.
With the new parsing flow introduced in
f7fb3b2dd92c ("x86/cpu: Provide an AMD/HYGON specific topology parser"),
parse_8000_001e() now unconditionally overwrites the initial_apicid already
parsed during cpu_parse_topology_ext().
Although this has not been a problem on baremetal platforms, on virtualized AMD
guests that feature more than 255 cores, QEMU zeros out the CPUID leaf
0x8000001e on CPUs with CoreID > 255 to prevent collision of these IDs in
EBX[7:0] which can only represent a maximum of 255 cores [1].
This results in the following FW_BUG being logged when booting a guest
with more than 255 cores:
[Firmware Bug]: CPU 512: APIC ID mismatch. CPUID: 0x0000 APIC: 0x0200
AMD64 Architecture Programmer's Manual Volume 2: System Programming Pub.
24593 Rev. 3.42 [2] Section 16.12 "x2APIC_ID" mentions the Extended
Enumeration leaf 0xb (Fn0000_000B_EDX[31:0])(which was later superseded by the
extended leaf 0x80000026) provides the full x2APIC ID under all circumstances
unlike the one reported by CPUID leaf 0x8000001e EAX which depends on the mode
in which APIC is configured.
Rely on the APIC ID parsed during cpu_parse_topology_ext() from CPUID leaf
0x80000026 or 0xb and only use the APIC ID from leaf 0x8000001e if
cpu_parse_topology_ext() failed (has_topoext is false).
On platforms that support the 0xb leaf (Zen2 or later, AMD guests on
QEMU) or the extended leaf 0x80000026 (Zen4 or later), the
initial_apicid is now set to the value parsed from EDX[31:0].
On older AMD/Hygon platforms that do not support the 0xb leaf but support the
TOPOEXT extension (families 0x15, 0x16, 0x17[Zen1], and Hygon), retain current
behavior where the initial_apicid is set using the 0x8000001e leaf.
Issue debugged by Naveen N Rao (AMD) <naveen@kernel.org> and Sairaj Kodilkar
<sarunkod@amd.com>.
min and dest_id are guest-controlled indices. Using array_index_nospec()
after the bounds checks clamps these values to mitigate speculative execution
side-channels.
When the driver is removed the clocks are disabled twice: once in
macb_remove and a second time by runtime pm. Disable wakeup in remove so
all the clocks are disabled and skip the second call to macb_clks_disable.
Always suspend the device as we always set it active in probe.
Fixes: d54f89af6cc4 ("net: macb: Add pm runtime support") Signed-off-by: Neil Mandir <neil.mandir@seco.com> Co-developed-by: Sean Anderson <sean.anderson@linux.dev> Signed-off-by: Sean Anderson <sean.anderson@linux.dev> Link: https://patch.msgid.link/20250826143022.935521-1-sean.anderson@linux.dev Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
If dentry->d_name.len < EFI_VARIABLE_GUID_LEN , 'guid' can become
negative, leadings to oob. The issue can be triggered by parallel
lookups using invalid filename:
T1 T2
lookup_open
->lookup
simple_lookup
d_add
// invalid dentry is added to hash list
lookup_open
d_alloc_parallel
__d_lookup_rcu
__d_lookup_rcu_op_compare
hlist_bl_for_each_entry_rcu
// invalid dentry can be retrieved
->d_compare
efivarfs_d_compare
// oob
Fix it by checking 'guid' before cmp.
Fixes: da27a24383b2 ("efivarfs: guid part of filenames are case-insensitive") Signed-off-by: Li Nan <linan122@huawei.com> Signed-off-by: Wu Guanghao <wuguanghao3@huawei.com> Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
A bit of digging showed that we were invoking the phylink_resume as a part
of the fbnic_up path when we were enabling the service task while not
holding the RTNL lock. We should be enabling this sooner as a part of the
ndo_open path and then just letting the service task come online later.
This will help to enforce the correct locking and brings the phylink
interface online at the same time as the network interface, instead of at a
later time.
I tested this on QEMU to verify this was working by putting the system to
sleep using "echo mem > /sys/power/state" to put the system to sleep in the
guest and then using the command "system_wakeup" in the QEMU monitor.
pppol2tp_session_get_sock() is using RCU, it must be ready
for sk_refcnt being zero.
Commit ee40fb2e1eb5 ("l2tp: protect sock pointer of
struct pppol2tp_session with RCU") was correct because it
had a call_rcu(..., pppol2tp_put_sk) which was later removed in blamed commit.
pppol2tp_recv() can use pppol2tp_session_get_sock() as well.
Fixes: c5cbaef992d6 ("l2tp: refactor ppp socket/session relationship") Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: James Chapman <jchapman@katalix.com> Reviewed-by: Guillaume Nault <gnault@redhat.com> Link: https://patch.msgid.link/20250826134435.1683435-1-edumazet@google.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Current implementation maintains two separate reference counting
mechanisms: the 'count' field in struct rose_neigh tracks references from
rose_node structures, while the 'use' field (now refcount_t) tracks
references from rose_sock.
This patch merges these two reference counting systems using 'use' field
for proper reference management. Specifically, this patch adds incrementing
and decrementing of rose_neigh->use when rose_neigh->count is incremented
or decremented.
This patch also modifies rose_rt_free(), rose_rt_device_down() and
rose_clear_route() to properly release references to rose_neigh objects
before freeing a rose_node through rose_remove_node().
These changes ensure rose_neigh structures are properly freed only when
all references, including those from rose_node structures, are released.
As a result, this resolves a slab-use-after-free issue reported by Syzbot.
The 'use' field in struct rose_neigh is used as a reference counter but
lacks atomicity. This can lead to race conditions where a rose_neigh
structure is freed while still being referenced by other code paths.
For example, when rose_neigh->use becomes zero during an ioctl operation
via rose_rt_ioctl(), the structure may be removed while its timer is
still active, potentially causing use-after-free issues.
This patch changes the type of 'use' from unsigned short to refcount_t and
updates all code paths to use rose_neigh_hold() and rose_neigh_put() which
operate reference counts atomically.
The current rose_remove_neigh() performs two distinct operations:
1. Removes rose_neigh from rose_neigh_list
2. Frees the rose_neigh structure
Split these operations into separate functions to improve maintainability
and prepare for upcoming refcount_t conversion. The timer cleanup remains
in rose_remove_neigh() because free operations can be called from timer
itself.
This patch introduce rose_neigh_put() to handle the freeing of rose_neigh
structures and modify rose_remove_neigh() to handle removal only.
Signed-off-by: Takamitsu Iwai <takamitz@amazon.co.jp> Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com> Link: https://patch.msgid.link/20250823085857.47674-2-takamitz@amazon.co.jp Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stable-dep-of: d860d1faa6b2 ("net: rose: convert 'use' field to refcount_t") Signed-off-by: Sasha Levin <sashal@kernel.org>
The hv_netvsc driver currently enables NAPI after opening the primary and
subchannels. This ordering creates a race: if the Hyper-V host places data
in the host -> guest ring buffer and signals the channel before
napi_enable() has been called, the channel callback will run but
napi_schedule_prep() will return false. As a result, the NAPI poller never
gets scheduled, the data in the ring buffer is not consumed, and the
receive queue may remain permanently stuck until another interrupt happens
to arrive.
Fix this by enabling NAPI and registering it with the RX/TX queues before
vmbus channel is opened. This guarantees that any early host signal after
open will correctly trigger NAPI scheduling and the ring buffer will be
drained.
Signed-off-by: Joe Damato <jdamato@fastly.com> Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com> Tested-by: Shradha Gupta <shradhagupta@linux.microsoft.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Stable-dep-of: 9448ccd85336 ("net: hv_netvsc: fix loss of early receive events from host during channel open.") Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently, in the AF_XDP transmit paths, the CIC bit of
TX Desc3 is set for all packets. Setting this bit for
packets transmitting through queues that don't support
checksum offloading causes the TX DMA to get stuck after
transmitting some packets. This patch ensures the CIC bit
of TX Desc3 is set only if the TX queue supports checksum
offloading.
Fixes: 132c32ee5bc0 ("net: stmmac: Add TX via XDP zero-copy socket") Signed-off-by: Rohan G Thomas <rohan.g.thomas@altera.com> Reviewed-by: Matthew Gerlach <matthew.gerlach@altera.com> Link: https://patch.msgid.link/20250825-xgmac-minor-fixes-v3-3-c225fe4444c0@altera.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Correct supported speed modes as per the XGMAC databook.
Commit 9cb54af214a7 ("net: stmmac: Fix IP-cores specific
MAC capabilities") removes support for 10M, 100M and
1000HD. 1000HD is not supported by XGMAC IP, but it does
support 10M and 100M FD mode for XGMAC version >= 2_20,
and it also supports 10M and 100M HD mode if the HDSEL bit
is set in the MAC_HW_FEATURE0 reg. This commit enables support
for 10M and 100M speed modes for XGMAC IP based on XGMAC
version and MAC capabilities.
Fixes: 9cb54af214a7 ("net: stmmac: Fix IP-cores specific MAC capabilities") Signed-off-by: Rohan G Thomas <rohan.g.thomas@altera.com> Reviewed-by: Matthew Gerlach <matthew.gerlach@altera.com> Link: https://patch.msgid.link/20250825-xgmac-minor-fixes-v3-2-c225fe4444c0@altera.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Enabling RX FIFO Overflow interrupts is counterproductive
and causes an interrupt storm when RX FIFO overflows.
Disabling this interrupt has no side effect and eliminates
interrupt storms when the RX FIFO overflows.
Commit 8a7cb245cf28 ("net: stmmac: Do not enable RX FIFO
overflow interrupts") disables RX FIFO overflow interrupts
for DWMAC4 IP and removes the corresponding handling of
this interrupt. This patch is doing the same thing for
XGMAC IP.
Fixes: 2142754f8b9c ("net: stmmac: Add MAC related callbacks for XGMAC2") Signed-off-by: Rohan G Thomas <rohan.g.thomas@altera.com> Reviewed-by: Matthew Gerlach <matthew.gerlach@altera.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Link: https://patch.msgid.link/20250825-xgmac-minor-fixes-v3-1-c225fe4444c0@altera.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The local Xoff value is being set before the firmware (FW) update.
In case of a failure where the FW is not updated with the new value,
there is no fallback to the previous value.
Update the local Xoff value after the FW has been successfully set.
Xon/Xoff sizes are derived from calculations that include
the port speed.
These settings need to be updated and applied whenever the
port speed is changed.
The port speed is typically set after the physical link goes down
and is negotiated as part of the link-up process between the two
connected interfaces.
Xon/Xoff parameters being updated at the point where the new
negotiated speed is established.
If PF (Physical Function) has SFs (Sub-Functions), since the SFs are not
taking part in the synchronization flow, sync reset can lead to fatal
error on the SFs, as the function will be closed unexpectedly from the
SF point of view.
Add a check to prevent sync reset when there are SFs on a PF device
which is not ECPF, as ECPF is teardowned gracefully before reset.
Fixes: 92501fa6e421 ("net/mlx5: Ack on sync_reset_request only if PF can do reset_now") Signed-off-by: Moshe Shemesh <moshe@nvidia.com> Reviewed-by: Parav Pandit <parav@nvidia.com> Reviewed-by: Tariq Toukan <tariqt@nvidia.com> Signed-off-by: Mark Bloch <mbloch@nvidia.com> Link: https://patch.msgid.link/20250825143435.598584-8-mbloch@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Fix lockdep assertion triggered during sync reset unload event. When the
sync reset flow is initiated using the devlink reload fw_activate
option, the PF already holds the devlink lock while handling unload
event. In this case, delegate sync reset unload event handling back to
the devlink callback process to avoid double-locking and resolve the
lockdep warning.
The devlink reload fw_activate command performs firmware activation
followed by driver reload, while devlink reload driver_reinit triggers
only driver reload. However, the driver reload logic differs between the
two modes, as on driver_reinit mode mlx5 also reloads auxiliary drivers,
while in fw_activate mode the auxiliary drivers are suspended where
applicable.
Additionally, following the cited commit, if the device has multiple PFs,
the behavior during fw_activate may vary between PFs: one PF may suspend
auxiliary drivers, while another reloads them.
Align devlink dev reload fw_activate behavior with devlink dev reload
driver_reinit, to reload all auxiliary drivers.
Fixes: 72ed5d5624af ("net/mlx5: Suspend auxiliary devices only in case of PCI device suspend") Signed-off-by: Moshe Shemesh <moshe@nvidia.com> Reviewed-by: Tariq Toukan <tariqt@nvidia.com> Reviewed-by: Akiva Goldberger <agoldberger@nvidia.com> Signed-off-by: Mark Bloch <mbloch@nvidia.com> Link: https://patch.msgid.link/20250825143435.598584-6-mbloch@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The HW resource reservation logic allows the L2 driver to use the
RoCE resources if the RoCE driver is not registered. When calculating
the stats contexts available for L2, we should not blindly subtract
the stats contexts reserved for RoCE unless the RoCE driver is
registered. This bug may cause the L2 rings to be less than the
number requested when we are close to running out of stats contexts.
Fixes: 2e4592dc9bee ("bnxt_en: Change MSIX/NQs allocation policy") Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com> Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Link: https://patch.msgid.link/20250825175927.459987-4-michael.chan@broadcom.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Before we accept an ethtool request to increase a resource (such as
rings), we call the FW to check that the requested resource is likely
available first before we commit. But it is still possible that
the actual reservation or allocation can fail. The existing code
is missing the logic to adjust the TX rings in case the reserved
TX rings are less than requested. Add a warning message (a similar
message for RX rings already exists) and add the logic to adjust
the TX rings. Without this fix, the number of TX rings reported
to the stack can exceed the actual TX rings and ethtool -l will
report more than the actual TX rings.
Fixes: 674f50a5b026 ("bnxt_en: Implement new method to reserve rings.") Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com> Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Link: https://patch.msgid.link/20250825175927.459987-3-michael.chan@broadcom.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
bnxt_set_dflt_rings() assumes that it is always called before any TC has
been created. So it doesn't take bp->num_tc into account and assumes
that it is always 0 or 1.
In the FW resource or capability change scenario, the FW will return
flags in bnxt_hwrm_if_change() that will cause the driver to
reinitialize and call bnxt_cancel_reservations(). This will lead to
bnxt_init_dflt_ring_mode() calling bnxt_set_dflt_rings() and bp->num_tc
may be greater than 1. This will cause bp->tx_ring[] to be sized too
small and cause memory corruption in bnxt_alloc_cp_rings().
Fix it by properly scaling the TX rings by bp->num_tc in the code
paths mentioned above. Add 2 helper functions to determine
bp->tx_nr_rings and bp->tx_nr_rings_per_tc.
Fixes: ec5d31e3c15d ("bnxt_en: Handle firmware reset status during IF_UP.") Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com> Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com> Signed-off-by: Sreekanth Reddy <sreekanth.reddy@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Link: https://patch.msgid.link/20250825175927.459987-2-michael.chan@broadcom.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
It looks like that every time when the interface was set down and up the
driver was creating a new ptp clock. On top of this the function
ptp_clock_unregister was never called.
Therefore fix this by calling ptp_clock_register and initialize the
mii_ts struct inside the probe function and call ptp_clock_unregister when
driver is removed.
On the first validate of an imported dma-buf (initial bind), the device
has no GPU mappings, so a rebind is unnecessary. Rebinding here is
harmful in multi-GPU setups and for VMs using preempt-fence mode, as it
would evict in-flight GPU work.
v2:
- Drop dma_buf_validated, check for XE_PL_SYSTEM (Thomas)
Marking ufence as signalled after copy_to_user() is too late.
Worker thread which signals ufence by memory write might be raced
with another userspace vm-bind call. In map/unmap scenario unmap
may still see ufence is not signalled causing -EBUSY. Change the
order of marking / write to user-fence fixes this issue.
The communication buffer allocated by setup_mm_hdr() is later on passed
to tee_shm_register_kernel_buf(). The latter expects those buffers to be
contiguous pages, but setup_mm_hdr() just uses kmalloc(). That can cause
various corruptions or BUGs, specifically since commit 9aec2fb0fd5e
("slab: allocate frozen pages"), though it was broken before as well.
Fix this by using alloc_pages_exact() instead of kmalloc().
`McstFramesRcvdOk` counts the number of received multicast packets, and
it reports the value correctly.
However, reading `McstFramesRcvdOk` clears the register to zero. As a
result, the driver was reporting only the packets since the last read,
instead of the accumulated total.
Fix this by updating the multicast statistics accumulatively instaed of
instantaneously.
None of MDP5 platforms have a LUT clock on the display-controller, it
was added by the mistake. Drop it, fixing DT warnings on MSM8976 /
MSM8956 platforms. Technically it's an ABI break, but no other platforms
are affected.
Fixes: 385c8ac763b3 ("dt-bindings: display/msm: convert MDP5 schema to YAML format") Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com> Acked-by: Rob Herring (Arm) <robh@kernel.org>
Patchwork: https://patchwork.freedesktop.org/patch/667822/ Signed-off-by: Rob Clark <robin.clark@oss.qualcomm.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently, the driver increments `alloc_page_failed` when buffer allocation fails
in `ice_clean_rx_irq()`. However, this counter is intended for page allocation
failures, not buffer allocation issues.
This patch corrects the counter by incrementing `alloc_buf_failed` instead,
ensuring accurate statistics reporting for buffer allocation failures.
Fixes: 2fba7dc5157b ("ice: Add support for XDP multi-buffer on Rx side") Reported-by: Jacob Keller <jacob.e.keller@intel.com> Suggested-by: Paul Menzel <pmenzel@molgen.mpg.de> Signed-off-by: Michal Kubiak <michal.kubiak@intel.com> Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de> Reviewed-by: Jason Xing <kerneljasonxing@gmail.com> Reviewed-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com> Tested-by: Priya Singh <priyax.singh@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The ice_adapter structure is used by the ice driver to connect multiple
physical functions of a device in software. It was introduced by
commit 0e2bddf9e5f9 ("ice: add ice_adapter for shared data across PFs on
the same NIC") and is primarily used for PTP support, as well as for
handling certain cross-PF synchronization.
The original design of ice_adapter used PCI address information to
determine which devices should be connected. This was extended to support
E825C devices by commit fdb7f54700b1 ("ice: Initial support for E825C
hardware in ice_adapter"), which used the device ID for E825C devices
instead of the PCI address.
Later, commit 0093cb194a75 ("ice: use DSN instead of PCI BDF for
ice_adapter index") replaced the use of Bus/Device/Function addressing with
use of the device serial number.
E825C devices may appear in "Dual NAC" configuration which has multiple
physical devices tied to the same clock source and which need to use the
same ice_adapter. Unfortunately, each "NAC" has its own NVM which has its
own unique Device Serial Number. Thus, use of the DSN for connecting
ice_adapter does not work properly. It "worked" in the pre-production
systems because the DSN was not initialized on the test NVMs and all the
NACs had the same zero'd serial number.
Since we cannot rely on the DSN, lets fall back to the logic in the
original E825C support which used the device ID. This is safe for E825C
only because of the embedded nature of the device. It isn't a discreet
adapter that can be plugged into an arbitrary system. All E825C devices on
a given system are connected to the same clock source and need to be
configured through the same PTP clock.
To make this separation clear, reserve bit 63 of the 64-bit index values as
a "fixed index" indicator. Always clear this bit when using the device
serial number as an index.
For E825C, use a fixed value defined as the 0x579C E825C backplane device
ID bitwise ORed with the fixed index indicator. This is slightly different
than the original logic of just using the device ID directly. Doing so
prevents a potential issue with systems where only one of the NACs is
connected with an external PHY over SGMII. In that case, one NAC would
have the E825C_SGMII device ID, but the other would not.
Separate the determination of the full 64-bit index from the 32-bit
reduction logic. Provide both ice_adapter_index() and a wrapping
ice_adapter_xa_index() which handles reducing the index to a long on 32-bit
systems. As before, cache the full index value in the adapter structure to
warn about collisions.
This fixes issues with E825C not initializing PTP on both NACs, due to
failure to connect the appropriate devices to the same ice_adapter.
Fixes: 0093cb194a75 ("ice: use DSN instead of PCI BDF for ice_adapter index") Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Reviewed-by: Grzegorz Nitka <grzegorz.nitka@intel.com> Reviewed-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com> Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com> Tested-by: Rinitha S <sx.rinitha@intel.com> (A Contingent worker at Intel) Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The ice_cfg_tx_topo function attempts to apply Tx scheduler topology
configuration based on NVM parameters, selecting either a 5 or 9 layer
topology.
As part of this flow, the driver acquires the "Global Configuration Lock",
which is a hardware resource associated with programming the DDP package
to the device. This "lock" is implemented by firmware as a way to
guarantee that only one PF can program the DDP for a device. Unlike a
traditional lock, once a PF has acquired this lock, no other PF will be
able to acquire it again (including that PF) until a CORER of the device.
Future requests to acquire the lock report that global configuration has
already completed.
The following flow is used to program the Tx topology:
* Read the DDP package for scheduler configuration data
* Acquire the global configuration lock
* Program Tx scheduler topology according to DDP package data
* Trigger a CORER which clears the global configuration lock
This is followed by the flow for programming the DDP package:
* Acquire the global configuration lock (again)
* Download the DDP package to the device
* Release the global configuration lock.
However, if configuration of the Tx topology fails, (i.e.
ice_get_set_tx_topo returns an error code), the driver exits
ice_cfg_tx_topo() immediately, and fails to trigger CORER.
While the global configuration lock is held, the firmware rejects most
AdminQ commands, as it is waiting for the DDP package download (or Tx
scheduler topology programming) to occur.
The current driver flows assume that the global configuration lock has been
reset by CORER after programming the Tx topology. Thus, the same PF
attempts to acquire the global lock again, and fails. This results in the
driver reporting "an unknown error occurred when loading the DDP package".
It then attempts to enter safe mode, but ultimately fails to finish
ice_probe() since nearly all AdminQ command report error codes, and the
driver stops loading the device at some point during its initialization.
The only currently known way that ice_get_set_tx_topo() can fail is with
certain older DDP packages which contain invalid topology configuration, on
firmware versions which strictly validate this data. The most recent
releases of the DDP have resolved the invalid data. However, it is still
poor practice to essentially brick the device, and prevent access to the
device even through safe mode or recovery mode. It is also plausible that
this command could fail for some other reason in the future.
We cannot simply release the global lock after a failed call to
ice_get_set_tx_topo(). Releasing the lock indicates to firmware that global
configuration (downloading of the DDP) has completed. Future attempts by
this or other PFs to load the DDP will fail with a report that the DDP
package has already been downloaded. Then, PFs will enter safe mode as they
realize that the package on the device does not meet the minimum version
requirement to load. The reported error messages are confusing, as they
indicate the version of the default "safe mode" package in the NVM, rather
than the version of the file loaded from /lib/firmware.
Instead, we need to trigger CORER to clear global configuration. This is
the lowest level of hardware reset which clears the global configuration
lock and related state. It also clears any already downloaded DDP.
Crucially, it does *not* clear the Tx scheduler topology configuration.
Refactor ice_cfg_tx_topo() to always trigger a CORER after acquiring the
global lock, regardless of success or failure of the topology
configuration.
We need to re-initialize the HW structure when we trigger the CORER. Thus,
it makes sense for this to be the responsibility of ice_cfg_tx_topo()
rather than its caller, ice_init_tx_topology(). This avoids needless
re-initialization in cases where we don't attempt to update the Tx
scheduler topology, such as if it has already been programmed.
There is one catch: failure to re-initialize the HW struct should stop
ice_probe(). If this function fails, we won't have a valid HW structure and
cannot ensure the device is functioning properly. To handle this, ensure
ice_cfg_tx_topo() returns a limited set of error codes. Set aside one
specifically, -ENODEV, to indicate that the ice_init_tx_topology() should
fail and stop probe.
Other error codes indicate failure to apply the Tx scheduler topology. This
is treated as a non-fatal error, with an informational message informing
the system administrator that the updated Tx topology did not apply. This
allows the device to load and function with the default Tx scheduler
topology, rather than failing to load entirely.
Note that this use of CORER will not result in loops with future PFs
attempting to also load the invalid Tx topology configuration. The first PF
will acquire the global configuration lock as part of programming the DDP.
Each PF after this will attempt to acquire the global lock as part of
programming the Tx topology, and will fail with the indication from
firmware that global configuration is already complete. Tx scheduler
topology configuration is only performed during driver init (probe or
devlink reload) and not during cleanup for a CORER that happens after probe
completes.
Fixes: 91427e6d9030 ("ice: Support 5 layer topology") Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Reviewed-by: Simon Horman <horms@kernel.org> Tested-by: Rinitha S <sx.rinitha@intel.com> (A Contingent worker at Intel) Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
projects / thirdparty / kernel / stable.git / log
