Only amd_pstate_epp driver (i.e. cppc_state = ACTIVE) enters the
amd_pstate_epp_offline() and amd_pstate_epp_cpu_online() functions,
so remove the unnecessary if condition checking if cppc_state is
equal to AMD_PSTATE_ACTIVE.
Signed-off-by: Dhananjay Ugwekar <Dhananjay.Ugwekar@amd.com> Reviewed-by: Mario Limonciello <mario.limonciello@amd.com> Reviewed-by: Gautham R. Shenoy <gautham.shenoy@amd.com> Link: https://lore.kernel.org/r/20241204144842.164178-5-Dhananjay.Ugwekar@amd.com Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Stable-dep-of: 3ace20038e19 ("cpufreq/amd-pstate: Fix cpufreq_policy ref counting") Signed-off-by: Sasha Levin <sashal@kernel.org>
This commit addresses an issue where clk_gating.state is being toggled in
ufshcd_setup_clocks() even if clock gating is not allowed.
The fix is to add a check for hba->clk_gating.is_initialized before toggling
clk_gating.state in ufshcd_setup_clocks().
Since clk_gating.lock is now initialized unconditionally, it can no longer
lead to the spinlock being used before it is properly initialized, but
instead it is mostly for documentation purposes.
Fixes: 1ab27c9cf8b6 ("ufs: Add support for clock gating") Reported-by: Geert Uytterhoeven <geert@linux-m68k.org> Tested-by: Geert Uytterhoeven <geert+renesas@glider.be> Signed-off-by: Avri Altman <avri.altman@wdc.com> Link: https://lore.kernel.org/r/20250128071207.75494-3-avri.altman@wdc.com Reviewed-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Introduce a new clock gating lock to serialize access to some of the clock
gating members instead of the host_lock.
While at it, simplify the code with the guard() macro and co for automatic
cleanup of the new lock. There are some explicit
spin_lock_irqsave()/spin_unlock_irqrestore() snaking instances I left
behind because I couldn't make heads or tails of it.
Additionally, move the trace_ufshcd_clk_gating() call from inside the
region protected by the lock as it doesn't needs protection.
Signed-off-by: Avri Altman <avri.altman@wdc.com> Link: https://lore.kernel.org/r/20241124070808.194860-4-avri.altman@wdc.com Reviewed-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Stable-dep-of: 839a74b5649c ("scsi: ufs: Fix toggling of clk_gating.state when clock gating is not allowed") Signed-off-by: Sasha Levin <sashal@kernel.org>
Remove hba->clk_gating.active_reqs check from ufshcd_is_ufs_dev_busy()
function to separate clock gating logic from general device busy checks.
Signed-off-by: Avri Altman <avri.altman@wdc.com> Link: https://lore.kernel.org/r/20241124070808.194860-3-avri.altman@wdc.com Reviewed-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Stable-dep-of: 839a74b5649c ("scsi: ufs: Fix toggling of clk_gating.state when clock gating is not allowed") Signed-off-by: Sasha Levin <sashal@kernel.org>
Prepare to remove hba->clk_gating.active_reqs check from
ufshcd_is_ufs_dev_busy().
Signed-off-by: Avri Altman <avri.altman@wdc.com> Link: https://lore.kernel.org/r/20241124070808.194860-2-avri.altman@wdc.com Reviewed-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Stable-dep-of: 839a74b5649c ("scsi: ufs: Fix toggling of clk_gating.state when clock gating is not allowed") Signed-off-by: Sasha Levin <sashal@kernel.org>
The following bug report happened with a PREEMPT_RT kernel:
BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48
in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 2012, name: kwatchdog
preempt_count: 1, expected: 0
RCU nest depth: 0, expected: 0
get_random_u32+0x4f/0x110
clocksource_verify_choose_cpus+0xab/0x1a0
clocksource_verify_percpu.part.0+0x6b/0x330
clocksource_watchdog_kthread+0x193/0x1a0
It is due to the fact that clocksource_verify_choose_cpus() is invoked with
preemption disabled. This function invokes get_random_u32() to obtain
random numbers for choosing CPUs. The batched_entropy_32 local lock and/or
the base_crng.lock spinlock in driver/char/random.c will be acquired during
the call. In PREEMPT_RT kernel, they are both sleeping locks and so cannot
be acquired in atomic context.
Fix this problem by using migrate_disable() to allow smp_processor_id() to
be reliably used without introducing atomic context. preempt_disable() is
then called after clocksource_verify_choose_cpus() but before the
clocksource measurement is being run to avoid introducing unexpected
latency.
Fixes: 7560c02bdffb ("clocksource: Check per-CPU clock synchronization when marked unstable") Suggested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Waiman Long <longman@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Link: https://lore.kernel.org/all/20250131173323.891943-2-longman@redhat.com Signed-off-by: Sasha Levin <sashal@kernel.org>
The "Checking clocksource synchronization" message is normally printed
when clocksource_verify_percpu() is called for a given clocksource if
both the CLOCK_SOURCE_UNSTABLE and CLOCK_SOURCE_VERIFY_PERCPU flags
are set.
It is an informational message and so pr_info() is the correct choice.
Signed-off-by: Waiman Long <longman@redhat.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Acked-by: John Stultz <jstultz@google.com> Link: https://lore.kernel.org/all/20250125015442.3740588-1-longman@redhat.com
Stable-dep-of: 6bb05a33337b ("clocksource: Use migrate_disable() to avoid calling get_random_u32() in atomic context") Signed-off-by: Sasha Levin <sashal@kernel.org>
Some lwtunnels have a dst cache for post-transformation dst.
If the packet destination did not change we may end up recording
a reference to the lwtunnel in its own cache, and the lwtunnel
state will never be freed.
Discovered by the ioam6.sh test, kmemleak was recently fixed
to catch per-cpu memory leaks. I'm not sure if rpl and seg6
can actually hit this, but in principle I don't see why not.
Fixes: 8cb3bf8bff3c ("ipv6: ioam: Add support for the ip6ip6 encapsulation") Fixes: 6c8702c60b88 ("ipv6: sr: add support for SRH encapsulation and injection with lwtunnels") Fixes: a7a29f9c361f ("net: ipv6: add rpl sr tunnel") Reviewed-by: Simon Horman <horms@kernel.org> Link: https://patch.msgid.link/20250130031519.2716843-2-kuba@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
This patch mitigates the two-reallocations issue with rpl_iptunnel by
providing the dst_entry (in the cache) to the first call to
skb_cow_head(). As a result, the very first iteration would still
trigger two reallocations (i.e., empty cache), while next iterations
would only trigger a single reallocation.
Performance tests before/after applying this patch, which clearly shows
there is no impact (it even shows improvement):
- before: https://ibb.co/nQJhqwc
- after: https://ibb.co/4ZvW6wV
Signed-off-by: Justin Iurman <justin.iurman@uliege.be> Cc: Alexander Aring <aahringo@redhat.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Stable-dep-of: 92191dd10730 ("net: ipv6: fix dst ref loops in rpl, seg6 and ioam6 lwtunnels") Signed-off-by: Sasha Levin <sashal@kernel.org>
This patch mitigates the two-reallocations issue with seg6_iptunnel by
providing the dst_entry (in the cache) to the first call to
skb_cow_head(). As a result, the very first iteration would still
trigger two reallocations (i.e., empty cache), while next iterations
would only trigger a single reallocation.
Performance tests before/after applying this patch, which clearly shows
the improvement:
- before: https://ibb.co/3Cg4sNH
- after: https://ibb.co/8rQ350r
Signed-off-by: Justin Iurman <justin.iurman@uliege.be> Cc: David Lebrun <dlebrun@google.com> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Stable-dep-of: 92191dd10730 ("net: ipv6: fix dst ref loops in rpl, seg6 and ioam6 lwtunnels") Signed-off-by: Sasha Levin <sashal@kernel.org>
This patch mitigates the two-reallocations issue with ioam6_iptunnel by
providing the dst_entry (in the cache) to the first call to
skb_cow_head(). As a result, the very first iteration may still trigger
two reallocations (i.e., empty cache), while next iterations would only
trigger a single reallocation.
Performance tests before/after applying this patch, which clearly shows
the improvement:
- inline mode:
- before: https://ibb.co/LhQ8V63
- after: https://ibb.co/x5YT2bS
- encap mode:
- before: https://ibb.co/3Cjm5m0
- after: https://ibb.co/TwpsxTC
- encap mode with tunsrc:
- before: https://ibb.co/Gpy9QPg
- after: https://ibb.co/PW1bZFT
This patch also fixes an incorrect behavior: after the insertion, the
second call to skb_cow_head() makes sure that the dev has enough
headroom in the skb for layer 2 and stuff. In that case, the "old"
dst_entry was used, which is now fixed. After discussing with Paolo, it
appears that both patches can be merged into a single one -this one-
(for the sake of readability) and target net-next.
Signed-off-by: Justin Iurman <justin.iurman@uliege.be> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Stable-dep-of: 92191dd10730 ("net: ipv6: fix dst ref loops in rpl, seg6 and ioam6 lwtunnels") Signed-off-by: Sasha Levin <sashal@kernel.org>
Add static inline dst_dev_overhead() function to include/net/dst.h. This
helper function is used by ioam6_iptunnel, rpl_iptunnel and
seg6_iptunnel to get the dev's overhead based on a cache entry
(dst_entry). If the cache is empty, the default and generic value
skb->mac_len is returned. Otherwise, LL_RESERVED_SPACE() over dst's dev
is returned.
Signed-off-by: Justin Iurman <justin.iurman@uliege.be> Cc: Alexander Lobakin <aleksander.lobakin@intel.com> Cc: Vadim Fedorenko <vadim.fedorenko@linux.dev> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Stable-dep-of: 92191dd10730 ("net: ipv6: fix dst ref loops in rpl, seg6 and ioam6 lwtunnels") Signed-off-by: Sasha Levin <sashal@kernel.org>
At btrfs_write_check() if our file's i_size is not sector size aligned and
we have a write that starts at an offset larger than the i_size that falls
within the same page of the i_size, then we end up not zeroing the file
range [i_size, write_offset).
ret = btrfs_cont_expand(BTRFS_I(inode), oldsize, end_pos);
if (ret)
return ret;
}
So if our file's i_size is 90269 bytes and a write at offset 90365 bytes
comes in, we get 'start_pos' set to 90112 bytes, which is less than the
i_size and therefore we don't zero out the range [90269, 90365) by
calling btrfs_cont_expand().
This is an old bug introduced in commit 9036c10208e1 ("Btrfs: update hole
handling v2"), from 2008, and the buggy code got moved around over the
years.
Fix this by discarding 'start_pos' and comparing against the write offset
('pos') without any alignment.
This bug was recently exposed by test case generic/363 which tests this
scenario by polluting ranges beyond EOF with an mmap write and than verify
that after a file increases we get zeroes for the range which is supposed
to be a hole and not what we wrote with the previous mmaped write.
We're only seeing this exposed now because generic/363 used to run only
on xfs until last Sunday's fstests update.
generic/363 0s ... [failed, exit status 1]- output mismatch (see /home/fdmanana/git/hub/xfstests/results//generic/363.out.bad)
# --- tests/generic/363.out 2025-02-05 15:31:14.013646509 +0000
# +++ /home/fdmanana/git/hub/xfstests/results//generic/363.out.bad 2025-02-05 17:25:33.112630781 +0000
@@ -1 +1,46 @@
QA output created by 363
+READ BAD DATA: offset = 0xdcad, size = 0xd921, fname = /home/fdmanana/btrfs-tests/dev/junk
+OFFSET GOOD BAD RANGE
+0x1609d 0x0000 0x3104 0x0
+operation# (mod 256) for the bad data may be 4
+0x1609e 0x0000 0x0472 0x1
+operation# (mod 256) for the bad data may be 4
...
(Run 'diff -u /home/fdmanana/git/hub/xfstests/tests/generic/363.out /home/fdmanana/git/hub/xfstests/results//generic/363.out.bad' to see the entire diff)
Ran: generic/363
Failures: generic/363
Failed 1 of 1 tests
Add a check for the return value of mlxsw_sp_port_get_stats_raw()
in __mlxsw_sp_port_get_stats(). If mlxsw_sp_port_get_stats_raw()
returns an error, exit the function to prevent further processing
with potentially invalid data.
Fixes: 614d509aa1e7 ("mlxsw: Move ethtool_ops to spectrum_ethtool.c") Cc: stable@vger.kernel.org # 5.9+ Signed-off-by: Wentao Liang <vulab@iscas.ac.cn> Reviewed-by: Petr Machata <petrm@nvidia.com> Link: https://patch.msgid.link/20250212152311.1332-1-vulab@iscas.ac.cn Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The netfs library could break down a read request into
multiple subrequests. When multichannel is used, there is
potential to improve performance when each of these
subrequests pick a different channel.
Today we call cifs_pick_channel when the main read request
is initialized in cifs_init_request. This change moves this to
cifs_prepare_read, which is the right place to pick channel since
it gets called for each subrequest.
Interestingly cifs_prepare_write already does channel selection
for individual subreq, but looks like it was missed for read.
This is especially important when multichannel is used with
increased rasize.
In my test setup, with rasize set to 8MB, a sequential read
of large file was taking 11.5s without this change. With the
change, it completed in 9s. The difference is even more signigicant
with bigger rasize.
Cc: <stable@vger.kernel.org> Cc: David Howells <dhowells@redhat.com> Signed-off-by: Shyam Prasad N <sprasad@microsoft.com> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Set the buffer type to IGC_TX_BUFFER_TYPE_SKB for empty frame in the
igc_init_empty_frame function. This ensures that the buffer type is
correctly identified and handled during Tx ring cleanup.
Fixes: db0b124f02ba ("igc: Enhance Qbv scheduling by using first flag bit") Cc: stable@vger.kernel.org # 6.2+ Signed-off-by: Song Yoong Siang <yoong.siang.song@intel.com> Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com> Reviewed-by: Simon Horman <horms@kernel.org> Tested-by: Mor Bar-Gabay <morx.bar.gabay@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For hs400(es) mode, the 'hs400-ds-delay' is typically configured in the
dts. However, some projects may only define 'mediatek,hs400-ds-dly3',
which can lead to initialization failures in hs400es mode. CMD13 reported
response crc error in the mmc_switch_status() just after switching to
hs400es mode.
Currently, the hs400_ds_dly3 value is set within the tuning function. This
means that the PAD_DS_DLY3 field is not configured before tuning process,
which is the reason for the above-mentioned CMD13 response crc error.
Move the PAD_DS_DLY3 field configuration into msdc_prepare_hs400_tuning(),
and add a value check of hs400_ds_delay to prevent overwriting by zero when
the 'hs400-ds-delay' is not set in the dts. In addition, since hs400(es)
only tune the PAD_DS_DLY1, the PAD_DS_DLY2_SEL bit should be cleared to
bypass it.
Fixes: c4ac38c6539b ("mmc: mtk-sd: Add HS400 online tuning support") Signed-off-by: Andy-ld Lu <andy-ld.lu@mediatek.com> Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20250123092644.7359-1-andy-ld.lu@mediatek.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A recent LLVM commit [1] started generating an .ARM.attributes section
similar to the one that exists for 32-bit, which results in orphan
section warnings (or errors if CONFIG_WERROR is enabled) from the linker
because it is not handled in the arm64 linker scripts.
ld.lld: error: arch/arm64/kernel/vdso/vgettimeofday.o:(.ARM.attributes) is being placed in '.ARM.attributes'
ld.lld: error: arch/arm64/kernel/vdso/vgetrandom.o:(.ARM.attributes) is being placed in '.ARM.attributes'
ld.lld: error: vmlinux.a(lib/vsprintf.o):(.ARM.attributes) is being placed in '.ARM.attributes'
ld.lld: error: vmlinux.a(lib/win_minmax.o):(.ARM.attributes) is being placed in '.ARM.attributes'
ld.lld: error: vmlinux.a(lib/xarray.o):(.ARM.attributes) is being placed in '.ARM.attributes'
Discard the new sections in the necessary linker scripts to resolve the
warnings, as the kernel and vDSO do not need to retain it, similar to
the .note.gnu.property section.
The iopf_queue_remove_device() helper removes a device from the per-iommu
iopf queue when PRI is disabled on the device. It responds to all
outstanding iopf's with an IOMMU_PAGE_RESP_INVALID code and detaches the
device from the queue.
However, it fails to release the group structure that represents a group
of iopf's awaiting for a response after responding to the hardware. This
can cause a memory leak if iopf_queue_remove_device() is called with
pending iopf's.
Fix it by calling iopf_free_group() after the iopf group is responded.
Fixes: 199112327135 ("iommu: Track iopf group instead of last fault") Cc: stable@vger.kernel.org Suggested-by: Kevin Tian <kevin.tian@intel.com> Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Link: https://lore.kernel.org/r/20250117055800.782462-1-baolu.lu@linux.intel.com Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
scx_move_task() is called from sched_move_task() and tells the BPF scheduler
that cgroup migration is being committed. sched_move_task() is used by both
cgroup and autogroup migrations and scx_move_task() tried to filter out
autogroup migrations by testing the destination cgroup and PF_EXITING but
this is not enough. In fact, without explicitly tagging the thread which is
doing the cgroup migration, there is no good way to tell apart
scx_move_task() invocations for racing migration to the root cgroup and an
autogroup migration.
This led to scx_move_task() incorrectly ignoring a migration from non-root
cgroup to an autogroup of the root cgroup triggering the following warning:
Fix it by adding an argument to sched_move_task() that indicates whether the
moving is for a cgroup or autogroup migration. After the change,
scx_move_task() is called only for cgroup migrations and renamed to
scx_cgroup_move_task().
- The bailout for a bad partoffset must use put_dev_sector(), since the
preceding read_part_sector() succeeded.
- If the partition table claims a silly sector size like 0xfff bytes
(which results in partition table entries straddling sector boundaries),
bail out instead of accessing out-of-bounds memory.
- We must not assume that the partition table contains proper NUL
termination - use strnlen() and strncmp() instead of strlen() and
strcmp().
The stmpe_reg_read function can fail, but its return value is not checked
in stmpe_gpio_irq_sync_unlock. This can lead to silent failures and
incorrect behavior if the hardware access fails.
This patch adds checks for the return value of stmpe_reg_read. If the
function fails, an error message is logged and the function returns
early to avoid further issues.
Fixes: b888fb6f2a27 ("gpio: stmpe: i2c transfer are forbiden in atomic context") Cc: stable@vger.kernel.org # 4.16+ Signed-off-by: Wentao Liang <vulab@iscas.ac.cn> Link: https://lore.kernel.org/r/20250212021849.275-1-vulab@iscas.ac.cn Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Spurious immediate wake up events are reported on Acer Nitro ANV14. GPIO 11 is
specified as an edge triggered input and also a wake source but this pin is
supposed to be an output pin for an LED, so it's effectively floating.
Block the interrupt from getting set up for this GPIO on this device.
In contrast to the commit message of the fixed commit VFs whose parent
PF is not configured are not always isolated, that is put on their own
PCI domain. This is because for VFs to be added to an existing PCI
domain it is enough for that PCI domain to share the same topology ID or
PCHID. Such a matching PCI domain without a parent PF may exist when
a PF from the same PCI card created the domain with the VF being a child
of a different, non accessible, PF. While not causing technical issues
it makes the rules which VFs are isolated inconsistent.
Fix this by explicitly checking that the parent PF exists on the PCI
domain determined by the topology ID or PCHID before registering the VF.
This works because a parent PF which is under control of this Linux
instance must be enabled and configured at the point where its child VFs
appear because otherwise SR-IOV could not have been enabled on the
parent.
This creates a new zpci_iov_find_parent_pf() function which a future
commit can use to find if a VF has a configured parent PF. Use
zdev->rid instead of zdev->devfn such that the new function can be used
before it has been decided if the RID will be exposed and zdev->devfn is
set. Also handle the hypotheical case that the RID is not available but
there is an otherwise matching zbus.
do_page_fault() and do_entUna() are special because they use
non-standard stack frame layout. Fix them manually.
Cc: stable@vger.kernel.org Tested-by: Maciej W. Rozycki <macro@orcam.me.uk> Tested-by: Magnus Lindholm <linmag7@gmail.com> Tested-by: Matt Turner <mattst88@gmail.com> Reviewed-by: Maciej W. Rozycki <macro@orcam.me.uk> Suggested-by: Maciej W. Rozycki <macro@orcam.me.uk> Signed-off-by: Ivan Kokshaysky <ink@unseen.parts> Signed-off-by: Matt Turner <mattst88@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This allows the assembly in entry.S to automatically keep in sync with
changes in the stack layout (struct pt_regs and struct switch_stack).
Cc: stable@vger.kernel.org Tested-by: Maciej W. Rozycki <macro@orcam.me.uk> Tested-by: Matt Turner <mattst88@gmail.com> Reviewed-by: Maciej W. Rozycki <macro@orcam.me.uk> Signed-off-by: Ivan Kokshaysky <ink@unseen.parts> Signed-off-by: Matt Turner <mattst88@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When flushing the serial port's buffer, uart_flush_buffer() calls
kfifo_reset() but if there is an outstanding DMA transfer then the
completion function will consume data from the kfifo via
uart_xmit_advance(), underflowing and leading to ongoing DMA as the
driver tries to transmit another 2^32 bytes.
This is readily reproduced with serial-generic and amidi sending even
short messages as closing the device on exit will wait for the fifo to
drain and in the underflow case amidi hangs for 30 seconds on exit in
tty_wait_until_sent(). A trace of that gives:
Since the port lock is held in when the kfifo is reset in
uart_flush_buffer() and in __dma_tx_complete(), adding a flush_buffer
hook to adjust the outstanding DMA byte count is sufficient to avoid the
kfifo underflow.
The documentation of the __uart_read_properties() states that
->iotype member is always altered after the function call, but
the code doesn't do that in the case when use_defaults == false
and the value of reg-io-width is unsupported. Make sure the code
follows the documentation.
Note, the current users of the uart_read_and_validate_port_properties()
will fail and the change doesn't affect their behaviour, neither
users of uart_read_port_properties() will be affected since the
alteration happens there even in the current code flow.
Fixes: e894b6005dce ("serial: port: Introduce a common helper to read properties") Cc: stable <stable@kernel.org> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Link: https://lore.kernel.org/r/20250124161530.398361-3-andriy.shevchenko@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently the ->iotype is always assigned to the UPIO_MEM when
the respective property is not found. However, this will not
support the cases when user wants to have UPIO_PORT to be set
or preserved. Support this scenario by checking ->iobase value
and default the ->iotype respectively.
Fixes: 1117a6fdc7c1 ("serial: 8250_of: Switch to use uart_read_port_properties()") Fixes: e894b6005dce ("serial: port: Introduce a common helper to read properties") Cc: stable <stable@kernel.org> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Link: https://lore.kernel.org/r/20250124161530.398361-2-andriy.shevchenko@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Tejun reported the following race between fork() and cgroup.kill at [1].
Tejun:
I was looking at cgroup.kill implementation and wondering whether there
could be a race window. So, __cgroup_kill() does the following:
k1. Set CGRP_KILL.
k2. Iterate tasks and deliver SIGKILL.
k3. Clear CGRP_KILL.
The copy_process() does the following:
c1. Copy a bunch of stuff.
c2. Grab siglock.
c3. Check fatal_signal_pending().
c4. Commit to forking.
c5. Release siglock.
c6. Call cgroup_post_fork() which puts the task on the css_set and tests
CGRP_KILL.
The intention seems to be that either a forking task gets SIGKILL and
terminates on c3 or it sees CGRP_KILL on c6 and kills the child. However, I
don't see what guarantees that k3 can't happen before c6. ie. After a
forking task passes c5, k2 can take place and then before the forking task
reaches c6, k3 can happen. Then, nobody would send SIGKILL to the child.
What am I missing?
This is indeed a race. One way to fix this race is by taking
cgroup_threadgroup_rwsem in write mode in __cgroup_kill() as the fork()
side takes cgroup_threadgroup_rwsem in read mode from cgroup_can_fork()
to cgroup_post_fork(). However that would be heavy handed as this adds
one more potential stall scenario for cgroup.kill which is usually
called under extreme situation like memory pressure.
To fix this race, let's maintain a sequence number per cgroup which gets
incremented on __cgroup_kill() call. On the fork() side, the
cgroup_can_fork() will cache the sequence number locally and recheck it
against the cgroup's sequence number at cgroup_post_fork() site. If the
sequence numbers mismatch, it means __cgroup_kill() can been called and
we should send SIGKILL to the newly created task.
Starting with Rust 1.86.0 (to be released 2025-04-03), Clippy will have
a new lint, `doc_overindented_list_items` [1], which catches cases of
overindented list items.
The lint has been added by Yutaro Ohno, based on feedback from the kernel
[2] on a patch that fixed a similar case -- commit 0c5928deada1 ("rust:
block: fix formatting in GenDisk doc").
Clippy reports a few cases in the kernel, apart from the one already
fixed in the commit above. One is this one:
error: doc list item overindented
--> rust/kernel/rbtree.rs:1152:5
|
1152 | /// null, it is a pointer to the root of the [`RBTree`].
| ^^^^ help: try using ` ` (2 spaces)
|
= help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#doc_overindented_list_items
= note: `-D clippy::doc-overindented-list-items` implied by `-D warnings`
= help: to override `-D warnings` add `#[allow(clippy::doc_overindented_list_items)]`
Starting with Rust 1.85.0 (currently in beta, to be released 2025-02-20),
under some kernel configurations with `CONFIG_RUST_DEBUG_ASSERTIONS=y`,
one may trigger a new `objtool` warning:
rust/kernel.o: warning: objtool: _R...securityNtB2_11SecurityCtx8as_bytes()
falls through to next function _R...core3ops4drop4Drop4drop()
due to a call to the `noreturn` symbol:
core::panicking::assert_failed::<usize, usize>
Thus add it to the list so that `objtool` knows it is actually `noreturn`.
Do so matching with `strstr` since it is a generic.
See commit 56d680dd23c3 ("objtool/rust: list `noreturn` Rust functions")
for more details.
Cc: stable@vger.kernel.org # Needed in 6.12.y and 6.13.y only (Rust is pinned in older LTSs). Fixes: 56d680dd23c3 ("objtool/rust: list `noreturn` Rust functions") Reviewed-by: Gary Guo <gary@garyguo.net> Link: https://lore.kernel.org/r/20250112143951.751139-1-ojeda@kernel.org
[ Updated Cc: stable@ to include 6.13.y. - Miguel ] Signed-off-by: Miguel Ojeda <ojeda@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Starting with Rust 1.85.0 (to be released 2025-02-20), `rustc` warns
[1] about disabling neon in the aarch64 hardfloat target:
warning: target feature `neon` cannot be toggled with
`-Ctarget-feature`: unsound on hard-float targets
because it changes float ABI
|
= note: this was previously accepted by the compiler but
is being phased out; it will become a hard error
in a future release!
= note: for more information, see issue #116344
<https://github.com/rust-lang/rust/issues/116344>
Thus, instead, use the softfloat target instead.
While trying it out, I found that the kernel sanitizers were not enabled
for that built-in target [2]. Upstream Rust agreed to backport
the enablement for the current beta so that it is ready for
the 1.85.0 release [3] -- thanks!
However, that still means that before Rust 1.85.0, we cannot switch
since sanitizers could be in use. Thus conditionally do so.
UEFI 2.11 introduced EFI_MEMORY_HOT_PLUGGABLE to annotate system memory
regions that are 'cold plugged' at boot, i.e., hot pluggable memory that
is available from early boot, and described as system RAM by the
firmware.
Existing loaders and EFI applications running in the boot context will
happily use this memory for allocating data structures that cannot be
freed or moved at runtime, and this prevents the memory from being
unplugged. Going forward, the new EFI_MEMORY_HOT_PLUGGABLE attribute
should be tested, and memory annotated as such should be avoided for
such allocations.
In the EFI stub, there are a couple of occurrences where, instead of the
high-level AllocatePages() UEFI boot service, a low-level code sequence
is used that traverses the EFI memory map and carves out the requested
number of pages from a free region. This is needed, e.g., for allocating
as low as possible, or for allocating pages at random.
While AllocatePages() should presumably avoid special purpose memory and
cold plugged regions, this manual approach needs to incorporate this
logic itself, in order to prevent the kernel itself from ending up in a
hot unpluggable region, preventing it from being unplugged.
So add the EFI_MEMORY_HOTPLUGGABLE macro definition, and check for it
where appropriate.
scripts/Makefile.clang was changed in the linked commit to move --target from
KBUILD_CFLAGS to KBUILD_CPPFLAGS, as that generally has a broader scope.
However that variable is not inspected by the userprogs logic,
breaking cross compilation on clang.
Use both variables to detect bitsize and target arguments for userprogs.
The Mediatek MT7922 WiFi device advertises FLR support, but it apparently
does not work, and all subsequent config reads return ~0:
pci 0000:01:00.0: [14c3:0616] type 00 class 0x028000 PCIe Endpoint
pciback 0000:01:00.0: not ready 65535ms after FLR; giving up
After an FLR, pci_dev_wait() waits for the device to become ready. Prior
to d591f6804e7e ("PCI: Wait for device readiness with Configuration RRS"),
it polls PCI_COMMAND until it is something other that PCI_POSSIBLE_ERROR
(~0). If it times out, pci_dev_wait() returns -ENOTTY and
__pci_reset_function_locked() tries the next available reset method.
Typically this is Secondary Bus Reset, which does work, so the MT7922 is
eventually usable.
After d591f6804e7e, if Configuration Request Retry Status Software
Visibility (RRS SV) is enabled, pci_dev_wait() polls PCI_VENDOR_ID until it
is something other than the special 0x0001 Vendor ID that indicates a
completion with RRS status.
When RRS SV is enabled, reads of PCI_VENDOR_ID should return either 0x0001,
i.e., the config read was completed with RRS, or a valid Vendor ID. On the
MT7922, it seems that all config reads after FLR return ~0 indefinitely.
When pci_dev_wait() reads PCI_VENDOR_ID and gets 0xffff, it assumes that's
a valid Vendor ID and the device is now ready, so it returns with success.
After pci_dev_wait() returns success, we restore config space and continue.
Since the MT7922 is not actually ready after the FLR, the restore fails and
the device is unusable.
We considered changing pci_dev_wait() to continue polling if a
PCI_VENDOR_ID read returns either 0x0001 or 0xffff. This "works" as it did
before d591f6804e7e, although we have to wait for the timeout and then fall
back to SBR. But it doesn't work for SR-IOV VFs, which *always* return
0xffff as the Vendor ID.
Mark Mediatek MT7922 WiFi devices to avoid the use of FLR completely. This
will cause fallback to another reset method, such as SBR.
In the US country code, to avoid including 6 GHz rules in the 5 GHz rules
list, the number of 5 GHz rules is set to a default constant value of 4
(REG_US_5G_NUM_REG_RULES). However, if there are more than 4 valid 5 GHz
rules, the current logic will bypass the legitimate 6 GHz rules.
For example, if there are 5 valid 5 GHz rules and 1 valid 6 GHz rule, the
current logic will only consider 4 of the 5 GHz rules, treating the last
valid rule as a 6 GHz rule. Consequently, the actual 6 GHz rule is never
processed, leading to the eventual disabling of 6 GHz channels.
To fix this issue, instead of hardcoding the value to 4, use a helper
function to determine the number of 6 GHz rules present in the 5 GHz rules
list and ignore only those rules.
The problem is that GCC expects 16-byte alignment of the incoming stack
since early 2004, as Maciej found out [1]:
Having actually dug speculatively I can see that the psABI was changed in
GCC 3.5 with commit e5e10fb4a350 ("re PR target/14539 (128-bit long double
improperly aligned)") back in Mar 2004, when the stack pointer alignment
was increased from 8 bytes to 16 bytes, and arch/alpha/kernel/entry.S has
various suspicious stack pointer adjustments, starting with SP_OFF which
is not a whole multiple of 16.
Also, as Magnus noted, "ALPHA Calling Standard" [2] required the same:
D.3.1 Stack Alignment
This standard requires that stacks be octaword aligned at the time a
new procedure is invoked.
However:
- the "normal" kernel stack is always misaligned by 8 bytes, thanks to
the odd number of 64-bit words in 'struct pt_regs', which is the very
first thing pushed onto the kernel thread stack;
- syscall, fault, interrupt etc. handlers may, or may not, receive aligned
stack depending on numerous factors.
Somehow we got away with it until recently, when we ended up with
a stack corruption in kernel/smp.c:smp_call_function_single() due to
its use of 32-byte aligned local data and the compiler doing clever
things allocating it on the stack.
This adds padding between the PAL-saved and kernel-saved registers
so that 'struct pt_regs' have an even number of 64-bit words.
This makes the stack properly aligned for most of the kernel
code, except two handlers which need special threatment.
Note: struct pt_regs doesn't belong in uapi/asm; this should be fixed,
but let's put this off until later.
The driver assumed that es58x_dev->udev->serial could never be NULL.
While this is true on commercially available devices, an attacker
could spoof the device identity providing a NULL USB serial number.
That would trigger a NULL pointer dereference.
Add a check on es58x_dev->udev->serial before accessing it.
Reported-by: yan kang <kangyan91@outlook.com> Reported-by: yue sun <samsun1006219@gmail.com> Closes: https://lore.kernel.org/linux-can/SY8P300MB0421E0013C0EBD2AA46BA709A1F42@SY8P300MB0421.AUSP300.PROD.OUTLOOK.COM/ Fixes: 9f06631c3f1f ("can: etas_es58x: export product information through devlink_ops::info_get()") Signed-off-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr> Link: https://patch.msgid.link/20250204154859.9797-2-mailhol.vincent@wanadoo.fr Cc: stable@vger.kernel.org Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The J1939 standard requires the transmission of messages of length 0.
For example proprietary messages are specified with a data length of 0
to 1785. The transmission of such messages is not possible. Sending
results in no error being returned but no corresponding can frame
being generated.
Enable the transmission of zero length J1939 messages. In order to
facilitate this two changes are necessary:
1) If the transmission of a new message is requested from user space
the message is segmented in j1939_sk_send_loop(). Let the segmentation
take into account zero length messages, do not terminate immediately,
queue the corresponding skb.
2) j1939_session_skb_get_by_offset() selects the next skb to transmit
for a session. Take into account that there might be zero length skbs
in the queue.
Signed-off-by: Alexander Hölzl <alexander.hoelzl@gmx.net> Acked-by: Oleksij Rempel <o.rempel@pengutronix.de> Link: https://patch.msgid.link/20250205174651.103238-1-alexander.hoelzl@gmx.net Fixes: 9d71dd0c7009 ("can: add support of SAE J1939 protocol") Cc: stable@vger.kernel.org
[mkl: commit message rephrased] Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Runtime PM is enabled as one of the last steps of probe(), so all
earlier gotos to "exit_free_device" label were not correct and were
leading to unbalanced runtime PM disable depth.
Fixes: 6e2fe01dd6f9 ("can: c_can: move runtime PM enable/disable to c_can_platform") Cc: stable@vger.kernel.org Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Reviewed-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr> Link: https://patch.msgid.link/20250112-syscon-phandle-args-can-v1-1-314d9549906f@linaro.org Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If skb allocation fails, the pointer to struct can_frame is NULL. This
is actually handled everywhere inside ctucan_err_interrupt() except for
the only place.
Add the missed NULL check.
Found by Linux Verification Center (linuxtesting.org) with SVACE static
analysis tool.
Fixes: 2dcb8e8782d8 ("can: ctucanfd: add support for CTU CAN FD open-source IP core - bus independent part.") Cc: stable@vger.kernel.org Signed-off-by: Fedor Pchelkin <pchelkin@ispras.ru> Acked-by: Pavel Pisa <pisa@cmp.felk.cvut.cz> Reviewed-by: Vincent Mailhol <mailhol.vincent@wanadoo.fr> Link: https://patch.msgid.link/20250114152138.139580-1-pchelkin@ispras.ru Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
MeiG Smart SLM828 is an LTE-A CAT6 modem with the mPCIe form factor. The
"Cls=ff(vend.) Sub=10 Prot=02" and "Cls=ff(vend.) Sub=10 Prot=03"
interfaces respond to AT commands. Add these interfaces.
The product ID the modem uses is shared across multiple modems. Therefore,
add comments to describe which interface is used for which modem.
device_del() can lead to new work being scheduled in gadget->work
workqueue. This is observed, for example, with the dwc3 driver with the
following call stack:
device_del()
gadget_unbind_driver()
usb_gadget_disconnect_locked()
dwc3_gadget_pullup()
dwc3_gadget_soft_disconnect()
usb_gadget_set_state()
schedule_work(&gadget->work)
Move flush_work() after device_del() to ensure the workqueue is cleaned
up.
Fixes: 5702f75375aa9 ("usb: gadget: udc-core: move sysfs_notify() to a workqueue") Cc: stable <stable@kernel.org> Signed-off-by: Roy Luo <royluo@google.com> Reviewed-by: Alan Stern <stern@rowland.harvard.edu> Reviewed-by: Thinh Nguyen <Thinh.Nguyen@synopsys.com> Link: https://lore.kernel.org/r/20250204233642.666991-1-royluo@google.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If we receive an initial fragment of size 8 bytes which specifies a wLength
of 1 byte (so the reassembled message is supposed to be 9 bytes long), and
we then receive a second fragment of size 9 bytes (which is not supposed to
happen), we currently wrongly bypass the fragment reassembly code but still
pass the pointer to the acm->notification_buffer to
acm_process_notification().
Make this less wrong by always going through fragment reassembly when we
expect more fragments.
Before this patch, receiving an overlong fragment could lead to `newctrl`
in acm_process_notification() being uninitialized data (instead of data
coming from the device).
If the first fragment is shorter than struct usb_cdc_notification, we can't
calculate an expected_size. Log an error and discard the notification
instead of reading lengths from memory outside the received data, which can
lead to memory corruption when the expected_size decreases between
fragments, causing `expected_size - acm->nb_index` to wrap.
This issue has been present since the beginning of git history; however,
it only leads to memory corruption since commit ea2583529cd1
("cdc-acm: reassemble fragmented notifications").
A mitigating factor is that acm_ctrl_irq() can only execute after userspace
has opened /dev/ttyACM*; but if ModemManager is running, ModemManager will
do that automatically depending on the USB device's vendor/product IDs and
its other interfaces.
Add Renesas R-Car D3 USB Download mode quirk and update comments
on all the other Renesas R-Car USB Download mode quirks to discern
them from each other. This follows R-Car Series, 3rd Generation
reference manual Rev.2.00 chapter 19.2.8 USB download mode .
Fixes: 6d853c9e4104 ("usb: cdc-acm: Add DISABLE_ECHO for Renesas USB Download mode") Cc: stable <stable@kernel.org> Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org> Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be> Link: https://lore.kernel.org/r/20250209145708.106914-1-marek.vasut+renesas@mailbox.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The cause of this error is that the device has two interfaces, and the
hub driver binds to interface 1 instead of interface 0, which is where
usb_hub_to_struct_hub() looks.
We can prevent the problem from occurring by refusing to accept hub
devices that violate the USB spec by having more than one
configuration or interface.
Reported-and-tested-by: Robert Morris <rtm@csail.mit.edu> Cc: stable <stable@kernel.org> Closes: https://lore.kernel.org/linux-usb/95564.1737394039@localhost/ Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Link: https://lore.kernel.org/r/c27f3bf4-63d8-4fb5-ac82-09e3cd19f61c@rowland.harvard.edu Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
While the MIDI jacks are configured correctly, and the MIDIStreaming
endpoint descriptors are filled with the correct information,
bNumEmbMIDIJack and bLength are set incorrectly in these descriptors.
This does not matter when the numbers of in and out ports are equal, but
when they differ the host will receive broken descriptors with
uninitialized stack memory leaking into the descriptor for whichever
value is smaller.
The precise meaning of "in" and "out" in the port counts is not clearly
defined and can be confusing. But elsewhere the driver consistently
uses this to match the USB meaning of IN and OUT viewed from the host,
so that "in" ports send data to the host and "out" ports receive data
from it.
Cc: stable <stable@kernel.org> Fixes: c8933c3f79568 ("USB: gadget: f_midi: allow a dynamic number of input and output ports") Signed-off-by: John Keeping <jkeeping@inmusicbrands.com> Reviewed-by: Takashi Iwai <tiwai@suse.de> Link: https://lore.kernel.org/r/20250130195035.3883857-1-jkeeping@inmusicbrands.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The fastboot tool for communicating with Android bootloaders does not
work reliably with this device if USB 2 Link Power Management (LPM)
is enabled.
Various fastboot commands are affected, including the
following, which usually reproduces the problem within two tries:
fastboot getvar kernel
getvar:kernel FAILED (remote: 'GetVar Variable Not found')
This issue was hidden on many systems up until commit 63a1f8454962
("xhci: stored cached port capability values in one place") as the xhci
driver failed to detect USB 2 LPM support if USB 3 ports were listed
before USB 2 ports in the "supported protocol capabilities".
Adding the quirk resolves the issue. No drawbacks are expected since
the device uses different USB product IDs outside of fastboot mode, and
since fastboot commands worked before, until LPM was enabled on the
tested system by the aforementioned commit.
Based on a patch from Forest <forestix@nom.one> from which most of the
code and commit message is taken.
Teclast disk used on Huawei hisi platforms doesn't work well,
losing connectivity intermittently if LPM is enabled.
Add quirk disable LPM to resolve the issue.
When usb_control_msg is used in the get_bMaxPacketSize0 function, the
USB pipe does not include the endpoint device number. This can cause
failures when a usb hub port is reinitialized after encountering a bad
cable connection. As a result, the system logs the following error
messages:
usb usb2-port1: cannot reset (err = -32)
usb usb2-port1: Cannot enable. Maybe the USB cable is bad?
usb usb2-port1: attempt power cycle
usb 2-1: new high-speed USB device number 5 using ci_hdrc
usb 2-1: device descriptor read/8, error -71
The problem began after commit 85d07c556216 ("USB: core: Unite old
scheme and new scheme descriptor reads"). There
usb_get_device_descriptor was replaced with get_bMaxPacketSize0. Unlike
usb_get_device_descriptor, the get_bMaxPacketSize0 function uses the
macro usb_rcvaddr0pipe, which does not include the endpoint device
number. usb_get_device_descriptor, on the other hand, used the macro
usb_rcvctrlpipe, which includes the endpoint device number.
By modifying the get_bMaxPacketSize0 function to use usb_rcvctrlpipe
instead of usb_rcvaddr0pipe, the issue can be resolved. This change will
ensure that the endpoint device number is included in the USB pipe,
preventing reinitialization failures. If the endpoint has not set the
device number yet, it will still work because the device number is 0 in
udev.
Cc: stable <stable@kernel.org> Fixes: 85d07c556216 ("USB: core: Unite old scheme and new scheme descriptor reads") Signed-off-by: Stefan Eichenberger <stefan.eichenberger@toradex.com> Reviewed-by: Alan Stern <stern@rowland.harvard.edu> Link: https://lore.kernel.org/r/20250203105840.17539-1-eichest@gmail.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
LS7A EHCI controller doesn't have extended capabilities, so the EECP
(EHCI Extended Capabilities Pointer) field of HCCPARAMS register should
be 0x0, but it reads as 0xa0 now. This is a hardware flaw and will be
fixed in future, now just clear the EECP field to avoid error messages
on boot:
Some Renesas HCs require firmware upload to work, this is handled by the
xhci_pci_renesas driver. Other variants of those chips load firmware from
a SPI flash and are ready to work with xhci_pci alone.
A refactor merged in v6.12 broke the latter configuration so that users
are finding their hardware ignored by the normal driver and are forced to
enable the firmware loader which isn't really necessary on their systems.
Let xhci_pci work with those chips as before when the firmware loader is
disabled by kernel configuration.
Fixes: 25f51b76f90f ("xhci-pci: Make xhci-pci-renesas a proper modular driver") Cc: stable <stable@kernel.org> Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219616 Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219726 Signed-off-by: Michal Pecio <michal.pecio@gmail.com> Tested-by: Nicolai Buchwitz <nb@tipi-net.de> Link: https://lore.kernel.org/r/20250128104529.58a79bfc@foxbook Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In dwc2_hsotg_udc_start(), e.g. when binding composite driver, "of_node"
is set to hsotg->dev->of_node.
It causes errors when binding the gadget driver several times, on
stm32mp157c-ev1 board. Below error is seen:
"pin PA10 already requested by 49000000.usb-otg; cannot claim for gadget.0"
The first time, no issue is seen as when registering the driver, of_node
isn't NULL:
-> gadget_dev_desc_UDC_store
-> usb_gadget_register_driver_owner
-> driver_register
...
-> really_probe -> pinctrl_bind_pins (no effect)
Then dwc2_hsotg_udc_start() sets of_node.
The second time (stop the gadget, reconfigure it, then start it again),
of_node has been set, so the probing code tries to acquire pins for the
gadget. These pins are hold by the controller, hence the error.
So clear gadget.dev.of_node in udc_stop() routine to avoid the issue.
drivers/usb/gadget/udc/renesas_usb3.c: In function 'renesas_usb3_probe':
drivers/usb/gadget/udc/renesas_usb3.c:2638:73: warning: '%d'
directive output may be truncated writing between 1 and 11 bytes into a
region of size 6 [-Wformat-truncation=]
2638 | snprintf(usb3_ep->ep_name, sizeof(usb3_ep->ep_name), "ep%d", i);
^~~~~~~~~~~~~~~~~~~~~~~~ ^~ ^
The role switch registration and set_role() can happen in parallel as they
are invoked independent of each other. There is a possibility that a driver
might spend significant amount of time in usb_role_switch_register() API
due to the presence of time intensive operations like component_add()
which operate under common mutex. This leads to a time window after
allocating the switch and before setting the registered flag where the set
role notifications are dropped. Below timeline summarizes this behavior
Thread1 | Thread2
usb_role_switch_register() |
| |
---> allocate switch |
| |
---> component_add() | usb_role_switch_set_role()
| | |
| | --> Drop role notifications
| | since sw->registered
| | flag is not set.
| |
--->Set registered flag.|
To avoid this, set the registered flag early on in the switch register
API.
Fixes: b787a3e78175 ("usb: roles: don't get/set_role() when usb_role_switch is unregistered") Cc: stable <stable@kernel.org> Signed-off-by: Elson Roy Serrao <quic_eserrao@quicinc.com> Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> Link: https://lore.kernel.org/r/20250206193950.22421-1-quic_eserrao@quicinc.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a frequent timeout during controller enter/exit from halt state
after toggling the run_stop bit by SW. This timeout occurs when
performing frequent role switches between host and device, causing
device enumeration issues due to the timeout. This issue was not present
when USB2 suspend PHY was disabled by passing the SNPS quirks
(snps,dis_u2_susphy_quirk and snps,dis_enblslpm_quirk) from the DTS.
However, there is a requirement to enable USB2 suspend PHY by setting of
GUSB2PHYCFG.ENBLSLPM and GUSB2PHYCFG.SUSPHY bits when controller starts
in gadget or host mode results in the timeout issue.
This commit addresses this timeout issue by ensuring that the bits
GUSB2PHYCFG.ENBLSLPM and GUSB2PHYCFG.SUSPHY are cleared before starting
the dwc3_gadget_run_stop sequence and restoring them after the
dwc3_gadget_run_stop sequence is completed.
The current implementation sets the wMaxPacketSize of bulk in/out
endpoints to 1024 bytes at the end of the f_midi_bind function. However,
in cases where there is a failure in the first midi bind attempt,
consider rebinding. This scenario may encounter an f_midi_bind issue due
to the previous bind setting the bulk endpoint's wMaxPacketSize to 1024
bytes, which exceeds the ep->maxpacket_limit where configured dwc3 TX/RX
FIFO's maxpacket size of 512 bytes for IN/OUT endpoints in support HS
speed only.
Here the term "rebind" in this context refers to attempting to bind the
MIDI function a second time in certain scenarios. The situations where
rebinding is considered include:
* When there is a failure in the first UDC write attempt, which may be
caused by other functions bind along with MIDI.
* Runtime composition change : Example : MIDI,ADB to MIDI. Or MIDI to
MIDI,ADB.
This commit addresses this issue by resetting the wMaxPacketSize before
endpoint claim. And here there is no need to reset all values in the usb
endpoint descriptor structure, as all members except wMaxPacketSize and
bEndpointAddress have predefined values.
This ensures that restores the endpoint to its expected configuration,
and preventing conflicts with value of ep->maxpacket_limit. It also
aligns with the approach used in other function drivers, which treat
endpoint descriptors as if they were full speed before endpoint claim.
The pages_touched field represents the number of subbuffers in the ring
buffer that have content that can be read. This is used in accounting of
"dirty_pages" and "buffer_percent" to allow the user to wait for the
buffer to be filled to a certain amount before it reads the buffer in
blocking mode.
The persistent buffer never updated this value so it was set to zero, and
this accounting would take it as it had no content. This would cause user
space to wait for content even though there's enough content in the ring
buffer that satisfies the buffer_percent.
The meta data for a mapped ring buffer contains an array of indexes of all
the subbuffers. The first entry is the reader page, and the rest of the
entries lay out the order of the subbuffers in how the ring buffer link
list is to be created.
The validator currently makes sure that all the entries are within the
range of 0 and nr_subbufs. But it does not check if there are any
duplicates.
While working on the ring buffer, I corrupted this array, where I added
duplicates. The validator did not catch it and created the ring buffer
link list on top of it. Luckily, the corruption was only that the reader
page was also in the writer path and only presented corrupted data but did
not crash the kernel. But if there were duplicates in the writer side,
then it could corrupt the ring buffer link list and cause a crash.
Create a bitmask array with the size of the number of subbuffers. Then
clear it. When walking through the subbuf array checking to see if the
entries are within the range, test if its bit is already set in the
subbuf_mask. If it is, then there is duplicates and fail the validation.
If not, set the corresponding bit and continue.
Cc: stable@vger.kernel.org Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Cc: Vincent Donnefort <vdonnefort@google.com> Link: https://lore.kernel.org/20250214102820.7509ddea@gandalf.local.home Fixes: c76883f18e59b ("ring-buffer: Add test if range of boot buffer is valid") Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
But virt_to_page() does not work with vmap()'d memory which is what the
persistent ring buffer has. It is rather trivial to allow this, but for
now just disable mmap() of instances that have their ring buffer from the
reserve_mem option.
If an mmap() is performed on a persistent buffer it will return -ENODEV
just like it would if the .mmap field wasn't defined in the
file_operations structure.
Cc: stable@vger.kernel.org Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Cc: Vincent Donnefort <vdonnefort@google.com> Link: https://lore.kernel.org/20250214115547.0d7287d3@gandalf.local.home Fixes: 9b7bdf6f6ece6 ("tracing: Have trace_printk not use binary prints if boot buffer") Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Memory mapping the tracing ring buffer will disable resizing the buffer.
But if there's an error in the memory mapping like an invalid parameter,
the function exits out without re-enabling the resizing of the ring
buffer, preventing the ring buffer from being resized after that.
Explicitly clear DEBUGCTL.LBR when a CPU is starting, prior to purging the
LBR MSRs themselves, as at least one system has been found to transfer
control to the kernel with LBRs enabled (it's unclear whether it's a BIOS
flaw or a CPU goof). Because the kernel preserves the original DEBUGCTL,
even when toggling LBRs, leaving DEBUGCTL.LBR as is results in running
with LBRs enabled at all times.
The EAX of the CPUID Leaf 023H enumerates the mask of valid sub-leaves.
To tell the availability of the sub-leaf 1 (enumerate the counter mask),
perf should check the bit 1 (0x2) of EAS, rather than bit 0 (0x1).
The error is not user-visible on bare metal. Because the sub-leaf 0 and
the sub-leaf 1 are always available. However, it may bring issues in a
virtualization environment when a VMM only enumerates the sub-leaf 0.
Introduce the cpuid35_e?x to replace the macros, which makes the
implementation style consistent.
Fixes: eb467aaac21e ("perf/x86/intel: Support Architectural PerfMon Extension leaf") Signed-off-by: Kan Liang <kan.liang@linux.intel.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: stable@vger.kernel.org Link: https://lkml.kernel.org/r/20250129154820.3755948-3-kan.liang@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When preparing vmcb02 for nested VMRUN (or state restore), "enter" guest
mode prior to initializing the MMU for nested NPT so that guest_mode is
set in the MMU's role. KVM's model is that all L2 MMUs are tagged with
guest_mode, as the behavior of hypervisor MMUs tends to be significantly
different than kernel MMUs.
Practically speaking, the bug is relatively benign, as KVM only directly
queries role.guest_mode in kvm_mmu_free_guest_mode_roots() and
kvm_mmu_page_ad_need_write_protect(), which SVM doesn't use, and in paths
that are optimizations (mmu_page_zap_pte() and
shadow_mmu_try_split_huge_pages()).
And while the role is incorprated into shadow page usage, because nested
NPT requires KVM to be using NPT for L1, reusing shadow pages across L1
and L2 is impossible as L1 MMUs will always have direct=1, while L2 MMUs
will have direct=0.
Hoist the TLB processing and setting of HF_GUEST_MASK to the beginning
of the flow instead of forcing guest_mode in the MMU, as nothing in
nested_vmcb02_prepare_control() between the old and new locations touches
TLB flush requests or HF_GUEST_MASK, i.e. there's no reason to present
inconsistent vCPU state to the MMU.
Fixes: 69cb877487de ("KVM: nSVM: move MMU setup to nested_prepare_vmcb_control") Cc: stable@vger.kernel.org Reported-by: Yosry Ahmed <yosry.ahmed@linux.dev> Reviewed-by: Yosry Ahmed <yosry.ahmed@linux.dev> Link: https://lore.kernel.org/r/20250130010825.220346-1-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Move the conditional loading of hardware DR6 with the guest's DR6 value
out of the core .vcpu_run() loop to fix a bug where KVM can load hardware
with a stale vcpu->arch.dr6.
When the guest accesses a DR and host userspace isn't debugging the guest,
KVM disables DR interception and loads the guest's values into hardware on
VM-Enter and saves them on VM-Exit. This allows the guest to access DRs
at will, e.g. so that a sequence of DR accesses to configure a breakpoint
only generates one VM-Exit.
For DR0-DR3, the logic/behavior is identical between VMX and SVM, and also
identical between KVM_DEBUGREG_BP_ENABLED (userspace debugging the guest)
and KVM_DEBUGREG_WONT_EXIT (guest using DRs), and so KVM handles loading
DR0-DR3 in common code, _outside_ of the core kvm_x86_ops.vcpu_run() loop.
But for DR6, the guest's value doesn't need to be loaded into hardware for
KVM_DEBUGREG_BP_ENABLED, and SVM provides a dedicated VMCB field whereas
VMX requires software to manually load the guest value, and so loading the
guest's value into DR6 is handled by {svm,vmx}_vcpu_run(), i.e. is done
_inside_ the core run loop.
Unfortunately, saving the guest values on VM-Exit is initiated by common
x86, again outside of the core run loop. If the guest modifies DR6 (in
hardware, when DR interception is disabled), and then the next VM-Exit is
a fastpath VM-Exit, KVM will reload hardware DR6 with vcpu->arch.dr6 and
clobber the guest's actual value.
The bug shows up primarily with nested VMX because KVM handles the VMX
preemption timer in the fastpath, and the window between hardware DR6
being modified (in guest context) and DR6 being read by guest software is
orders of magnitude larger in a nested setup. E.g. in non-nested, the
VMX preemption timer would need to fire precisely between #DB injection
and the #DB handler's read of DR6, whereas with a KVM-on-KVM setup, the
window where hardware DR6 is "dirty" extends all the way from L1 writing
DR6 to VMRESUME (in L1).
L1's view:
==========
<L1 disables DR interception>
CPU 0/KVM-7289 [023] d.... 2925.640961: kvm_entry: vcpu 0
A: L1 Writes DR6
CPU 0/KVM-7289 [023] d.... 2925.640963: <hack>: Set DRs, DR6 = 0xffff0ff1
Reported-by: John Stultz <jstultz@google.com> Closes: https://lkml.kernel.org/r/CANDhNCq5_F3HfFYABqFGCA1bPd_%2BxgNj-iDQhH4tDk%2Bwi8iZZg%40mail.gmail.com Fixes: 375e28ffc0cf ("KVM: X86: Set host DR6 only on VMX and for KVM_DEBUGREG_WONT_EXIT") Fixes: d67668e9dd76 ("KVM: x86, SVM: isolate vcpu->arch.dr6 from vmcb->save.dr6") Cc: stable@vger.kernel.org Cc: Jim Mattson <jmattson@google.com> Tested-by: John Stultz <jstultz@google.com> Link: https://lore.kernel.org/r/20250125011833.3644371-1-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Advertise support for Hyper-V's SEND_IPI and SEND_IPI_EX hypercalls if and
only if the local API is emulated/virtualized by KVM, and explicitly reject
said hypercalls if the local APIC is emulated in userspace, i.e. don't rely
on userspace to opt-in to KVM_CAP_HYPERV_ENFORCE_CPUID.
Rejecting SEND_IPI and SEND_IPI_EX fixes a NULL-pointer dereference if
Hyper-V enlightenments are exposed to the guest without an in-kernel local
APIC:
Note, checking the sending vCPU is sufficient, as the per-VM irqchip_mode
can't be modified after vCPUs are created, i.e. if one vCPU has an
in-kernel local APIC, then all vCPUs have an in-kernel local APIC.
It malicious user provides a small pptable through sysfs and then
a bigger pptable, it may cause buffer overflow attack in function
smu_sys_set_pp_table().
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com> Signed-off-by: Jiang Liu <gerry@linux.alibaba.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The ELP worker needs to calculate new metric values for all neighbors
"reachable" over an interface. Some of the used metric sources require
locks which might need to sleep. This sleep is incompatible with the RCU
list iterator used for the recorded neighbors. The initial approach to work
around of this problem was to queue another work item per neighbor and then
run this in a new context.
Even when this solved the RCU vs might_sleep() conflict, it has a major
problems: Nothing was stopping the work item in case it is not needed
anymore - for example because one of the related interfaces was removed or
the batman-adv module was unloaded - resulting in potential invalid memory
accesses.
Directly canceling the metric worker also has various problems:
* cancel_work_sync for a to-be-deactivated interface is called with
rtnl_lock held. But the code in the ELP metric worker also tries to use
rtnl_lock() - which will never return in this case. This also means that
cancel_work_sync would never return because it is waiting for the worker
to finish.
* iterating over the neighbor list for the to-be-deactivated interface is
currently done using the RCU specific methods. Which means that it is
possible to miss items when iterating over it without the associated
spinlock - a behaviour which is acceptable for a periodic metric check
but not for a cleanup routine (which must "stop" all still running
workers)
The better approch is to get rid of the per interface neighbor metric
worker and handle everything in the interface worker. The original problems
are solved by:
* creating a list of neighbors which require new metric information inside
the RCU protected context, gathering the metric according to the new list
outside the RCU protected context
* only use rcu_trylock inside metric gathering code to avoid a deadlock
when the cancel_delayed_work_sync is called in the interface removal code
(which is called with the rtnl_lock held)
Cc: stable@vger.kernel.org Fixes: c833484e5f38 ("batman-adv: ELP - compute the metric based on the estimated throughput") Signed-off-by: Sven Eckelmann <sven@narfation.org> Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If a temporary error happened in the evaluation of the neighbor throughput
information, then the invalid throughput result should not be stored in the
throughtput EWMA.
Cc: stable@vger.kernel.org Signed-off-by: Sven Eckelmann <sven@narfation.org> Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reference counting is used to ensure that
batadv_hardif_neigh_node and batadv_hard_iface
are not freed before/during
batadv_v_elp_throughput_metric_update work is
finished.
But there isn't a guarantee that the hard if will
remain associated with a soft interface up until
the work is finished.
This fixes a crash triggered by reboot that looks
like this:
(the batadv_v_mesh_free call is misleading,
and does not actually happen)
I was able to make the issue happen more reliably
by changing hardif_neigh->bat_v.metric_work work
to be delayed work. This allowed me to track down
and confirm the fix.
Cc: stable@vger.kernel.org Fixes: c833484e5f38 ("batman-adv: ELP - compute the metric based on the estimated throughput") Signed-off-by: Andy Strohman <andrew@andrewstrohman.com>
[sven@narfation.org: prevent entering batadv_v_elp_get_throughput without
soft_iface] Signed-off-by: Sven Eckelmann <sven@narfation.org> Signed-off-by: Simon Wunderlich <sw@simonwunderlich.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
GCC 15 introduces a regression in "= { 0 }" style initialization of
unions that Linux has depended on for eliminating uninitialized variable
contents. GCC does not seem likely to fix it[1], instead suggesting[2]
that affected projects start using -fzero-init-padding-bits=unions.
To avoid future surprises beyond just the current situation with unions,
enable -fzero-init-padding-bits=all when available (GCC 15+). This will
correctly zero padding bits in unions and structs that might have been
left uninitialized, and will make sure there is no immediate regression
in union initializations. As seen in the stackinit KUnit selftest union
cases, which were passing before, were failing under GCC 15:
not ok 18 test_small_start_old_zero
ok 29 test_small_start_dynamic_partial # SKIP XFAIL uninit bytes: 63
ok 32 test_small_start_assigned_dynamic_partial # SKIP XFAIL uninit bytes: 63
ok 67 test_small_start_static_partial # SKIP XFAIL uninit bytes: 63
ok 70 test_small_start_static_all # SKIP XFAIL uninit bytes: 56
ok 73 test_small_start_dynamic_all # SKIP XFAIL uninit bytes: 56
ok 82 test_small_start_assigned_static_partial # SKIP XFAIL uninit bytes: 63
ok 85 test_small_start_assigned_static_all # SKIP XFAIL uninit bytes: 56
ok 88 test_small_start_assigned_dynamic_all # SKIP XFAIL uninit bytes: 56
The above all now pass again with -fzero-init-padding-bits=all added.
This also fixes the following cases for struct initialization that had
been XFAIL until now because there was no compiler support beyond the
larger "-ftrivial-auto-var-init=zero" option:
ok 38 test_small_hole_static_all # SKIP XFAIL uninit bytes: 3
ok 39 test_big_hole_static_all # SKIP XFAIL uninit bytes: 124
ok 40 test_trailing_hole_static_all # SKIP XFAIL uninit bytes: 7
ok 42 test_small_hole_dynamic_all # SKIP XFAIL uninit bytes: 3
ok 43 test_big_hole_dynamic_all # SKIP XFAIL uninit bytes: 124
ok 44 test_trailing_hole_dynamic_all # SKIP XFAIL uninit bytes: 7
ok 58 test_small_hole_assigned_static_all # SKIP XFAIL uninit bytes: 3
ok 59 test_big_hole_assigned_static_all # SKIP XFAIL uninit bytes: 124
ok 60 test_trailing_hole_assigned_static_all # SKIP XFAIL uninit bytes: 7
ok 62 test_small_hole_assigned_dynamic_all # SKIP XFAIL uninit bytes: 3
ok 63 test_big_hole_assigned_dynamic_all # SKIP XFAIL uninit bytes: 124
ok 64 test_trailing_hole_assigned_dynamic_all # SKIP XFAIL uninit bytes: 7
All of the above now pass when built under GCC 15. Tests can be seen
with:
./tools/testing/kunit/kunit.py run stackinit --arch=x86_64 \
--make_option CC=gcc-15
Clang continues to fully initialize these kinds of variables[3] without
additional flags.
The Vexia EDU ATLA 10 tablet comes in 2 different versions with
significantly different mainboards. The only outward difference is that
the charging barrel on one is marked 5V and the other is marked 9V.
The 5V version mostly works with the BYTCR defaults, except that it is
missing a CHAN package in its ACPI tables and the default of using
SSP0-AIF2 is wrong, instead SSP0-AIF1 must be used. That and its jack
detect signal is not inverted as it usually is.
Add a DMI quirk for the 5V version to fix sound not working.
I got a syzbot report: slab-out-of-bounds Read in
orangefs_debug_write... several people suggested fixes,
I tested Al Viro's suggestion and made this patch.
Signed-off-by: Mike Marshall <hubcap@omnibond.com> Reported-by: syzbot+fc519d7875f2d9186c1f@syzkaller.appspotmail.com Signed-off-by: Sasha Levin <sashal@kernel.org>
Setting and clearing CPU bits in the mm_cpumask is only ever done
by the CPU itself, from the context switch code or the TLB flush
code.
Synchronization is handled by switch_mm_irqs_off() blocking interrupts.
Sending TLB flush IPIs to CPUs that are in the mm_cpumask, but no
longer running the program causes a regression in the will-it-scale
tlbflush2 test. This test is contrived, but a large regression here
might cause a small regression in some real world workload.
Instead of always sending IPIs to CPUs that are in the mm_cpumask,
but no longer running the program, send these IPIs only once a second.
The rest of the time we can skip over CPUs where the loaded_mm is
different from the target mm.
Reported-by: kernel test roboto <oliver.sang@intel.com> Signed-off-by: Rik van Riel <riel@surriel.com> Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Link: https://lore.kernel.org/r/20241204210316.612ee573@fangorn Closes: https://lore.kernel.org/oe-lkp/202411282207.6bd28eae-lkp@intel.com/ Signed-off-by: Sasha Levin <sashal@kernel.org>
The Vexia EDU ATLA 10 tablet comes in 2 different versions with
significantly different mainboards. The only outward difference is that
the charging barrel on one is marked 5V and the other is marked 9V.
Both ship with Android 4.4 as factory OS and have the usual broken DSDT
issues for x86 Android tablets.
Add a quirk to skip ACPI I2C client enumeration for the 5V version to
complement the existing quirk for the 9V version.
Since upstream commit 8bd76b3d3f3a ("gpio: sim: lock up configfs that an
instantiated device depends on"), rmdir for an active virtual devices
been prohibited.
Update gpio-sim selftest to align with the change.
Function xen_pin_page calls xen_pte_lock, which in turn grab page
table lock (ptlock). When locking, xen_pte_lock expect mm->page_table_lock
to be held before grabbing ptlock, but this does not happen when pinning
is caused by xen_mm_pin_all.
This commit addresses lockdep warning below, which shows up when
suspending a Xen VM.
There is a HW defect on Grace Hopper (GH) to support the
Multi-Instance GPU (MIG) feature [1] that necessiated the presence
of a 1G region carved out from the device memory and mapped as
uncached. The 1G region is shown as a fake BAR (comprising region 2 and 3)
to workaround the issue.
The Grace Blackwell systems (GB) differ from GH systems in the following
aspects:
1. The aforementioned HW defect is fixed on GB systems.
2. There is a usable BAR1 (region 2 and 3) on GB systems for the
GPUdirect RDMA feature [2].
This patch accommodate those GB changes by showing the 64b physical
device BAR1 (region2 and 3) to the VM instead of the fake one. This
takes care of both the differences.
Moreover, the entire device memory is exposed on GB as cacheable to
the VM as there is no carveout required.
NVIDIA's recently introduced Grace Blackwell (GB) Superchip is a
continuation with the Grace Hopper (GH) superchip that provides a
cache coherent access to CPU and GPU to each other's memory with
an internal proprietary chip-to-chip cache coherent interconnect.
There is a HW defect on GH systems to support the Multi-Instance
GPU (MIG) feature [1] that necessiated the presence of a 1G region
with uncached mapping carved out from the device memory. The 1G
region is shown as a fake BAR (comprising region 2 and 3) to
workaround the issue. This is fixed on the GB systems.
The presence of the fix for the HW defect is communicated by the
device firmware through the DVSEC PCI config register with ID 3.
The module reads this to take a different codepath on GB vs GH.
Scan through the DVSEC registers to identify the correct one and use
it to determine the presence of the fix. Save the value in the device's
nvgrace_gpu_pci_core_device structure.
name is char[64] where the size of clnt->cl_program->name remains
unknown. Invoking strcat() directly will also lead to potential buffer
overflow. Change them to strscpy() and strncat() to fix potential
issues.
Signed-off-by: Zichen Xie <zichenxie0106@gmail.com> Reviewed-by: Benjamin Coddington <bcodding@redhat.com> Signed-off-by: Anna Schumaker <anna.schumaker@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Definitions of ioread64 and iowrite64 macros in asm/io.h called by vfio
pci implementations are enclosed inside check for CONFIG_GENERIC_IOMAP.
They don't get defined if CONFIG_GENERIC_IOMAP is defined. Include
linux/io-64-nonatomic-lo-hi.h to define iowrite64 and ioread64 macros
when they are not defined. io-64-nonatomic-lo-hi.h maps the macros to
generic implementation in lib/iomap.c. The generic implementation does
64 bit rw if readq/writeq is defined for the architecture, otherwise it
would do 32 bit back to back rw.
Note that there are two versions of the generic implementation that
differs in the order the 32 bit words are written if 64 bit support is
not present. This is not the little/big endian ordering, which is
handled separately. This patch uses the lo followed by hi word ordering
which is consistent with current back to back implementation in the
vfio/pci code.
If the <kunit/platform_device.h> header is included in a test without
certain other headers, it produces compiler warnings like:
In file included from [...]
../include/kunit/platform_device.h:15:57: warning: ‘struct completion’
declared inside parameter list will not be visible outside of this
definition or declaration
15 | struct completion *x);
| ^~~~~~~~~~
Add a 'struct completion' forward declaration to resolve this.
Reported-by: kernel test robot <lkp@intel.com> Closes: https://lore.kernel.org/oe-kbuild-all/202412241958.dbAImJsA-lkp@intel.com/ Signed-off-by: Brian Norris <briannorris@chromium.org> Reviewed-by: David Gow <davidgow@google.com> Link: https://lore.kernel.org/r/20241213180841.3023843-1-briannorris@chromium.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In the B0 revision, the RTS pin remains high due to incorrect hardware
mapping. To address this issue, enable auto-direction control with the
RTS bit in ADCL_CFG_REG. This configuration ensures that the RTS pin
goes low when the terminal is opened and high when the terminal is
closed. Additionally, we reset the step counter for Rx and Tx engines
by writing into FRAC_DIV_CFG_REG.
projects / thirdparty / kernel / stable.git / log
