The dacloffset field was originally typed as int and used in an
unchecked addition, which could overflow and bypass the existing
bounds check in both smb_check_perm_dacl() and smb_inherit_dacl().
This could result in out-of-bounds memory access and a kernel crash
when dereferencing the DACL pointer.
This patch converts dacloffset to unsigned int and uses
check_add_overflow() to validate access to the DACL.
Cc: stable@vger.kernel.org Signed-off-by: Norbert Szetei <norbert@doyensec.com> Acked-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is a race condition between session setup and
ksmbd_sessions_deregister. The session can be freed before the connection
is added to channel list of session.
This patch check reference count of session before freeing it.
Cc: stable@vger.kernel.org Reported-by: Sean Heelan <seanheelan@gmail.com> Signed-off-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In multichannel mode, UAF issue can occur in session_deregister
when the second channel sets up a session through the connection of
the first channel. session that is freed through the global session
table can be accessed again through ->sessions of connection.
If KVM rejects an AP Creation event, leave the target vCPU state as-is.
Nothing in the GHCB suggests the hypervisor is *allowed* to muck with vCPU
state on failure, let alone required to do so. Furthermore, kicking only
in the !ON_INIT case leads to divergent behavior, and even the "kick" case
is non-deterministic.
E.g. if an ON_INIT request fails, the guest can successfully retry if the
fixed AP Creation request is made prior to sending INIT. And if a !ON_INIT
fails, the guest can successfully retry if the fixed AP Creation request is
handled before the target vCPU processes KVM's
KVM_REQ_UPDATE_PROTECTED_GUEST_STATE.
Fixes: e366f92ea99e ("KVM: SEV: Support SEV-SNP AP Creation NAE event") Cc: stable@vger.kernel.org Reviewed-by: Tom Lendacky <thomas.lendacky@amd.com> Reviewed-by: Pankaj Gupta <pankaj.gupta@amd.com> Link: https://lore.kernel.org/r/20250227012541.3234589-5-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We have received reports about cards can become corrupt related to the
aggressive PM support. Let's make a partial revert of the change that
enabled the feature.
Reported-by: David Owens <daowens01@gmail.com> Reported-by: Romain Naour <romain.naour@smile.fr> Reported-by: Robert Nelson <robertcnelson@gmail.com> Tested-by: Robert Nelson <robertcnelson@gmail.com> Fixes: 3edf588e7fe0 ("mmc: sdhci-omap: Allow SDIO card power off and enable aggressive PM") Cc: stable@vger.kernel.org Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Acked-by: Adrian Hunter <adrian.hunter@intel.com> Reviewed-by: Tony Lindgren <tony@atomide.com> Link: https://lore.kernel.org/r/20250312121712.1168007-1-ulf.hansson@linaro.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Set the MMC_CAP_NEED_RSP_BUSY capability for the sdhci-pxav3 host to
prevent conversion of R1B responses to R1. Without this, the eMMC card
in the samsung,coreprimevelte smartphone using the Marvell PXA1908 SoC
with this mmc host doesn't probe with the ETIMEDOUT error originating in
__mmc_poll_for_busy.
Note that the other issues reported for this phone and host, namely
floods of "Tuning failed, falling back to fixed sampling clock" dmesg
messages for the eMMC and unstable SDIO are not mitigated by this
change.
Add err_free_host label to properly pair mmc_alloc_host() with
mmc_free_host() in GPIO error paths. The allocated host memory was
leaked when GPIO lookups failed.
Fixes: e519f0bb64ef ("ARM/mmc: Convert old mmci-omap to GPIO descriptors") Signed-off-by: Miaoqian Lin <linmq006@gmail.com> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20250318140226.19650-1-linmq006@gmail.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It's no longer practical for the OMAP IOMMU driver to trick
arm_setup_iommu_dma_ops() into ignoring its presence, so let's use the
same tactic as other IOMMU API users on 32-bit ARM and explicitly kick
the arch code's dma_iommu_mapping out of the way to avoid problems.
Fixes: 4720287c7bf7 ("iommu: Remove struct iommu_ops *iommu from arch_setup_dma_ops()") Cc: stable@vger.kernel.org Signed-off-by: Robin Murphy <robin.murphy@arm.com> Tested-by: Sicelo A. Mhlongo <absicsz@gmail.com> Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com> Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Without this, the vectors are removed if LD_DEAD_CODE_DATA_ELIMINATION
is enabled. At startup, the CPU (silently) hangs in the undefined
instruction exception as soon as the first timer interrupt arrives.
On my setup, the system also boots fine without the 2nd and 3rd KEEP()
statements, so I cannot tell whether these are actually required.
[nathan: Use OVERLAY_KEEP() to avoid breaking old ld.lld versions]
Cc: stable@vger.kernel.org Fixes: ed0f94102251 ("ARM: 9404/1: arm32: enable HAVE_LD_DEAD_CODE_DATA_ELIMINATION") Signed-off-by: Christian Eggers <ceggers@arri.de> Reviewed-by: Linus Walleij <linus.walleij@linaro.org> Signed-off-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Like the ASUS Vivobook X1504VAP and Vivobook X1704VAP, the ASUS Vivobook 14
X1404VAP has its keyboard IRQ (1) described as ActiveLow in the DSDT, which
the kernel overrides to EdgeHigh breaking the keyboard.
$ sudo dmidecode
[…]
System Information
Manufacturer: ASUSTeK COMPUTER INC.
Product Name: ASUS Vivobook 14 X1404VAP_X1404VA
[…]
$ grep -A 30 PS2K dsdt.dsl | grep IRQ -A 1
IRQ (Level, ActiveLow, Exclusive, )
{1}
Add the X1404VAP to the irq1_level_low_skip_override[] quirk table to fix
this.
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219224 Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Paul Menzel <pmenzel@molgen.mpg.de> Reviewed-by: Hans de Goede <hdegoede@redhat.com> Tested-by: Anton Shyndin <mrcold.il@gmail.com> Link: https://patch.msgid.link/20250318160903.77107-1-pmenzel@molgen.mpg.de Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Syzkaller has reported a warning in to_nfit_bus_uuid(): "only secondary
bus families can be translated". This warning is emited if the argument
is equal to NVDIMM_BUS_FAMILY_NFIT == 0. Function acpi_nfit_ctl() first
verifies that a user-provided value call_pkg->nd_family of type u64 is
not equal to 0. Then the value is converted to int, and only after that
is compared to NVDIMM_BUS_FAMILY_MAX. This can lead to passing an invalid
argument to acpi_nfit_ctl(), if call_pkg->nd_family is non-zero, while
the lower 32 bits are zero.
Furthermore, it is best to return EINVAL immediately upon seeing the
invalid user input. The WARNING is insufficient to prevent further
undefined behavior based on other invalid user input.
All checks of the input value should be applied to the original variable
call_pkg->nd_family.
The code for handling ACPI configuration in CLC was copied from the mt7921
driver but is not utilized in the mt7925 implementation. So removes the
unused functionality to clean up the codebase.
Cc: stable@vger.kernel.org Fixes: c948b5da6bbe ("wifi: mt76: mt7925: add Mediatek Wi-Fi7 driver for mt7925 chips") Signed-off-by: Ming Yen Hsieh <mingyen.hsieh@mediatek.com> Link: https://patch.msgid.link/20250304113649.867387-4-mingyen.hsieh@mediatek.com Signed-off-by: Felix Fietkau <nbd@nbd.name> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On the following path, flush_tlb_range() can be used for zapping normal
PMD entries (PMD entries that point to page tables) together with the PTE
entries in the pointed-to page table:
The arm64 version of flush_tlb_range() has a comment describing that it can
be used for page table removal, and does not use any last-level
invalidation optimizations. Fix the X86 version by making it behave the
same way.
Currently, X86 only uses this information for the following two purposes,
which I think means the issue doesn't have much impact:
- In native_flush_tlb_multi() for checking if lazy TLB CPUs need to be
IPI'd to avoid issues with speculative page table walks.
- In Hyper-V TLB paravirtualization, again for lazy TLB stuff.
The patch "x86/mm: only invalidate final translations with INVLPGB" which
is currently under review (see
<https://lore.kernel.org/all/20241230175550.4046587-13-riel@surriel.com/>)
would probably be making the impact of this a lot worse.
TSC could be reset in deep ACPI sleep states, even with invariant TSC.
That's the reason we have sched_clock() save/restore functions, to deal
with this situation. But what happens is that such functions are guarded
with a check for the stability of sched_clock - if not considered stable,
the save/restore routines aren't executed.
On top of that, we have a clear comment in native_sched_clock() saying
that *even* with TSC unstable, we continue using TSC for sched_clock due
to its speed.
In other words, if we have a situation of TSC getting detected as unstable,
it marks the sched_clock as unstable as well, so subsequent S3 sleep cycles
could bring bogus sched_clock values due to the lack of the save/restore
mechanism, causing warnings like this:
[22.954918] ------------[ cut here ]------------
[22.954923] Delta way too big! 18446743750843854390 ts=18446744072977390405 before=322133536015 after=322133536015 write stamp=18446744072977390405
[22.954923] If you just came from a suspend/resume,
[22.954923] please switch to the trace global clock:
[22.954923] echo global > /sys/kernel/tracing/trace_clock
[22.954923] or add trace_clock=global to the kernel command line
[22.954937] WARNING: CPU: 2 PID: 5728 at kernel/trace/ring_buffer.c:2890 rb_add_timestamp+0x193/0x1c0
Notice that the above was reproduced even with "trace_clock=global".
The fix for that is to _always_ save/restore the sched_clock on suspend
cycle _if TSC is used_ as sched_clock - only if we fallback to jiffies
the sched_clock_stable() check becomes relevant to save/restore the
sched_clock.
Debugged-by: Thadeu Lima de Souza Cascardo <cascardo@igalia.com> Signed-off-by: Guilherme G. Piccoli <gpiccoli@igalia.com> Signed-off-by: Ingo Molnar <mingo@kernel.org> Cc: stable@vger.kernel.org Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Link: https://lore.kernel.org/r/20250215210314.351480-1-gpiccoli@igalia.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The sizeof(struct napi_struct) can change. Don't hardcode the size to
400 bytes and instead use "sizeof(struct napi_struct)".
Suggested-by: Alexander Lobakin <aleksander.lobakin@intel.com> Signed-off-by: Joe Damato <jdamato@fastly.com> Acked-by: Alexander Lobakin <aleksander.lobakin@intel.com> Link: https://patch.msgid.link/20241004105407.73585-1-jdamato@fastly.com Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[Yifei: In Linux-6.12.y, it still hard code the size of napi_struct,
adding a member will lead the entire build failed] Signed-off-by: Yifei Liu <yifei.l.liu@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Jann reported a possible issue when trampoline_check_ip returns
address near the bottom of the address space that is allowed to
call into the syscall if uretprobes are not set up:
The WARN_ON(this_cpu_read(cpu_hw_events.enabled)) in the
intel_pmu_save_and_restart_reload() is triggered, when sampling read
topdown events.
In a NMI handler, the cpu_hw_events.enabled is set and used to indicate
the status of core PMU. The generic pmu->pmu_disable_count, updated in
the perf_pmu_disable/enable pair, is not touched.
However, the perf_pmu_disable/enable pair is invoked when sampling read
in a NMI handler. The cpuc->enabled is mistakenly set by the
perf_pmu_enable().
Avoid disabling PMU if the core PMU is already disabled.
Merge the logic together.
Fixes: 7b2c05a15d29 ("perf/x86/intel: Generic support for hardware TopDown metrics") Suggested-by: Peter Zijlstra (Intel) <peterz@infradead.org> 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/20250121152303.3128733-2-kan.liang@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The x86_pmu_drain_pebs static call was introduced in commit 7c9903c9bf71
("x86/perf, static_call: Optimize x86_pmu methods"), but it's not really
used to replace the old method.
Apply the static call for drain_pebs.
Fixes: 7c9903c9bf71 ("x86/perf, static_call: Optimize x86_pmu methods") Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> 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/20250121152303.3128733-1-kan.liang@linux.intel.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The function call “dmaengine_unmap_put(unmap)” was used in an if branch.
The same call was immediately triggered by a subsequent goto statement.
Thus avoid such a call repetition.
This issue was detected by using the Coccinelle software.
Fixes: 5648e56d03fa ("NTB: ntb_perf: Add full multi-port NTB API support") Cc: stable@vger.kernel.org Signed-off-by: Markus Elfring <elfring@users.sourceforge.net> Signed-off-by: Jon Mason <jdmason@kudzu.us> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
After resume/online turbo limit ratio (TRL) is restored partially if
the admin explicitly changed TRL from user space.
A hash table is used to store SST mail box and MSR settings when modified
to restore those settings after resume or online. This uses a struct
isst_cmd field "data" to store these settings. This is a 64 bit field.
But isst_store_new_cmd() is only assigning as u32. This results in
truncation of 32 bits.
Change the argument to u64 from u32.
Fixes: f607874f35cb ("platform/x86: ISST: Restore state on resume") Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20250328224749.2691272-1-srinivas.pandruvada@linux.intel.com Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com> Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
T495, T495s, and E560 laptops have the FANG+FANW ACPI methods
(therefore fang_handle and fanw_handle are not NULL) but they do not
actually work, which results in a "No such device or address" error.
The DSDT table code for the FANG+FANW methods doesn't seem to do
anything special regarding the fan being secondary. The bug was
introduced in commit 57d0557dfa49 ("platform/x86: thinkpad_acpi: Add
Thinkpad Edge E531 fan support"), which added a new fan control method
via the FANG+FANW ACPI methods.
Add a quirk for T495, T495s, and E560 to avoid the FANG+FANW methods.
Fan access and control is restored after forcing the legacy non-ACPI
fan control method by setting both fang_handle and fanw_handle to NULL.
Reported-by: Vlastimil Holer <vlastimil.holer@gmail.com> Fixes: 57d0557dfa49 ("platform/x86: thinkpad_acpi: Add Thinkpad Edge E531 fan support") Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219643 Cc: stable@vger.kernel.org Tested-by: Alireza Elikahi <scr0lll0ck1s4b0v3h0m3k3y@gmail.com> Reviewed-by: Kurt Borja <kuurtb@gmail.com> Signed-off-by: Eduard Christian Dumitrescu <eduard.c.dumitrescu@gmail.com> Co-developed-by: Seyediman Seyedarab <ImanDevel@gmail.com> Signed-off-by: Seyediman Seyedarab <ImanDevel@gmail.com> Link: https://lore.kernel.org/r/20250324152442.106113-1-ImanDevel@gmail.com Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com> Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Depending on the secureboot signature on EFI\BOOT\BOOTX86.EFI the
Lenovo Yoga Tab 3 UEFI will switch its OSID ACPI variable between
1 (Windows) and 4 (Android(GMIN)).
In Windows mode a GPIO event handler gets installed for GPO1 pin 5,
causing Linux' x86-android-tables code which deals with the general
brokenness of this device's ACPI tables to fail to probe with:
Direct HLT instruction execution causes #VEs for TDX VMs which is routed
to hypervisor via TDCALL. If HLT is executed in STI-shadow, resulting #VE
handler will enable interrupts before TDCALL is routed to hypervisor
leading to missed wakeup events, as current TDX spec doesn't expose
interruptibility state information to allow #VE handler to selectively
enable interrupts.
Commit bfe6ed0c6727 ("x86/tdx: Add HLT support for TDX guests")
prevented the idle routines from executing HLT instruction in STI-shadow.
But it missed the paravirt routine which can be reached via this path
as an example:
To reliably handle arch_safe_halt() for TDX VMs, introduce explicit
dependency on CONFIG_PARAVIRT and override paravirt halt()/safe_halt()
routines with TDX-safe versions that execute direct TDCALL and needed
interrupt flag updates. Executing direct TDCALL brings in additional
benefit of avoiding HLT related #VEs altogether.
As tested by Ryan Afranji:
"Tested with the specjbb2015 benchmark. It has heavy lock contention which leads
to many halt calls. TDX VMs suffered a poor score before this patchset.
Verified the major performance improvement with this patchset applied."
Fixes: bfe6ed0c6727 ("x86/tdx: Add HLT support for TDX guests") Signed-off-by: Vishal Annapurve <vannapurve@google.com> Signed-off-by: Ingo Molnar <mingo@kernel.org> Reviewed-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Tested-by: Ryan Afranji <afranji@google.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Brian Gerst <brgerst@gmail.com> Cc: Juergen Gross <jgross@suse.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20250228014416.3925664-3-vannapurve@google.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Patch series "mm/hwpoison: Fix regressions in memory failure handling",
v4.
## 1. What am I trying to do:
This patchset resolves two critical regressions related to memory failure
handling that have appeared in the upstream kernel since version 5.17, as
compared to 5.10 LTS.
- copyin case: poison found in user page while kernel copying from user space
- instr case: poison found while instruction fetching in user space
## 2. What is the expected outcome and why
- For copyin case:
Kernel can recover from poison found where kernel is doing get_user() or
copy_from_user() if those places get an error return and the kernel return
-EFAULT to the process instead of crashing. More specifily, MCE handler
checks the fixup handler type to decide whether an in kernel #MC can be
recovered. When EX_TYPE_UACCESS is found, the PC jumps to recovery code
specified in _ASM_EXTABLE_FAULT() and return a -EFAULT to user space.
- For instr case:
If a poison found while instruction fetching in user space, full recovery
is possible. User process takes #PF, Linux allocates a new page and fills
by reading from storage.
## 3. What actually happens and why
- For copyin case: kernel panic since v5.17
Commit 4c132d1d844a ("x86/futex: Remove .fixup usage") introduced a new
extable fixup type, EX_TYPE_EFAULT_REG, and later patches updated the
extable fixup type for copy-from-user operations, changing it from
EX_TYPE_UACCESS to EX_TYPE_EFAULT_REG. It breaks previous EX_TYPE_UACCESS
handling when posion found in get_user() or copy_from_user().
- For instr case: user process is killed by a SIGBUS signal due to #CMCI
and #MCE race
When an uncorrected memory error is consumed there is a race between the
CMCI from the memory controller reporting an uncorrected error with a UCNA
signature, and the core reporting and SRAR signature machine check when
the data is about to be consumed.
### Background: why *UN*corrected errors tied to *C*MCI in Intel platform [1]
Prior to Icelake memory controllers reported patrol scrub events that
detected a previously unseen uncorrected error in memory by signaling a
broadcast machine check with an SRAO (Software Recoverable Action
Optional) signature in the machine check bank. This was overkill because
it's not an urgent problem that no core is on the verge of consuming that
bad data. It's also found that multi SRAO UCE may cause nested MCE
interrupts and finally become an IERR.
Hence, Intel downgrades the machine check bank signature of patrol scrub
from SRAO to UCNA (Uncorrected, No Action required), and signal changed to
#CMCI. Just to add to the confusion, Linux does take an action (in
uc_decode_notifier()) to try to offline the page despite the UC*NA*
signature name.
### Background: why #CMCI and #MCE race when poison is consuming in
Intel platform [1]
Having decided that CMCI/UCNA is the best action for patrol scrub errors,
the memory controller uses it for reads too. But the memory controller is
executing asynchronously from the core, and can't tell the difference
between a "real" read and a speculative read. So it will do CMCI/UCNA if
an error is found in any read.
Thus:
1) Core is clever and thinks address A is needed soon, issues a
speculative read.
2) Core finds it is going to use address A soon after sending the read
request
3) The CMCI from the memory controller is in a race with MCE from the
core that will soon try to retire the load from address A.
Quite often (because speculation has got better) the CMCI from the memory
controller is delivered before the core is committed to the instruction
reading address A, so the interrupt is taken, and Linux offlines the page
(marking it as poison).
## Why user process is killed for instr case
Commit 046545a661af ("mm/hwpoison: fix error page recovered but reported
"not recovered"") tries to fix noise message "Memory error not recovered"
and skips duplicate SIGBUSs due to the race. But it also introduced a bug
that kill_accessing_process() return -EHWPOISON for instr case, as result,
kill_me_maybe() send a SIGBUS to user process.
# 4. The fix, in my opinion, should be:
- For copyin case:
The key point is whether the error context is in a read from user memory.
We do not care about the ex-type if we know its a MOV reading from
userspace.
is_copy_from_user() return true when both of the following two checks are
true:
- the current instruction is copy
- source address is user memory
Then do_machine_check() will try fixup_exception() first.
- For instr case: let kill_accessing_process() return 0 to prevent a SIGBUS.
- For patch 3:
The return value of memory_failure() is quite important while discussed
instr case regression with Tony and Miaohe for patch 2, so add comment
about the return value.
This patch (of 3):
Commit 4c132d1d844a ("x86/futex: Remove .fixup usage") introduced a new
extable fixup type, EX_TYPE_EFAULT_REG, and commit 4c132d1d844a
("x86/futex: Remove .fixup usage") updated the extable fixup type for
copy-from-user operations, changing it from EX_TYPE_UACCESS to
EX_TYPE_EFAULT_REG. The error context for copy-from-user operations no
longer functions as an in-kernel recovery context. Consequently, the
error context for copy-from-user operations no longer functions as an
in-kernel recovery context, resulting in kernel panics with the message:
"Machine check: Data load in unrecoverable area of kernel."
To address this, it is crucial to identify if an error context involves a
read operation from user memory. The function is_copy_from_user() can be
utilized to determine:
- the current operation is copy
- when reading user memory
When these conditions are met, is_copy_from_user() will return true,
confirming that it is indeed a direct copy from user memory. This check
is essential for correctly handling the context of errors in these
operations without relying on the extable fixup types that previously
allowed for in-kernel recovery.
So, use is_copy_from_user() to determine if a context is copy user directly.
Link: https://lkml.kernel.org/r/20250312112852.82415-1-xueshuai@linux.alibaba.com Link: https://lkml.kernel.org/r/20250312112852.82415-2-xueshuai@linux.alibaba.com Fixes: 4c132d1d844a ("x86/futex: Remove .fixup usage") Signed-off-by: Shuai Xue <xueshuai@linux.alibaba.com> Suggested-by: Peter Zijlstra <peterz@infradead.org> Acked-by: Borislav Petkov (AMD) <bp@alien8.de> Tested-by: Tony Luck <tony.luck@intel.com> Cc: Baolin Wang <baolin.wang@linux.alibaba.com> Cc: Borislav Betkov <bp@alien8.de> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Josh Poimboeuf <jpoimboe@kernel.org> Cc: Miaohe Lin <linmiaohe@huawei.com> Cc: Naoya Horiguchi <nao.horiguchi@gmail.com> Cc: Ruidong Tian <tianruidong@linux.alibaba.com> Cc: Thomas Gleinxer <tglx@linutronix.de> Cc: Yazen Ghannam <yazen.ghannam@amd.com> Cc: Jane Chu <jane.chu@oracle.com> Cc: Jarkko Sakkinen <jarkko@kernel.org> Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
KVM_CAP_SYNC_REGS does not make sense for VMs with protected guest state,
since the register values cannot actually be written. Return 0
when using the VM-level KVM_CHECK_EXTENSION ioctl, and accordingly
return -EINVAL from KVM_RUN if the valid/dirty fields are nonzero.
However, on exit from KVM_RUN userspace could have placed a nonzero
value into kvm_run->kvm_valid_regs, so check guest_state_protected
again and skip store_regs() in that case.
Cc: stable@vger.kernel.org Fixes: 517987e3fb19 ("KVM: x86: add fields to struct kvm_arch for CoCo features") Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20250306202923.646075-1-pbonzini@redhat.com> Reviewed-by: Pankaj Gupta <pankaj.gupta@amd.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Vincent reported that running XDP synproxy program on LoongArch results
in the following error:
JIT doesn't support bpf-to-bpf calls
With dmesg:
multi-func JIT bug 1391 != 1390
The root cause is that verifier will refill the imm with the correct
addresses of bpf_calls for BPF_PSEUDO_FUNC instructions and then run
the last pass of JIT. So we generate different JIT code for the same
instruction in two passes (one for placeholder and the other for the
real address). Let's use move_addr() instead.
See commit 64f50f6575721ef0 ("LoongArch, bpf: Use 4 instructions for
function address in JIT") for a similar fix.
Cc: stable@vger.kernel.org Fixes: 69c087ba6225 ("bpf: Add bpf_for_each_map_elem() helper") Fixes: bb035ef0cc91 ("LoongArch: BPF: Support mixing bpf2bpf and tailcalls") Reported-by: Vincent Li <vincent.mc.li@gmail.com> Tested-by: Vincent Li <vincent.mc.li@gmail.com> Closes: https://lore.kernel.org/loongarch/CAK3+h2yfM9FTNiXvEQBkvtuoJrvzmN4c_NZsFXqEk4Cj1tsBNA@mail.gmail.com/T/#u Signed-off-by: Hengqi Chen <hengqi.chen@gmail.com> Signed-off-by: Huacai Chen <chenhuacai@loongson.cn> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The verifier test `calls: div by 0 in subprog` triggers a panic at the
ld.bu instruction. The ld.bu insn is trying to load byte from memory
address returned by the subprog. The subprog actually set the correct
address at the a5 register (dedicated register for BPF return values).
But at commit 73c359d1d356 ("LoongArch: BPF: Sign-extend return values")
we also sign extended a5 to the a0 register (return value in LoongArch).
For function call insn, we later propagate the a0 register back to a5
register. This is right for native calls but wrong for bpf2bpf calls
which expect zero-extended return value in a5 register. So only move a0
to a5 for native calls (i.e. non-BPF_PSEUDO_CALL).
Vincent reported that running BPF progs with tailcalls on LoongArch
causes kernel hard lockup. Debugging the issues shows that the JITed
image missing a jirl instruction at the end of the epilogue.
There are two passes in JIT compiling, the first pass set the flags and
the second pass generates JIT code based on those flags. With BPF progs
mixing bpf2bpf and tailcalls, build_prologue() generates N insns in the
first pass and then generates N+1 insns in the second pass. This makes
epilogue_offset off by one and we will jump to some unexpected insn and
cause lockup. Fix this by inserting a nop insn.
Cc: stable@vger.kernel.org Fixes: 5dc615520c4d ("LoongArch: Add BPF JIT support") Fixes: bb035ef0cc91 ("LoongArch: BPF: Support mixing bpf2bpf and tailcalls") Reported-by: Vincent Li <vincent.mc.li@gmail.com> Tested-by: Vincent Li <vincent.mc.li@gmail.com> Closes: https://lore.kernel.org/loongarch/CAK3+h2w6WESdBN3UCr3WKHByD7D6Q_Ve1EDAjotVrnx6Or_c8g@mail.gmail.com/ Closes: https://lore.kernel.org/bpf/CAK3+h2woEjG_N=-XzqEGaAeCmgu2eTCUc7p6bP4u8Q+DFHm-7g@mail.gmail.com/ Signed-off-by: Hengqi Chen <hengqi.chen@gmail.com> Signed-off-by: Huacai Chen <chenhuacai@loongson.cn> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Begin with Loongson-3C6000, the number of PCI host can be as many as
8 for multi-chip machines, and this number should be the same for I/O
interrupt controllers. To support these machines we also increase the
MAX_IO_PICS up to 8.
ARCH_DMA_MINALIGN is 1 by default, but some LoongArch-specific devices
(such as APBDMA) require 16 bytes alignment. When the data buffer length
is too small, the hardware may make an error writing cacheline. Thus, it
is dangerous to allocate a small memory buffer for DMA. It's always safe
to define ARCH_DMA_MINALIGN as L1_CACHE_BYTES but unnecessary (kmalloc()
need small memory objects). Therefore, just increase it to 16.
This patch fixes a build issue on LoongArch when Rust is enabled and
compiled with GCC by explicitly setting the bindgen target and skipping
C flags that Clang doesn't support.
Cc: stable@vger.kernel.org Acked-by: Miguel Ojeda <ojeda@kernel.org> Signed-off-by: WANG Rui <wangrui@loongson.cn> Signed-off-by: Huacai Chen <chenhuacai@loongson.cn> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Missing usbnet_going_away Check in Critical Path.
The usb_submit_urb function lacks a usbnet_going_away
validation, whereas __usbnet_queue_skb includes this check.
This inconsistency creates a race condition where:
A URB request may succeed, but the corresponding SKB data
fails to be queued.
Subsequent processes:
(e.g., rx_complete → defer_bh → __skb_unlink(skb, list))
attempt to access skb->next, triggering a NULL pointer
dereference (Kernel Panic).
No need to overwrite the whole UARTMODIR register before waiting the
transmit engine complete, actually our target here is only to disable
CTS flow control to avoid the dirty data in TX FIFO may block the
transmit engine complete.
Also delete the following duplicate CTS disable configuration.
Fixes: d5a2e0834364 ("tty: serial: lpuart: disable flow control while waiting for the transmit engine to complete") Cc: stable <stable@kernel.org> Signed-off-by: Sherry Sun <sherry.sun@nxp.com> Link: https://lore.kernel.org/r/20250307065446.1122482-1-sherry.sun@nxp.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Most lpuart functions have the parameter struct uart_port *port, but
still use the &sport->port to get the uart_port instead of use it
directly, let's simply the code logic, directly use this struct instead
of covert it from struct sport.
The commit b824766504e4 ("cgroup/rstat: add force idle show helper")
retrieves forceidle_time outside cgroup_rstat_lock for non-root cgroups
which can be potentially inconsistent with other stats.
Rather than reverting that commit, fix it in a way that retains the
effort of cleaning up the ifdef-messes.
Fixes: b824766504e4 ("cgroup/rstat: add force idle show helper") Signed-off-by: Abel Wu <wuyun.abel@bytedance.com> Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Cgroup-level CPU statistics currently include time spent on
user/system processes, but do not include niced CPU time (despite
already being tracked). This patch exposes niced CPU time to the
userspace, allowing users to get a better understanding of their
hardware limits and can facilitate more informed workload distribution.
A new field 'ntime' is added to struct cgroup_base_stat as opposed to
struct task_cputime to minimize footprint.
Signed-off-by: Joshua Hahn <joshua.hahnjy@gmail.com> Signed-off-by: Tejun Heo <tj@kernel.org>
Stable-dep-of: c4af66a95aa3 ("cgroup/rstat: Fix forceidle time in cpu.stat") Signed-off-by: Sasha Levin <sashal@kernel.org>
The function event_{hist,hist_debug}_open() maintains the refcount of
'file->tr' and 'file' through tracing_open_file_tr(). However, it does
not roll back these counts on subsequent failure paths, resulting in a
refcount leak.
A very obvious case is that if the hist/hist_debug file belongs to a
specific instance, the refcount leak will prevent the deletion of that
instance, as it relies on the condition 'tr->ref == 1' within
__remove_instance().
Fix this by calling tracing_release_file_tr() on all failure paths in
event_{hist,hist_debug}_open() to correct the refcount.
Cc: stable@vger.kernel.org Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Cc: Zheng Yejian <zhengyejian1@huawei.com> Link: https://lore.kernel.org/20250314065335.1202817-1-wutengda@huaweicloud.com Fixes: 1cc111b9cddc ("tracing: Fix uaf issue when open the hist or hist_debug file") Signed-off-by: Tengda Wu <wutengda@huaweicloud.com> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Since POLLIN will not be flushed until the hist file is read, the user
needs to repeatedly read() and poll() on the hist file for monitoring the
event continuously. But the read() is somewhat redundant when the user is
only monitoring for event updates.
Add POLLPRI poll event on the hist file so the event returns when a
histogram is updated after open(), poll() or read(). Thus it is possible
to wait for the next event without having to issue a read().
Cc: Shuah Khan <shuah@kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Link: https://lore.kernel.org/173527248770.464571.2536902137325258133.stgit@devnote2 Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org> Reviewed-by: Tom Zanussi <zanussi@kernel.org> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Stable-dep-of: 0b4ffbe4888a ("tracing: Correct the refcount if the hist/hist_debug file fails to open") Signed-off-by: Sasha Levin <sashal@kernel.org>
Add poll syscall support on the `hist` file. The Waiter will be waken
up when the histogram is updated with POLLIN.
Currently, there is no way to wait for a specific event in userspace.
So user needs to peek the `trace` periodicaly, or wait on `trace_pipe`.
But it is not a good idea to peek at the `trace` for an event that
randomly happens. And `trace_pipe` is not coming back until a page is
filled with events.
This allows a user to wait for a specific event on the `hist` file. User
can set a histogram trigger on the event which they want to monitor
and poll() on its `hist` file. Since this poll() returns POLLIN, the next
poll() will return soon unless a read() happens on that hist file.
NOTE: To read the hist file again, you must set the file offset to 0,
but just for monitoring the event, you may not need to read the
histogram.
Cc: Shuah Khan <shuah@kernel.org> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Link: https://lore.kernel.org/173527247756.464571.14236296701625509931.stgit@devnote2 Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org> Reviewed-by: Tom Zanussi <zanussi@kernel.org> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Stable-dep-of: 0b4ffbe4888a ("tracing: Correct the refcount if the hist/hist_debug file fails to open") Signed-off-by: Sasha Levin <sashal@kernel.org>
There are a couple functions in trace_events_hist.c that have "goto out" or
equivalent on error in order to release locks that were taken. This can be
error prone or just simply make the code more complex.
Switch every location that ends with unlocking a mutex on error over to
using the guard(mutex)() infrastructure to let the compiler worry about
releasing locks. This makes the code easier to read and understand.
Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: https://lore.kernel.org/20241219201345.694601480@goodmis.org Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Stable-dep-of: 0b4ffbe4888a ("tracing: Correct the refcount if the hist/hist_debug file fails to open") Signed-off-by: Sasha Levin <sashal@kernel.org>
The CoreThr column displays total thermal throttling events
since boot time.
Change it to report events during the measurement interval.
This is more useful for showing a user the current conditions.
Total events since boot time are still available to the user via
/sys/devices/system/cpu/cpu*/thermal_throttle/*
Document CoreThr on turbostat.8
Fixes: eae97e053fe30 ("turbostat: Support thermal throttle count print") Reported-by: Arjan van de Ven <arjan@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com> Cc: Chen Yu <yu.c.chen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The two events above are opened on two different CPU PMUs, for example,
each event is opened for a cluster in an Arm big.LITTLE system, they
will never run on the same CPU. In theory, the total enabled time should
be same for both events, as two events are opened and closed together.
As the result show, the two events' total enabled time including
child event is different (2289429840 vs 1950025700).
This is because child events are not accounted properly
if a event is INACTIVE state when the task exits:
The problem is the time accumulation happens prior to child event's
time updating. Thus, it misses to account the last period's time when
the event exits.
The perf core layer follows the rule that timekeeping is tied to state
change. To address the issue, make __perf_remove_from_context()
handle the task exit case by passing 'DETACH_EXIT' to it and
invoke perf_event_state() for state alongside with accounting the time.
Then, perf_child_detach() populates the time into the parent's time metrics.
In ThinPro, we use the convention <upstream_ver>+hp<patchlevel> for
the kernel package. This does not have a dash in the name or version.
This is built by editing ".version" before a build, and setting
EXTRAVERSION="+hp" and KDEB_PKGVERSION make variables:
echo 68 > .version
make -j<n> EXTRAVERSION="+hp" bindeb-pkg KDEB_PKGVERSION=6.12.2+hp69
Since commit 7d4f07d5cb71 ("kbuild: deb-pkg: squash
scripts/package/deb-build-option to debian/rules"), this no longer
works. The deb build logic changed, even though, the commit message
implies that the logic should be unmodified.
Before, KBUILD_BUILD_VERSION was not set if the KDEB_PKGVERSION did
not contain a dash. After the change KBUILD_BUILD_VERSION is always
set to KDEB_PKGVERSION. Since this determines UTS_VERSION, the uname
output to look off:
(now) uname -a: version 6.12.2+hp ... #6.12.2+hp69
(expected) uname -a: version 6.12.2+hp ... #69
Update the debian/rules logic to restore the original behavior.
Fixes: 7d4f07d5cb71 ("kbuild: deb-pkg: squash scripts/package/deb-build-option to debian/rules") Signed-off-by: Alexandru Gagniuc <alexandru.gagniuc@hp.com> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
v2:
- Created a single error handling unlock and exit in veth_pool_store
- Greatly expanded commit message with previous explanatory-only text
Summary: Use rtnl_mutex to synchronize veth_pool_store with itself,
ibmveth_close and ibmveth_open, preventing multiple calls in a row to
napi_disable.
Background: Two (or more) threads could call veth_pool_store through
writing to /sys/devices/vio/30000002/pool*/*. You can do this easily
with a little shell script. This causes a hang.
I configured LOCKDEP, compiled ibmveth.c with DEBUG, and built a new
kernel. I ran this test again and saw:
The control APIs are not idempotent. Control API calls are safe
against concurrent use of datapath APIs but an incorrect sequence of
control API calls may result in crashes, deadlocks, or race
conditions. For example, calling napi_disable() multiple times in a
row will deadlock.
In the normal open and close paths, rtnl_mutex is acquired to prevent
other callers. This is missing from veth_pool_store. Use rtnl_mutex in
veth_pool_store fixes these hangs.
Signed-off-by: Dave Marquardt <davemarq@linux.ibm.com> Fixes: 860f242eb534 ("[PATCH] ibmveth change buffer pools dynamically") Reviewed-by: Nick Child <nnac123@linux.ibm.com> Reviewed-by: Simon Horman <horms@kernel.org> Link: https://patch.msgid.link/20250402154403.386744-1-davemarq@linux.ibm.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
devm_kasprintf() returns NULL when memory allocation fails. Currently,
com20020pci_probe() does not check for this case, which results in a
NULL pointer dereference.
Add NULL check after devm_kasprintf() to prevent this issue and ensure
no resources are left allocated.
Fixes: 6b17a597fc2f ("arcnet: restoring support for multiple Sohard Arcnet cards") Signed-off-by: Henry Martin <bsdhenrymartin@gmail.com> Link: https://patch.msgid.link/20250402135036.44697-1-bsdhenrymartin@gmail.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Nexthops whose link is down are not supposed to be considered during
path selection when the "ignore_routes_with_linkdown" sysctl is set.
This is done by assigning them a negative region boundary.
However, when comparing the computed hash (unsigned) with the region
boundary (signed), the negative region boundary is treated as unsigned,
resulting in incorrect nexthop selection.
Fix by treating the computed hash as signed. Note that the computed hash
is always in range of [0, 2^31 - 1].
Fixes: 3d709f69a3e7 ("ipv6: Use hash-threshold instead of modulo-N") Signed-off-by: Ido Schimmel <idosch@nvidia.com> Reviewed-by: Willem de Bruijn <willemb@google.com> Link: https://patch.msgid.link/20250402114224.293392-3-idosch@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Cited commit transitioned IPv6 path selection to use hash-threshold
instead of modulo-N. With hash-threshold, each nexthop is assigned a
region boundary in the multipath hash function's output space and a
nexthop is chosen if the calculated hash is smaller than the nexthop's
region boundary.
Hash-threshold does not work correctly if path selection does not start
with the first nexthop. For example, if fib6_select_path() is always
passed the last nexthop in the group, then it will always be chosen
because its region boundary covers the entire hash function's output
space.
Fix this by starting the selection process from the first nexthop and do
not consider nexthops for which rt6_score_route() provided a negative
score.
Fixes: 3d709f69a3e7 ("ipv6: Use hash-threshold instead of modulo-N") Reported-by: Stanislav Fomichev <stfomichev@gmail.com> Closes: https://lore.kernel.org/netdev/Z9RIyKZDNoka53EO@mini-arch/ Signed-off-by: Ido Schimmel <idosch@nvidia.com> Link: https://patch.msgid.link/20250402114224.293392-2-idosch@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
struct geneve_opt uses 5 bit length for each single option, which
means every vary size option should be smaller than 128 bytes.
However, all current related Netlink policies cannot promise this
length condition and the attacker can exploit a exact 128-byte size
option to *fake* a zero length option and confuse the parsing logic,
further achieve heap out-of-bounds read.
Fix these issues by enforing correct length condition in related
policies.
Fixes: 925d844696d9 ("netfilter: nft_tunnel: add support for geneve opts") Fixes: 4ece47787077 ("lwtunnel: add options setting and dumping for geneve") Fixes: 0ed5269f9e41 ("net/sched: add tunnel option support to act_tunnel_key") Fixes: 0a6e77784f49 ("net/sched: allow flower to match tunnel options") Signed-off-by: Lin Ma <linma@zju.edu.cn> Reviewed-by: Xin Long <lucien.xin@gmail.com> Acked-by: Cong Wang <xiyou.wangcong@gmail.com> Link: https://patch.msgid.link/20250402165632.6958-1-linma@zju.edu.cn Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The mv88e6xxx has an internal PPU that polls PHY state. If we want to
access the internal PHYs, we need to disable the PPU first. Because
that is a slow operation, a 10ms timer is used to re-enable it,
canceled with every access, so bulk operations effectively only
disable it once and re-enable it some 10ms after the last access.
If a PHY is accessed and then the mv88e6xxx module is removed before
the 10ms are up, the PPU re-enable ends up accessing a dangling pointer.
This especially affects probing during bootup. The MDIO bus and PHY
registration may succeed, but registration with the DSA framework
may fail later on (e.g. because the CPU port depends on another,
very slow device that isn't done probing yet, returning -EPROBE_DEFER).
In this case, probe() fails, but the MDIO subsystem may already have
accessed the MIDO bus or PHYs, arming the timer.
This is fixed as follows:
- If probe fails after mv88e6xxx_phy_init(), make sure we also call
mv88e6xxx_phy_destroy() before returning
- In mv88e6xxx_remove(), make sure we do the teardown in the correct
order, calling mv88e6xxx_phy_destroy() after unregistering the
switch device.
- In mv88e6xxx_phy_destroy(), destroy both the timer and the work item
that the timer might schedule, synchronously waiting in case one of
the callbacks already fired and destroying the timer first, before
waiting for the work item.
- Access to the PPU is guarded by a mutex, the worker acquires it
with a mutex_trylock(), not proceeding with the expensive shutdown
if that fails. We grab the mutex in mv88e6xxx_phy_destroy() to make
sure the slow PPU shutdown is already done or won't even enter, when
we wait for the work item.
Fixes: 2e5f032095ff ("dsa: add support for the Marvell 88E6131 switch chip") Signed-off-by: David Oberhollenzer <david.oberhollenzer@sigma-star.at> Reviewed-by: Vladimir Oltean <olteanv@gmail.com> Link: https://patch.msgid.link/20250401135705.92760-1-david.oberhollenzer@sigma-star.at Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Using RTEXT_FILTER_SKIP_STATS is incorrectly skipping non-stats IPv6
netlink attributes on link dump. This causes issues on userspace tools,
e.g iproute2 is not rendering address generation mode as it should due
to missing netlink attribute.
Move the filling of IFLA_INET6_STATS and IFLA_INET6_ICMP6STATS to a
helper function guarded by a flag check to avoid hitting the same
situation in the future.
Fixes: d5566fd72ec1 ("rtnetlink: RTEXT_FILTER_SKIP_STATS support to avoid dumping inet/inet6 stats") Signed-off-by: Fernando Fernandez Mancera <ffmancera@riseup.net> Reviewed-by: Simon Horman <horms@kernel.org> Link: https://patch.msgid.link/20250402121751.3108-1-ffmancera@riseup.net Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When handling multiple NFTA_TUNNEL_KEY_OPTS_GENEVE attributes, the
parsing logic should place every geneve_opt structure one by one
compactly. Hence, when deciding the next geneve_opt position, the
pointer addition should be in units of char *.
However, the current implementation erroneously does type conversion
before the addition, which will lead to heap out-of-bounds write.
Fix this bug with correct pointer addition and conversion in parse
and dump code.
Fixes: 925d844696d9 ("netfilter: nft_tunnel: add support for geneve opts") Signed-off-by: Lin Ma <linma@zju.edu.cn> Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Upstream fix ac888d58869b ("net: do not delay dst_entries_add() in
dst_release()") moved decrementing the dst count from dst_destroy to
dst_release to avoid accessing already freed data in case of netns
dismantle. However in case CONFIG_DST_CACHE is enabled and OvS+tunnels
are used, this fix is incomplete as the same issue will be seen for
cached dsts:
Because skb_tunnel_check_pmtu() doesn't handle PACKET_HOST packets,
commit 30a92c9e3d6b ("openvswitch: Set the skbuff pkt_type for proper
pmtud support.") forced skb->pkt_type to PACKET_OUTGOING for
openvswitch packets that are sent using the OVS_ACTION_ATTR_OUTPUT
action. This allowed such packets to invoke the
iptunnel_pmtud_check_icmp() or iptunnel_pmtud_check_icmpv6() helpers
and thus trigger PMTU update on the input device.
However, this also broke other parts of PMTU discovery. Since these
packets don't have the PACKET_HOST type anymore, they won't trigger the
sending of ICMP Fragmentation Needed or Packet Too Big messages to
remote hosts when oversized (see the skb_in->pkt_type condition in
__icmp_send() for example).
These two skb->pkt_type checks are therefore incompatible as one
requires skb->pkt_type to be PACKET_HOST, while the other requires it
to be anything but PACKET_HOST.
It makes sense to not trigger ICMP messages for non-PACKET_HOST packets
as these messages should be generated only for incoming l2-unicast
packets. However there doesn't seem to be any reason for
skb_tunnel_check_pmtu() to ignore PACKET_HOST packets.
Allow both cases to work by allowing skb_tunnel_check_pmtu() to work on
PACKET_HOST packets and not overriding skb->pkt_type in openvswitch
anymore.
Fixes: 30a92c9e3d6b ("openvswitch: Set the skbuff pkt_type for proper pmtud support.") Fixes: 4cb47a8644cc ("tunnels: PMTU discovery support for directly bridged IP packets") Signed-off-by: Guillaume Nault <gnault@redhat.com> Reviewed-by: Stefano Brivio <sbrivio@redhat.com> Reviewed-by: Aaron Conole <aconole@redhat.com> Tested-by: Aaron Conole <aconole@redhat.com> Link: https://patch.msgid.link/eac941652b86fddf8909df9b3bf0d97bc9444793.1743208264.git.gnault@redhat.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When a peer attempts to establish a connection, vsock_connect() contains
a loop that waits for the state to be TCP_ESTABLISHED. However, the
other peer can be fast enough to accept the connection and close it
immediately, thus moving the state to TCP_CLOSING.
When this happens, the peer in the vsock_connect() is properly woken up,
but since the state is not TCP_ESTABLISHED, it goes back to sleep
until the timeout expires, returning -ETIMEDOUT.
If the socket state is TCP_CLOSING, waiting for the timeout is pointless.
vsock_connect() can return immediately without errors or delay since the
connection actually happened. The socket will be in a closing state,
but this is not an issue, and subsequent calls will fail as expected.
We discovered this issue while developing a test that accepts and
immediately closes connections to stress the transport switch between
two connect() calls, where the first one was interrupted by a signal
(see Closes link).
Reported-by: Luigi Leonardi <leonardi@redhat.com> Closes: https://lore.kernel.org/virtualization/bq6hxrolno2vmtqwcvb5bljfpb7mvwb3kohrvaed6auz5vxrfv@ijmd2f3grobn/ Fixes: d021c344051a ("VSOCK: Introduce VM Sockets") Signed-off-by: Stefano Garzarella <sgarzare@redhat.com> Acked-by: Paolo Abeni <pabeni@redhat.com> Tested-by: Luigi Leonardi <leonardi@redhat.com> Reviewed-by: Luigi Leonardi <leonardi@redhat.com> Link: https://patch.msgid.link/20250328141528.420719-1-sgarzare@redhat.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Matt Dowling reported a weird UDP memory usage issue.
Under normal operation, the UDP memory usage reported in /proc/net/sockstat
remains close to zero. However, it occasionally spiked to 524,288 pages
and never dropped. Moreover, the value doubled when the application was
terminated. Finally, it caused intermittent packet drops.
We can reproduce the issue with the script below [0]:
The application set INT_MAX to SO_RCVBUF, which triggered an integer
overflow in udp_rmem_release().
When a socket is close()d, udp_destruct_common() purges its receive
queue and sums up skb->truesize in the queue. This total is calculated
and stored in a local unsigned integer variable.
The total size is then passed to udp_rmem_release() to adjust memory
accounting. However, because the function takes a signed integer
argument, the total size can wrap around, causing an overflow.
Then, the released amount is calculated as follows:
1) Add size to sk->sk_forward_alloc.
2) Round down sk->sk_forward_alloc to the nearest lower multiple of
PAGE_SIZE and assign it to amount.
3) Subtract amount from sk->sk_forward_alloc.
4) Pass amount >> PAGE_SHIFT to __sk_mem_reduce_allocated().
When the issue occurred, the total in udp_destruct_common() was 2147484480
(INT_MAX + 833), which was cast to -2147482816 in udp_rmem_release().
At 1) sk->sk_forward_alloc is changed from 3264 to -2147479552, and
2) sets -2147479552 to amount. 3) reverts the wraparound, so we don't
see a warning in inet_sock_destruct(). However, udp_memory_allocated
ends up doubling at 4).
Since commit 3cd3399dd7a8 ("net: implement per-cpu reserves for
memory_allocated"), memory usage no longer doubles immediately after
a socket is close()d because __sk_mem_reduce_allocated() caches the
amount in udp_memory_per_cpu_fw_alloc. However, the next time a UDP
socket receives a packet, the subtraction takes effect, causing UDP
memory usage to double.
This issue makes further memory allocation fail once the socket's
sk->sk_rmem_alloc exceeds net.ipv4.udp_rmem_min, resulting in packet
drops.
To prevent this issue, let's use unsigned int for the calculation and
call sk_forward_alloc_add() only once for the small delta.
Note that first_packet_length() also potentially has the same problem.
[0]:
from socket import *
SO_RCVBUFFORCE = 33
INT_MAX = (2 ** 31) - 1
s = socket(AF_INET, SOCK_DGRAM)
s.bind(('', 0))
s.setsockopt(SOL_SOCKET, SO_RCVBUFFORCE, INT_MAX)
c = socket(AF_INET, SOCK_DGRAM)
c.connect(s.getsockname())
data = b'a' * 100
while True:
c.send(data)
Fixes: f970bd9e3a06 ("udp: implement memory accounting helpers") Reported-by: Matt Dowling <madowlin@amazon.com> Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com> Reviewed-by: Willem de Bruijn <willemb@google.com> Link: https://patch.msgid.link/20250401184501.67377-3-kuniyu@amazon.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
__udp_enqueue_schedule_skb() has the following condition:
if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf)
goto drop;
sk->sk_rcvbuf is initialised by net.core.rmem_default and later can
be configured by SO_RCVBUF, which is limited by net.core.rmem_max,
or SO_RCVBUFFORCE.
If we set INT_MAX to sk->sk_rcvbuf, the condition is always false
as sk->sk_rmem_alloc is also signed int.
Then, the size of the incoming skb is added to sk->sk_rmem_alloc
unconditionally.
This results in integer overflow (possibly multiple times) on
sk->sk_rmem_alloc and allows a single socket to have skb up to
net.core.udp_mem[1].
For example, if we set a large value to udp_mem[1] and INT_MAX to
sk->sk_rcvbuf and flood packets to the socket, we can see multiple
overflows:
# cat /proc/net/sockstat | grep UDP:
UDP: inuse 3 mem 7956736 <-- (7956736 << 12) bytes > INT_MAX * 15
^- PAGE_SHIFT
# ss -uam
State Recv-Q ...
UNCONN -1757018048 ... <-- flipping the sign repeatedly
skmem:(r2537949248,rb2147483646,t0,tb212992,f1984,w0,o0,bl0,d0)
Previously, we had a boundary check for INT_MAX, which was removed by
commit 6a1f12dd85a8 ("udp: relax atomic operation on sk->sk_rmem_alloc").
A complete fix would be to revert it and cap the right operand by
INT_MAX:
# ss -uam
State Recv-Q ...
UNCONN -2147482816 ... <-- INT_MAX + 831 bytes
skmem:(r2147484480,rb2147483646,t0,tb212992,f3264,w0,o0,bl0,d14468947)
So, let's define rmem and rcvbuf as unsigned int and check skb->truesize
only when rcvbuf is large enough to lower the overflow possibility.
Note that we still have a small chance to see overflow if multiple skbs
to the same socket are processed on different core at the same time and
each size does not exceed the limit but the total size does.
Note also that we must ignore skb->truesize for a small buffer as
explained in commit 363dc73acacb ("udp: be less conservative with
sock rmem accounting").
Fixes: 6a1f12dd85a8 ("udp: relax atomic operation on sk->sk_rmem_alloc") Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com> Reviewed-by: Willem de Bruijn <willemb@google.com> Link: https://patch.msgid.link/20250401184501.67377-2-kuniyu@amazon.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Protect the parser TCAM/SRAM memory, and the cached (shadow) SRAM
information, from concurrent modifications.
Both the TCAM and SRAM tables are indirectly accessed by configuring
an index register that selects the row to read or write to. This means
that operations must be atomic in order to, e.g., avoid spreading
writes across multiple rows. Since the shadow SRAM array is used to
find free rows in the hardware table, it must also be protected in
order to avoid TOCTOU errors where multiple cores allocate the same
row.
This issue was detected in a situation where `mvpp2_set_rx_mode()` ran
concurrently on two CPUs. In this particular case the
MVPP2_PE_MAC_UC_PROMISCUOUS entry was corrupted, causing the
classifier unit to drop all incoming unicast - indicated by the
`rx_classifier_drops` counter.
Fixes: 3f518509dedc ("ethernet: Add new driver for Marvell Armada 375 network unit") Signed-off-by: Tobias Waldekranz <tobias@waldekranz.com> Reviewed-by: Maxime Chevallier <maxime.chevallier@bootlin.com> Tested-by: Maxime Chevallier <maxime.chevallier@bootlin.com> Link: https://patch.msgid.link/20250401065855.3113635-1-tobias@waldekranz.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In the current implementation, skbprio enqueue/dequeue contains an assertion
that fails under certain conditions when SKBPRIO is used as a child qdisc under
TBF with specific parameters. The failure occurs because TBF sometimes peeks at
packets in the child qdisc without actually dequeuing them when tokens are
unavailable.
This peek operation creates a discrepancy between the parent and child qdisc
queue length counters. When TBF later receives a high-priority packet,
SKBPRIO's queue length may show a different value than what's reflected in its
internal priority queue tracking, triggering the assertion.
The fix removes this overly strict assertions in SKBPRIO, they are not
necessary at all.
When calling netlbl_conn_setattr(), addr->sa_family is used
to determine the function behavior. If sk is an IPv4 socket,
but the connect function is called with an IPv6 address,
the function calipso_sock_setattr() is triggered.
Inside this function, the following code is executed:
sk_fullsock(__sk) ? inet_sk(__sk)->pinet6 : NULL;
Since sk is an IPv4 socket, pinet6 is NULL, leading to a
null pointer dereference.
This patch fixes the issue by checking if inet6_sk(sk)
returns a NULL pointer before accessing pinet6.
Signed-off-by: Debin Zhu <mowenroot@163.com> Signed-off-by: Bitao Ouyang <1985755126@qq.com> Acked-by: Paul Moore <paul@paul-moore.com> Fixes: ceba1832b1b2 ("calipso: Set the calipso socket label to match the secattr.") Link: https://patch.msgid.link/20250401124018.4763-1-mowenroot@163.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
conncount has its own GC handler which determines when to reap stale
elements, this is convenient for dynamic sets. However, this also reaps
non-dynamic sets with static configurations coming from control plane.
Always run connlimit gc handler but honor feedback to reap element if
this set is dynamic.
With SRIOV enabled, idpf ends up calling into idpf_remove() twice.
First via idpf_shutdown() and then again when idpf_remove() calls into
sriov_disable(), because the VF devices use the idpf driver, hence the
same remove routine. When that happens, it is possible for the adapter
to be NULL from the first call to idpf_remove(), leading to a NULL
pointer dereference.
Replace the direct idpf_remove() call in idpf_shutdown() with
idpf_vc_core_deinit() and idpf_deinit_dflt_mbx(), which perform
the bulk of the cleanup, such as stopping the init task, freeing IRQs,
destroying the vports and freeing the mailbox. This avoids the calls to
sriov_disable() in addition to a small netdev cleanup, and destroying
workqueues, which don't seem to be required on shutdown.
Reported-by: Yuying Ma <yuma@redhat.com> Fixes: e850efed5e15 ("idpf: add module register and probe functionality") Reviewed-by: Madhu Chittim <madhu.chittim@intel.com> Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com> Reviewed-by: Simon Horman <horms@kernel.org> Tested-by: Samuel Salin <Samuel.salin@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Starting from Meteor Lake, the Kumeran interface between the integrated
MAC and the I219 PHY works at a different frequency. This causes sporadic
MDI errors when accessing the PHY, and in rare circumstances could lead
to packet corruption.
To overcome this, introduce minor changes to the Kumeran idle
state (K1) parameters during device initialization. Hardware reset
reverts this configuration, therefore it needs to be applied in a few
places.
Fixes: cc23f4f0b6b9 ("e1000e: Add support for Meteor Lake") Signed-off-by: Vitaly Lifshits <vitaly.lifshits@intel.com> Tested-by: Avigail Dahan <avigailx.dahan@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The lookup table forces the use of the "pinctrl-bcm2835" GPIO chip
provider and essentially assumes that there is going to be such a
provider, and if not, we will fail to set-up the SPI device.
While this is true on Raspberry Pi based systems (2835/36/37, 2711,
2712), this is not true on 7712/77122 Broadcom STB systems which use the
SPI driver, but not the GPIO driver.
There used to be an early check:
chip = gpiochip_find("pinctrl-bcm2835", chip_match_name);
if (!chip)
return 0;
which would accomplish that nicely, bring something similar back by
checking for the compatible strings matched by the pinctrl-bcm2835.c
driver, if there is no Device Tree node matching those compatible
strings, then we won't find any GPIO provider registered by the
"pinctrl-bcm2835" driver.
Fixes: 21f252cd29f0 ("spi: bcm2835: reduce the abuse of the GPIO API") Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com> Link: https://patch.msgid.link/20250401233603.2938955-1-florian.fainelli@broadcom.com Acked-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
There is another VivoBook model which built-in mic got broken recently
by the fix of the pin sort. Apply the correct quirk
ALC256_FIXUP_ASUS_MIC_NO_PRESENCE to this model for addressing the
regression, too.
If we are unable to lookup the chip-select GPIO, the error path will
call bcm2835_spi_cleanup() which unconditionally calls gpiod_put() on
the cs->gpio variable which we just determined was invalid.
devm_kasprintf() returns NULL when memory allocation fails. Currently,
imx_card_probe() does not check for this case, which results in a NULL
pointer dereference.
Add NULL check after devm_kasprintf() to prevent this issue.
Fixes: aa736700f42f ("ASoC: imx-card: Add imx-card machine driver") Signed-off-by: Henry Martin <bsdhenrymartin@gmail.com> Reviewed-by: Frank Li <Frank.Li@nxp.com> Link: https://patch.msgid.link/20250401142510.29900-1-bsdhenrymartin@gmail.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The vectorized io_uring NVMe passthru opcodes don't yet support fixed
buffers. But since userspace can trigger this condition based on the
io_uring SQE parameters, it shouldn't cause a kernel warning.
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com> Reviewed-by: Jens Axboe <axboe@kernel.dk> Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Fixes: 23fd22e55b76 ("nvme: wire up fixed buffer support for nvme passthrough") Signed-off-by: Keith Busch <kbusch@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When a crashkernel is launched on RISC-V, the entry to purgatory is
done by trapping via the stvec CSR. From riscv_kexec_norelocate():
| ...
| /*
| * Switch to physical addressing
| * This will also trigger a jump to CSR_STVEC
| * which in this case is the address of the new
| * kernel.
| */
| csrw CSR_STVEC, a2
| csrw CSR_SATP, zero
stvec requires that the address is 4B aligned, which was not the case,
e.g.:
Commit 58ff537109ac ("riscv: Omit optimized string routines when
using KASAN") introduced calls to EXPORT_SYMBOL() in assembly string
routines, which result in R_RISCV_64 relocations against
.export_symbol section. As these rountines are reused by RISC-V
purgatory and our relocator doesn't recognize these relocations, this
fails kexec-file-load with dmesg like
Ryan sent a fix [1] for arm64 that applies to riscv too: in some hugetlb
functions, we must not use the pte value to get the size of a mapping
because the pte may not be present.
So use the already present size parameter for huge_pte_clear() and the
newly introduced size parameter for huge_ptep_get_and_clear(). And make
sure to gather A/D bits only on present ptes.
If requested_clk > 128, cdns_mrvl_xspi_setup_clock() iterates over the
entire cdns_mrvl_xspi_clk_div_list array without breaking out early,
causing 'i' to go beyond the array bounds.
Fix that by stopping the loop when it gets to the last entry, clamping
the clock to the minimum 6.25 MHz.
Fixes the following warning with an UBSAN kernel:
vmlinux.o: warning: objtool: cdns_mrvl_xspi_setup_clock: unexpected end of section .text.cdns_mrvl_xspi_setup_clock
Should an error occur after a successful regulator_bulk_enable() call,
regulator_bulk_disable() should be called, as already done in the remove
function.
Instead of adding an error handling path in the probe, switch from
devm_regulator_bulk_get() to devm_regulator_bulk_get_enable() and
simplify the remove function and some other places accordingly.
Finally, add a missing const when defining rt5665_supply_names to please
checkpatch and constify a few bytes.
When lowcore relocation is enabled, the machine check handler doesn't
use the lowcore address when setting _CIF_MCCK_GUEST. Fix this by
adding the missing base register.
Fixes: 0001b7bbc53a ("s390/entry: Make mchk_int_handler() ready for lowcore relocation") Reported-by: Heiko Carstens <hca@linux.ibm.com> Reviewed-by: Heiko Carstens <hca@linux.ibm.com> Signed-off-by: Sven Schnelle <svens@linux.ibm.com> Signed-off-by: Vasily Gorbik <gor@linux.ibm.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Now ublk driver depends on `ubq->canceling` for deciding if the request
can be dispatched via uring_cmd & io_uring_cmd_complete_in_task().
Once ubq->canceling is set, the uring_cmd can be done via ublk_cancel_cmd()
and io_uring_cmd_done().
So set ubq->canceling when queue is frozen, this way makes sure that the
flag can be observed from ublk_queue_rq() reliably, and avoids
use-after-free on uring_cmd.
Fixes: 216c8f5ef0f2 ("ublk: replace monitor with cancelable uring_cmd") Signed-off-by: Ming Lei <ming.lei@redhat.com> Link: https://lore.kernel.org/r/20250327095123.179113-2-ming.lei@redhat.com Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Sasha Levin <sashal@kernel.org>
History of the performance regression:
======================================
Since the following series of user copy updates were merged upstream
~2 years ago via:
a5624566431d ("Merge branch 'x86-rep-insns': x86 user copy clarifications")
.. copy_user_generic() on x86_64 stopped doing alignment of the
writes to the destination to a 8 byte boundary for the non FSRM case.
Previously, this was done through the ALIGN_DESTINATION macro that
was used in the now removed copy_user_generic_unrolled function.
Turns out this change causes some loss of performance/throughput on
some use cases and specific CPU/platforms without FSRM and ERMS.
Lately I got two reports of performance/throughput issues after a
RHEL 9 kernel pulled the same upstream series with updates to user
copy functions. Both reports consisted of running specific
networking/TCP related testing using iperf3.
Partial upstream fix
====================
The first report was related to a Linux Bridge testing using VMs on a
specific machine with an AMD CPU (EPYC 7402), and after a brief
investigation it turned out that the later change via:
ca96b162bfd2 ("x86: bring back rep movsq for user access on CPUs without ERMS")
... helped/fixed the performance issue.
However, after the later commit/fix was applied, then I got another
regression reported in a multistream TCP test on a 100Gbit mlx5 nic, also
running on an AMD based platform (AMD EPYC 7302 CPU), again that was using
iperf3 to run the test. That regression was after applying the later
fix/commit, but only this didn't help in telling the whole history.
Testing performed to pinpoint residual regression
=================================================
So I narrowed down the second regression use case, but running it
without traffic through a NIC, on localhost, in trying to narrow down
CPU usage and not being limited by other factor like network bandwidth.
I used another system also with an AMD CPU (AMD EPYC 7742). Basically,
I run iperf3 in server and client mode in the same system, for example:
- Start the server binding it to CPU core/thread 19:
$ taskset -c 19 iperf3 -D -s -B 127.0.0.1 -p 12000
- Start the client always binding/running on CPU core/thread 17, using
perf to get statistics:
$ perf stat -o stat.txt taskset -c 17 iperf3 -c 127.0.0.1 -b 0/1000 -V \
-n 50G --repeating-payload -l 16384 -p 12000 --cport 12001 2>&1 \
> stat-19.txt
For the client, always running/pinned to CPU 17. But for the iperf3 in
server mode, I did test runs using CPUs 19, 21, 23 or not pinned to any
specific CPU. So it basically consisted with four runs of the same
commands, just changing the CPU which the server is pinned, or without
pinning by removing the taskset call before the server command. The CPUs
were chosen based on NUMA node they were on, this is the relevant output
of lscpu on the system:
$ lscpu
...
Model name: AMD EPYC 7742 64-Core Processor
...
Caches (sum of all):
L1d: 2 MiB (64 instances)
L1i: 2 MiB (64 instances)
L2: 32 MiB (64 instances)
L3: 256 MiB (16 instances)
NUMA:
NUMA node(s): 4
NUMA node0 CPU(s): 0,1,8,9,16,17,24,25,32,33,40,41,48,49,56,57,64,65,72,73,80,81,88,89,96,97,104,105,112,113,120,121
NUMA node1 CPU(s): 2,3,10,11,18,19,26,27,34,35,42,43,50,51,58,59,66,67,74,75,82,83,90,91,98,99,106,107,114,115,122,123
NUMA node2 CPU(s): 4,5,12,13,20,21,28,29,36,37,44,45,52,53,60,61,68,69,76,77,84,85,92,93,100,101,108,109,116,117,124,125
NUMA node3 CPU(s): 6,7,14,15,22,23,30,31,38,39,46,47,54,55,62,63,70,71,78,79,86,87,94,95,102,103,110,111,118,119,126,127
...
So for the server run, when picking a CPU, I chose CPUs to be not on the same
node. The reason is with that I was able to get/measure relevant
performance differences when changing the alignment of the writes to the
destination in copy_user_generic.
Testing shows up to +81% performance improvement under iperf3
=============================================================
Here's a summary of the iperf3 runs:
# Vanilla upstream alignment:
CPU RATE SYS TIME sender-receiver
Server bind 19: 13.0Gbits/sec 28.371851000 33.233499566 86.9%-70.8%
Server bind 21: 12.9Gbits/sec 28.283381000 33.586486621 85.8%-69.9%
Server bind 23: 11.1Gbits/sec 33.660190000 39.012243176 87.7%-64.5%
Server bind none: 18.9Gbits/sec 19.215339000 22.875117865 86.0%-80.5%
# With the attached patch (aligning writes in non ERMS/FSRM case):
CPU RATE SYS TIME sender-receiver
Server bind 19: 20.8Gbits/sec 14.897284000 20.811101382 75.7%-89.0%
Server bind 21: 20.4Gbits/sec 15.205055000 21.263165909 75.4%-89.7%
Server bind 23: 20.2Gbits/sec 15.433801000 21.456175000 75.5%-89.8%
Server bind none: 26.1Gbits/sec 12.534022000 16.632447315 79.8%-89.6%
So I consistently got better results when aligning the write. The
results above were run on 6.14.0-rc6/rc7 based kernels. The sys is sys
time and then the total time to run/transfer 50G of data. The last
field is the CPU usage of sender/receiver iperf3 process. It's also
worth to note that each pair of iperf3 runs may get slightly different
results on each run, but I always got consistent higher results with
the write alignment for this specific test of running the processes
on CPUs in different NUMA nodes.
Linus Torvalds helped/provided this version of the patch. Initially I
proposed a version which aligned writes for all cases in
rep_movs_alternative, however it used two extra registers and thus
Linus provided an enhanced version that only aligns the write on the
large_movsq case, which is sufficient since the problem happens only
on those AMD CPUs like ones mentioned above without ERMS/FSRM, and
also doesn't require using extra registers. Also, I validated that
aligning only on large_movsq case is really enough for getting the
performance back.
I also tested this patch on an old Intel based non-ERMS/FRMS system
(with Xeon E5-2667 - Sandy Bridge based) and didn't get any problems:
no performance enhancement but also no regression either, using the
same iperf3 based benchmark. Also newer Intel processors after
Sandy Bridge usually have ERMS and should not be affected by this change.
[ mingo: Updated the changelog. ]
Fixes: ca96b162bfd2 ("x86: bring back rep movsq for user access on CPUs without ERMS") Fixes: 034ff37d3407 ("x86: rewrite '__copy_user_nocache' function") Reported-by: Ondrej Lichtner <olichtne@redhat.com> Co-developed-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Herton R. Krzesinski <herton@redhat.com> Signed-off-by: Ingo Molnar <mingo@kernel.org> Link: https://lore.kernel.org/r/20250320142213.2623518-1-herton@redhat.com Signed-off-by: Sasha Levin <sashal@kernel.org>
We're trying to mix non-PIC/PIE objects into the otherwise-PIE
relocatable kernels, to avoid GOT/PLT references during early boot
alternative resolution (which happens before the GOT/PLT are set up).
riscv64-unknown-linux-gnu-ld: arch/riscv/errata/sifive/errata.o: relocation R_RISCV_HI20 against `tlb_flush_all_threshold' can not be used when making a shared object; recompile with -fPIC
riscv64-unknown-linux-gnu-ld: arch/riscv/errata/thead/errata.o: relocation R_RISCV_HI20 against `riscv_cbom_block_size' can not be used when making a shared object; recompile with -fPIC
ASUS VivoBook X515JA with PCI SSID 1043:14f2 also hits the same issue
as other VivoBook model about the mic pin assignment, and the same
workaround is required to apply ALC256_FIXUP_ASUS_MIC_NO_PRESENCE
quirk.
Set ret = 0 on successful completion of the processing loop in
cs_dsp_load() and cs_dsp_load_coeff() to ensure that the function
returns 0 on success.
All normal firmware files will have at least one data block, and
processing this block will set ret == 0, from the result of either
regmap_raw_write() or cs_dsp_parse_coeff().
The kunit tests create a dummy firmware file that contains only the
header, without any data blocks. This gives cs_dsp a file to "load"
that will not cause any side-effects. As there aren't any data blocks,
the processing loop will not set ret == 0.
Originally there was a line after the processing loop:
ret = regmap_async_complete(regmap);
which would set ret == 0 before the function returned.
Commit fe08b7d5085a ("firmware: cs_dsp: Remove async regmap writes")
changed the regmap write to a normal sync write, so the call to
regmap_async_complete() wasn't necessary and was removed. It was
overlooked that the ret here wasn't only to check the result of
regmap_async_complete(), it also set the final return value of the
function.
Fixes: fe08b7d5085a ("firmware: cs_dsp: Remove async regmap writes") Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com> Link: https://patch.msgid.link/20250323170529.197205-1-rf@opensource.cirrus.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
There is a kernel API ntb_mw_clear_trans() would pass 0 to both addr and
size. This would make xlate_pos negative.
[ 23.734156] switchtec switchtec0: MW 0: part 0 addr 0x0000000000000000 size 0x0000000000000000
[ 23.734158] ================================================================================
[ 23.734172] UBSAN: shift-out-of-bounds in drivers/ntb/hw/mscc/ntb_hw_switchtec.c:293:7
[ 23.734418] shift exponent -1 is negative
Ensuring xlate_pos is a positive or zero before BIT.
Fixes: 1e2fd202f859 ("ntb_hw_switchtec: Check for alignment of the buffer in mw_set_trans()") Signed-off-by: Yajun Deng <yajun.deng@linux.dev> Reviewed-by: Logan Gunthorpe <logang@deltatee.com> Signed-off-by: Jon Mason <jdmason@kudzu.us> Signed-off-by: Sasha Levin <sashal@kernel.org>
This patch adds parentheses to parameters caller and callee of macros
make_call_t0 and make_call_ra. Every existing invocation of these two
macros uses a single variable for each argument, so the absence of the
parentheses seems okay. However, future invocations might use more
complex expressions as arguments. For example, a future invocation might
look like this: make_call_t0(a - b, c, call). Without parentheses in the
macro definition, the macro invocation expands to:
...
unsigned int offset = (unsigned long) c - (unsigned long) a - b;
...
which is clearly wrong.
The use of parentheses ensures arguments are correctly evaluated and
potentially saves future users of make_call_t0 and make_call_ra debugging
trouble.
Fixes: 6724a76cff85 ("riscv: ftrace: Reduce the detour code size to half") Signed-off-by: Juhan Jin <juhan.jin@foxmail.com> Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com> Link: https://lore.kernel.org/r/tencent_AE90AA59903A628E87E9F80E563DA5BA5508@qq.com Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When a 9p tree was mounted with option 'posixacl', parent directory had a
default ACL set for its subdirectories, e.g.:
setfacl -m default:group:simpsons:rwx parentdir
then creating a subdirectory crashed 9p client, as v9fs_fid_add() call in
function v9fs_vfs_mkdir_dotl() sets the passed 'fid' pointer to NULL
(since dafbe689736) even though the subsequent v9fs_set_create_acl() call
expects a valid non-NULL 'fid' pointer:
prior to "[POWERPC] spufs: Fix gang destroy leaks" we used to have
a problem with gang lifetimes - creation of a gang returns opened
gang directory, which normally gets removed when that gets closed,
but if somebody has created a context belonging to that gang and
kept it alive until the gang got closed, removal failed and we
ended up with a leak.
Unfortunately, it had been fixed the wrong way. Dentry of gang
directory was no longer pinned, and rmdir on close was gone.
One problem was that failure of open kept calling simple_rmdir()
as cleanup, which meant an unbalanced dput(). Another bug was
in the success case - gang creation incremented link count on
root directory, but that was no longer undone when gang got
destroyed.
Fix consists of
* reverting the commit in question
* adding a counter to gang, protected by ->i_rwsem
of gang directory inode.
* having it set to 1 at creation time, dropped
in both spufs_dir_close() and spufs_gang_close() and bumped
in spufs_create_context(), provided that it's not 0.
* using simple_recursive_removal() to take the gang
directory out when counter reaches zero.
Fixes: 877907d37da9 "[POWERPC] spufs: Fix gang destroy leaks" Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Sasha Levin <sashal@kernel.org>
It's called from spufs_fill_dir(), and caller of that will do
spufs_rmdir() in case of failure. That does remove everything
we'd managed to create, but... the problem dentry is still
negative. IOW, it needs to be explicitly dropped.
projects / thirdparty / kernel / stable.git / log
