Currently, __split_vma() triggers hugetlb page table unsharing through
vm_ops->may_split(). This happens before the VMA lock and rmap locks are
taken - which is too early, it allows racing VMA-locked page faults in our
process and racing rmap walks from other processes to cause page tables to
be shared again before we actually perform the split.
Fix it by explicitly calling into the hugetlb unshare logic from
__split_vma() in the same place where THP splitting also happens. At that
point, both the VMA and the rmap(s) are write-locked.
An annoying detail is that we can now call into the helper
hugetlb_unshare_pmds() from two different locking contexts:
1. from hugetlb_split(), holding:
- mmap lock (exclusively)
- VMA lock
- file rmap lock (exclusively)
2. hugetlb_unshare_all_pmds(), which I think is designed to be able to
call us with only the mmap lock held (in shared mode), but currently
only runs while holding mmap lock (exclusively) and VMA lock
Backporting note:
This commit fixes a racy protection that was introduced in commit b30c14cd6102 ("hugetlb: unshare some PMDs when splitting VMAs"); that
commit claimed to fix an issue introduced in 5.13, but it should actually
also go all the way back.
Like many Dell laptops, the 3.5mm port by default can not detect a
combined headphones+mic headset or even a pure microphone. This
change enables the port's functionality.
Signed-off-by: Jonathan Lane <jon@borg.moe> Cc: <stable@vger.kernel.org> Link: https://patch.msgid.link/20250611193124.26141-2-jon@borg.moe Signed-off-by: Takashi Iwai <tiwai@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
syzbort reported a uninitialize wiphy_work_lock in cfg80211_dev_free. [1]
After rfkill allocation fails, the wiphy release process will be performed,
which will cause cfg80211_dev_free to access the uninitialized wiphy_work
related data.
Move the initialization of wiphy_work to before rfkill initialization to
avoid this issue.
[1]
INFO: trying to register non-static key.
The code is fine but needs lockdep annotation, or maybe
you didn't initialize this object before use?
turning off the locking correctness validator.
CPU: 0 UID: 0 PID: 5935 Comm: syz-executor550 Not tainted 6.14.0-rc6-syzkaller-00103-g4003c9e78778 #0
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2~bpo12+1 04/01/2014
Call Trace:
<TASK>
__dump_stack lib/dump_stack.c:94 [inline]
dump_stack_lvl+0x116/0x1f0 lib/dump_stack.c:120
assign_lock_key kernel/locking/lockdep.c:983 [inline]
register_lock_class+0xc39/0x1240 kernel/locking/lockdep.c:1297
__lock_acquire+0x135/0x3c40 kernel/locking/lockdep.c:5103
lock_acquire.part.0+0x11b/0x380 kernel/locking/lockdep.c:5851
__raw_spin_lock_irqsave include/linux/spinlock_api_smp.h:110 [inline]
_raw_spin_lock_irqsave+0x3a/0x60 kernel/locking/spinlock.c:162
cfg80211_dev_free+0x30/0x3d0 net/wireless/core.c:1196
device_release+0xa1/0x240 drivers/base/core.c:2568
kobject_cleanup lib/kobject.c:689 [inline]
kobject_release lib/kobject.c:720 [inline]
kref_put include/linux/kref.h:65 [inline]
kobject_put+0x1e4/0x5a0 lib/kobject.c:737
put_device+0x1f/0x30 drivers/base/core.c:3774
wiphy_free net/wireless/core.c:1224 [inline]
wiphy_new_nm+0x1c1f/0x2160 net/wireless/core.c:562
ieee80211_alloc_hw_nm+0x1b7a/0x2260 net/mac80211/main.c:835
mac80211_hwsim_new_radio+0x1d6/0x54e0 drivers/net/wireless/virtual/mac80211_hwsim.c:5185
hwsim_new_radio_nl+0xb42/0x12b0 drivers/net/wireless/virtual/mac80211_hwsim.c:6242
genl_family_rcv_msg_doit+0x202/0x2f0 net/netlink/genetlink.c:1115
genl_family_rcv_msg net/netlink/genetlink.c:1195 [inline]
genl_rcv_msg+0x565/0x800 net/netlink/genetlink.c:1210
netlink_rcv_skb+0x16b/0x440 net/netlink/af_netlink.c:2533
genl_rcv+0x28/0x40 net/netlink/genetlink.c:1219
netlink_unicast_kernel net/netlink/af_netlink.c:1312 [inline]
netlink_unicast+0x53c/0x7f0 net/netlink/af_netlink.c:1338
netlink_sendmsg+0x8b8/0xd70 net/netlink/af_netlink.c:1882
sock_sendmsg_nosec net/socket.c:718 [inline]
__sock_sendmsg net/socket.c:733 [inline]
____sys_sendmsg+0xaaf/0xc90 net/socket.c:2573
___sys_sendmsg+0x135/0x1e0 net/socket.c:2627
__sys_sendmsg+0x16e/0x220 net/socket.c:2659
do_syscall_x64 arch/x86/entry/common.c:52 [inline]
do_syscall_64+0xcd/0x250 arch/x86/entry/common.c:83
Fixes: 72d520476a2f ("wifi: cfg80211: cancel wiphy_work before freeing wiphy") Reported-by: syzbot+aaf0488c83d1d5f4f029@syzkaller.appspotmail.com
Close: https://syzkaller.appspot.com/bug?extid=aaf0488c83d1d5f4f029 Tested-by: syzbot+aaf0488c83d1d5f4f029@syzkaller.appspotmail.com Signed-off-by: Edward Adam Davis <eadavis@qq.com> Link: https://patch.msgid.link/tencent_258DD9121DDDB9DD9A1939CFAA0D8625B107@qq.com Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: WangYuli <wangyuli@uniontech.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We should count the terminating NUL byte as part of the ctx_len.
Otherwise, UBSAN logs a warning:
UBSAN: array-index-out-of-bounds in security/selinux/xfrm.c:99:14
index 60 is out of range for type 'char [*]'
The allocation itself is correct so there is no actual out of bounds
indexing, just a warning.
If client set ->PreviousSessionId on kerberos session setup stage,
NULL pointer dereference error will happen. Since sess->user is not
set yet, It can pass the user argument as NULL to destroy_previous_session.
sess->user will be set in ksmbd_krb5_authenticate(). So this patch move
calling destroy_previous_session() after ksmbd_krb5_authenticate().
Cc: stable@vger.kernel.org Reported-by: zdi-disclosures@trendmicro.com # ZDI-CAN-27391 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>
When FRED is enabled, if the Trap Flag (TF) is set without an external
debugger attached, it can lead to an infinite loop in the SIGTRAP
handler. To avoid this, the software event flag in the augmented SS
must be cleared, ensuring that no single-step trap remains pending when
ERETU completes.
This test checks for that specific scenario—verifying whether the kernel
correctly prevents an infinite SIGTRAP loop in this edge case when FRED
is enabled.
The test should _always_ pass with IDT event delivery, thus no need to
disable the test even when FRED is not enabled.
Signed-off-by: Xin Li (Intel) <xin@zytor.com> Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Tested-by: Sohil Mehta <sohil.mehta@intel.com>
Cc:stable@vger.kernel.org Link: https://lore.kernel.org/all/20250609084054.2083189-3-xin%40zytor.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use common wrappers operating directly on the struct sg_table objects to
fix incorrect use of scatterlists sync calls. dma_sync_sg_for_*()
functions have to be called with the number of elements originally passed
to dma_map_sg_*() function, not the one returned in sgtable's nents.
Fixes: 1ffe09590121 ("udmabuf: fix dma-buf cpu access") CC: stable@vger.kernel.org Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com> Acked-by: Vivek Kasireddy <vivek.kasireddy@intel.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Christian König <christian.koenig@amd.com> Link: https://lore.kernel.org/r/20250507160913.2084079-3-m.szyprowski@samsung.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Improve the usability of the unit_add sysfs attribute by ensuring that
the associated FCP LUN scan processing is completed synchronously. This
enables configuration tooling to consistently determine the end of the
scan process to allow for serialization of follow-on actions.
While the scan process associated with unit_add typically completes
synchronously, it is deferred to an asynchronous background process if
unit_add is used before initial remote port scanning has completed. This
occurs when unit_add is used immediately after setting the associated FCP
device online.
To ensure synchronous unit_add processing, wait for remote port scanning
to complete before initiating the FCP LUN scan.
Cc: stable@vger.kernel.org Reviewed-by: M Nikhil <nikh1092@linux.ibm.com> Reviewed-by: Nihar Panda <niharp@linux.ibm.com> Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com> Signed-off-by: Nihar Panda <niharp@linux.ibm.com> Link: https://lore.kernel.org/r/20250603182252.2287285-2-niharp@linux.ibm.com Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently storvsc_timeout is only used in storvsc_sdev_configure(), and
5s and 10s are used elsewhere. It turns out that rarely the 5s is not
enough on Azure, so let's use storvsc_timeout everywhere.
In case a timeout happens and storvsc_channel_init() returns an error,
close the VMBus channel so that any host-to-guest messages in the
channel's ringbuffer, which might come late, can be safely ignored.
Add a "const" to storvsc_timeout.
Cc: stable@kernel.org Signed-off-by: Dexuan Cui <decui@microsoft.com> Link: https://lore.kernel.org/r/1749243459-10419-1-git-send-email-decui@microsoft.com Reviewed-by: Long Li <longli@microsoft.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently, cached directory contents were not reused across subsequent
'ls' operations because the cache validity check relied on comparing
the ctx pointer, which changes with each readdir invocation. As a
result, the cached dir entries was not marked as valid and the cache was
not utilized for subsequent 'ls' operations.
This change uses the file pointer, which remains consistent across all
readdir calls for a given directory instance, to associate and validate
the cache. As a result, cached directory contents can now be
correctly reused, improving performance for repeated directory listings.
Performance gains with local windows SMB server:
Without the patch and default actimeo=1:
1000 directory enumeration operations on dir with 10k files took 135.0s
With this patch and actimeo=0:
1000 directory enumeration operations on dir with 10k files took just 5.1s
Signed-off-by: Bharath SM <bharathsm@microsoft.com> Reviewed-by: Shyam Prasad N <sprasad@microsoft.com> Cc: stable@vger.kernel.org Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fuzzing hit another invalid pointer dereference due to the lack of
checking whether jffs2_prealloc_raw_node_refs() completed successfully.
Subsequent logic implies that the node refs have been allocated.
Handle that. The code is ready for propagating the error upwards.
Syzkaller detected a kernel bug in jffs2_link_node_ref, caused by fault
injection in jffs2_prealloc_raw_node_refs. jffs2_sum_write_sumnode doesn't
check return value of jffs2_prealloc_raw_node_refs and simply lets any
error propagate into jffs2_sum_write_data, which eventually calls
jffs2_link_node_ref in order to link the summary to an expectedly allocated
node.
When building kernel with LLVM there are occasionally such errors:
In file included from ./include/linux/spinlock.h:59:
In file included from ./include/linux/irqflags.h:17:
arch/loongarch/include/asm/irqflags.h:38:3: error: must not be $r0 or $r1
38 | "csrxchg %[val], %[mask], %[reg]\n\t"
| ^
<inline asm>:1:16: note: instantiated into assembly here
1 | csrxchg $a1, $ra, 0
| ^
To prevent the compiler from allocating $r0 or $r1 for the "mask" of the
csrxchg instruction, the 'q' constraint must be used but Clang < 21 does
not support it. So force to use $t0 in the inline asm, in order to avoid
using $r0/$r1 while keeping the backward compatibility.
Cc: stable@vger.kernel.org Link: https://github.com/llvm/llvm-project/pull/141037 Reviewed-by: Yanteng Si <si.yanteng@linux.dev> Suggested-by: WANG Rui <wangrui@loongson.cn> Signed-off-by: Huacai Chen <chenhuacai@loongson.cn> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Without correct unregisteration, ACPI notify handlers and the platform
drivers installed by generic_subdriver_init() will become dangling
references after removing the loongson_laptop module, triggering various
kernel faults when a hotkey is sent or at kernel shutdown.
Previously during driver probe, 1 is unconditionally taken as current
brightness value and set to props.brightness, which will be considered
as the brightness before suspend and restored to EC on resume. Since a
brightness value of 1 almost never matches EC's state on coldboot (my
laptop's EC defaults to 80), this causes surprising changes of screen
brightness on the first time of resume after coldboot.
Let's get brightness from EC and take it as the current brightness on
probe of the laptop driver to avoid the surprising behavior. Tested on
TongFang L860-T2 Loongson-3A5000 laptop.
cm_chan_msg_send() checks that userspace didn't send too much data but
riocm_ch_send() failed to check that userspace sent sufficient data. The
result is that riocm_ch_send() can write to fields in the rio_ch_chan_hdr
which were outside the bounds of the space which cm_chan_msg_send()
allocated.
Address this by teaching riocm_ch_send() to check that the entire
rio_ch_chan_hdr was copied in from userspace.
commit 7adb96687ce8 ("x86/bugs: Make spectre user default depend on
MITIGATION_SPECTRE_V2") depends on commit 72c70f480a70 ("x86/bugs: Add
a separate config for Spectre V2"), which introduced
MITIGATION_SPECTRE_V2.
commit 72c70f480a70 ("x86/bugs: Add a separate config for Spectre V2")
never landed in stable tree, thus, stable tree doesn't have
MITIGATION_SPECTRE_V2, that said, commit 7adb96687ce8 ("x86/bugs: Make
spectre user default depend on MITIGATION_SPECTRE_V2") has no value if
the dependecy was not applied.
Revert commit 7adb96687ce8 ("x86/bugs: Make spectre user default
depend on MITIGATION_SPECTRE_V2") in stable kernel which landed in in
5.4.294, 5.10.238, 5.15.185, 6.1.141 and 6.6.93 stable versions.
VFIO EEH recovery for PCI passthrough devices fails on PowerNV and pseries
platforms due to missing host-side PE bridge reconfiguration. In the
current implementation, eeh_pe_configure() only performs RTAS or OPAL-based
bridge reconfiguration for native host devices, but skips it entirely for
PEs managed through VFIO in guest passthrough scenarios.
This leads to incomplete EEH recovery when a PCI error affects a
passthrough device assigned to a QEMU/KVM guest. Although VFIO triggers the
EEH recovery flow through VFIO_EEH_PE_ENABLE ioctl, the platform-specific
bridge reconfiguration step is silently bypassed. As a result, the PE's
config space is not fully restored, causing subsequent config space access
failures or EEH freeze-on-access errors inside the guest.
This patch fixes the issue by ensuring that eeh_pe_configure() always
invokes the platform's configure_bridge() callback (e.g.,
pseries_eeh_phb_configure_bridge) even for VFIO-managed PEs. This ensures
that RTAS or OPAL calls to reconfigure the PE bridge are correctly issued
on the host side, restoring the PE's configuration space after an EEH
event.
This fix is essential for reliable EEH recovery in QEMU/KVM guests using
VFIO PCI passthrough on PowerNV and pseries systems.
Tested with:
- QEMU/KVM guest using VFIO passthrough (IBM Power9,(lpar)Power11 host)
- Injected EEH errors with pseries EEH errinjct tool on host, recovery
verified on qemu guest.
- Verified successful config space access and CAP_EXP DevCtl restoration
after recovery
The dell_rbu driver will use memset() to clear the data held by each
packet when it is no longer needed (when the driver is unloaded, the
packet size is changed, etc).
The amount of memory that is cleared (before this patch) is the normal
packet size. However, the last packet in the list may be smaller.
Fix this to only clear the memory actually used by each packet, to prevent
it from writing past the end of data buffer.
Because the packet data buffers are allocated with __get_free_pages() (in
page-sized increments), this bug could only result in a buffer being
overwritten when a packet size larger than one page is used. The only user
of the dell_rbu module should be the Dell BIOS update program, which uses
a packet size of 4096, so no issues should be seen without the patch, it
just blocks the possiblity.
Fixes: 6c54c28e69f2 ("[PATCH] dell_rbu: new Dell BIOS update driver") Signed-off-by: Stuart Hayes <stuart.w.hayes@gmail.com> Link: https://lore.kernel.org/r/20250609184659.7210-5-stuart.w.hayes@gmail.com Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com> Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Pass the correct list head to list_for_each_entry*() when looping through
the packet list.
Without this patch, reading the packet data via sysfs will show the data
incorrectly (because it starts at the wrong packet), and clearing the
packet list will result in a NULL pointer dereference.
Fixes: d19f359fbdc6 ("platform/x86: dell_rbu: don't open code list_for_each_entry*()") Signed-off-by: Stuart Hayes <stuart.w.hayes@gmail.com> Link: https://lore.kernel.org/r/20250609184659.7210-3-stuart.w.hayes@gmail.com Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com> Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
It breaks target-module@2b300050 ("ti,sysc-omap2") probe on AM62x in a case
when minimally-configured system tries to network-boot:
[ 6.888776] probe of 2b300050.target-module returned 517 after 258 usecs
[ 17.129637] probe of 2b300050.target-module returned 517 after 708 usecs
[ 17.137397] platform 2b300050.target-module: deferred probe pending: (reason unknown)
[ 26.878471] Waiting up to 100 more seconds for network.
There are minimal configurations possible when the deferred device is not
being probed any more (because everything else has been successfully
probed) and deferral lists are not processed any more.
Stable mmc enumeration can be achieved by filling /aliases node properly
(4700a00755fb commit's rationale).
The current code around TEE_IOCTL_PARAM_SIZE() is a bit wrong on
32-bit kernels: Multiplying a user-provided 32-bit value with the
size of a structure can wrap around on such platforms.
Fix it by using saturating arithmetic for the size calculation.
This has no security consequences because, in all users of
TEE_IOCTL_PARAM_SIZE(), the subsequent kcalloc() implicitly checks
for wrapping.
Don't put the l4ls clk domain to sleep in case of standby.
Since CM3 PM FW[1](ti-v4.1.y) doesn't wake-up/enable the l4ls clk domain
upon wake-up, CM3 PM FW fails to wake-up the MPU.
In case the MC firmware runs in debug mode with extensive prints pushed
to the console, the current timeout of 500ms is not enough.
Increase the timeout value so that we don't have any chance of wrongly
assuming that the firmware is not responding when it's just taking more
time.
We have to wait at least the minimium time for the watchdog window
(TWDMIN) before writings to the wdt register after the
watchdog is activated.
Otherwise the chip will assert TWD_ERROR and power down to reset mode.
It looks like attempting to write to the "store_modes" sysfs node will
run afoul of unregistered consoles:
UBSAN: array-index-out-of-bounds in drivers/video/fbdev/core/fbcon.c:122:28
index -1 is out of range for type 'fb_info *[32]'
...
fbcon_info_from_console+0x192/0x1a0 drivers/video/fbdev/core/fbcon.c:122
fbcon_new_modelist+0xbf/0x2d0 drivers/video/fbdev/core/fbcon.c:3048
fb_new_modelist+0x328/0x440 drivers/video/fbdev/core/fbmem.c:673
store_modes+0x1c9/0x3e0 drivers/video/fbdev/core/fbsysfs.c:113
dev_attr_store+0x55/0x80 drivers/base/core.c:2439
If con2fb_map contains a -1 things go wrong here. Instead, return NULL,
as callers of fbcon_info_from_console() are trying to compare against
existing "info" pointers, so error handling should kick in correctly.
It is desireable to push the hardware accelerator to also
process non-segmented TCP frames: we pass the skb->len
to the "TOE/TSO" offloader and it will handle them.
Without this quirk the driver becomes unstable and lock
up and and crash.
I do not know exactly why, but it is probably due to the
TOE (TCP offload engine) feature that is coupled with the
segmentation feature - it is not possible to turn one
part off and not the other, either both TOE and TSO are
active, or neither of them.
Not having the TOE part active seems detrimental, as if
that hardware feature is not really supposed to be turned
off.
The datasheet says:
"Based on packet parsing and TCP connection/NAT table
lookup results, the NetEngine puts the packets
belonging to the same TCP connection to the same queue
for the software to process. The NetEngine puts
incoming packets to the buffer or series of buffers
for a jumbo packet. With this hardware acceleration,
IP/TCP header parsing, checksum validation and
connection lookup are offloaded from the software
processing."
After numerous tests with the hardware locking up after
something between minutes and hours depending on load
using iperf3 I have concluded this is necessary to stabilize
the hardware.
We call skb_bpf_redirect_clear() to clean _sk_redir before handling skb in
backlog, but when sk_psock_handle_skb() return EAGAIN due to sk_rcvbuf
limit, the redirect info in _sk_redir is not recovered.
Fix skb redir loss during EAGAIN retries by restoring _sk_redir
information using skb_bpf_set_redir().
Before this patch:
'''
./bench sockmap -c 2 -p 1 -a --rx-verdict-ingress
Setting up benchmark 'sockmap'...
create socket fd c1:13 p1:14 c2:15 p2:16
Benchmark 'sockmap' started.
Send Speed 1343.172 MB/s, BPF Speed 1343.238 MB/s, Rcv Speed 65.271 MB/s
Send Speed 1352.022 MB/s, BPF Speed 1352.088 MB/s, Rcv Speed 0 MB/s
Send Speed 1354.105 MB/s, BPF Speed 1354.105 MB/s, Rcv Speed 0 MB/s
Send Speed 1355.018 MB/s, BPF Speed 1354.887 MB/s, Rcv Speed 0 MB/s
'''
Due to the high send rate, the RX processing path may frequently hit the
sk_rcvbuf limit. Once triggered, incorrect _sk_redir will cause the flow
to mistakenly enter the "!ingress" path, leading to send failures.
(The Rcv speed depends on tcp_rmem).
In case the rule already exists and another VSI wants to subscribe to it
new VSI list is being created and both VSIs are moved to it.
Currently, the check for already existing VSI with the same rule is done
based on fdw_id.hw_vsi_id, which applies only to LOOKUP_RX flag.
Change it to vsi_handle. This is software VSI ID, but it can be applied
here, because vsi_map itself is also based on it.
Additionally change return status in case the VSI already exists in the
VSI map to "Already exists". Such case should be handled by the caller.
The strlcat() with FORTIFY support is triggering a panic because it
thinks the target buffer will overflow although the correct target
buffer size is passed in.
Anyway, instead of memset() with 0 followed by a strlcat(), just use
memcpy() and ensure that the resulting buffer is NULL terminated.
BIOSVersion is only used for the lpfc_printf_log() which expects a
properly terminated string.
At startup, the driver just assumes that all registers have their
default values. But after a soft reset, the chip will just be in the
state it was, and some pins may have been configured as outputs. Any
modification of the output register will cause these pins to be driven
low, which leads to unexpected/unwanted effects. To prevent this from
happening, set the chip's IO configuration register to a known safe
mode (all inputs) before toggling any other bits.
software_node_get_reference_args() wants to get @index-th element, so
the property value requires at least '(index + 1) * sizeof(*ref)' bytes
but that can not be guaranteed by current OOB check, and may cause OOB
for malformed property.
Fix by using as OOB check '((index + 1) * sizeof(*ref) > prop->length)'.
FDB entries are allocated in an atomic context as they can be added from
the data path when learning is enabled.
After converting the FDB hash table to rhashtable, the insertion rate
will be much higher (*) which will entail a much higher rate of per-CPU
allocations via dst_cache_init().
When adding a large number of entries (e.g., 256k) in a batch, a small
percentage (< 0.02%) of these per-CPU allocations will fail [1]. This
does not happen with the current code since the insertion rate is low
enough to give the per-CPU allocator a chance to asynchronously create
new chunks of per-CPU memory.
Given that:
a. Only a small percentage of these per-CPU allocations fail.
b. The scenario where this happens might not be the most realistic one.
c. The driver can work correctly without dst caches. The dst_cache_*()
APIs first check that the dst cache was properly initialized.
d. The dst caches are not always used (e.g., 'tos inherit').
It seems reasonable to not treat these allocation failures as fatal.
Therefore, do not bail when dst_cache_init() fails and suppress warnings
by specifying '__GFP_NOWARN'.
[1] percpu: allocation failed, size=40 align=8 atomic=1, atomic alloc failed, no space left
(*) 97% reduction in average latency of vxlan_fdb_update() when adding
256k entries in a batch.
Reviewed-by: Petr Machata <petrm@nvidia.com> Signed-off-by: Ido Schimmel <idosch@nvidia.com> Link: https://patch.msgid.link/20250415121143.345227-14-idosch@nvidia.com Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org> Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When a bridge port STP state is changed from BLOCKING/DISABLED to
FORWARDING, the port's igmp query timer will NOT re-arm itself if the
bridge has been configured as per-VLAN multicast snooping.
Solve this by choosing the correct multicast context(s) to enable/disable
port multicast based on whether per-VLAN multicast snooping is enabled or
not, i.e. using per-{port, VLAN} context in case of per-VLAN multicast
snooping by re-implementing br_multicast_enable_port() and
br_multicast_disable_port() functions.
Before the patch, the IGMP query does not happen in the last step of the
following test sequence, i.e. no growth for tx counter:
# ip link add name br1 up type bridge vlan_filtering 1 mcast_snooping 1 mcast_vlan_snooping 1 mcast_querier 1 mcast_stats_enabled 1
# bridge vlan global set vid 1 dev br1 mcast_snooping 1 mcast_querier 1 mcast_query_interval 100 mcast_startup_query_count 0
# ip link add name swp1 up master br1 type dummy
# bridge link set dev swp1 state 0
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
1
# sleep 1
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
1
# bridge link set dev swp1 state 3
# sleep 2
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
1
After the patch, the IGMP query happens in the last step of the test:
# ip link add name br1 up type bridge vlan_filtering 1 mcast_snooping 1 mcast_vlan_snooping 1 mcast_querier 1 mcast_stats_enabled 1
# bridge vlan global set vid 1 dev br1 mcast_snooping 1 mcast_querier 1 mcast_query_interval 100 mcast_startup_query_count 0
# ip link add name swp1 up master br1 type dummy
# bridge link set dev swp1 state 0
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
1
# sleep 1
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
1
# bridge link set dev swp1 state 3
# sleep 2
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
3
Signed-off-by: Yong Wang <yongwang@nvidia.com> Reviewed-by: Andy Roulin <aroulin@nvidia.com> Reviewed-by: Ido Schimmel <idosch@nvidia.com> Signed-off-by: Petr Machata <petrm@nvidia.com> Acked-by: Nikolay Aleksandrov <razor@blackwall.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
When the vlan STP state is changed, which could be manipulated by
"bridge vlan" commands, similar to port STP state, this also impacts
multicast behaviors such as igmp query. In the scenario of per-VLAN
snooping, there's a need to update the corresponding multicast context
to re-arm the port query timer when vlan state becomes "forwarding" etc.
Update br_vlan_set_state() function to enable vlan multicast context
in such scenario.
Before the patch, the IGMP query does not happen in the last step of the
following test sequence, i.e. no growth for tx counter:
# ip link add name br1 up type bridge vlan_filtering 1 mcast_snooping 1 mcast_vlan_snooping 1 mcast_querier 1 mcast_stats_enabled 1
# bridge vlan global set vid 1 dev br1 mcast_snooping 1 mcast_querier 1 mcast_query_interval 100 mcast_startup_query_count 0
# ip link add name swp1 up master br1 type dummy
# sleep 1
# bridge vlan set vid 1 dev swp1 state 4
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
1
# sleep 1
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
1
# bridge vlan set vid 1 dev swp1 state 3
# sleep 2
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
1
After the patch, the IGMP query happens in the last step of the test:
# ip link add name br1 up type bridge vlan_filtering 1 mcast_snooping 1 mcast_vlan_snooping 1 mcast_querier 1 mcast_stats_enabled 1
# bridge vlan global set vid 1 dev br1 mcast_snooping 1 mcast_querier 1 mcast_query_interval 100 mcast_startup_query_count 0
# ip link add name swp1 up master br1 type dummy
# sleep 1
# bridge vlan set vid 1 dev swp1 state 4
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
1
# sleep 1
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
1
# bridge vlan set vid 1 dev swp1 state 3
# sleep 2
# ip -j -p stats show dev swp1 group xstats_slave subgroup bridge suite mcast | jq '.[]["multicast"]["igmp_queries"]["tx_v2"]'
3
Signed-off-by: Yong Wang <yongwang@nvidia.com> Reviewed-by: Andy Roulin <aroulin@nvidia.com> Reviewed-by: Ido Schimmel <idosch@nvidia.com> Signed-off-by: Petr Machata <petrm@nvidia.com> Acked-by: Nikolay Aleksandrov <razor@blackwall.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
In lpfc_check_sli_ndlp(), the get_job_els_rsp64_did remote_id assignment
does not apply for GEN_REQUEST64 commands as it only has meaning for a
ELS_REQUEST64 command. So, if (iocb->ndlp == ndlp) is false, we could
erroneously return the wrong value. Fix by replacing the fallthrough
statement with a break statement before the remote_id check.
issues were observed in deduplication between modules and kernel BTF
such that a large number of kernel types were not deduplicated so
were found in module BTF (task_struct, bpf_prog etc). The root cause
appeared to be a failure to dedup struct types, specifically those
with members that were pointers with __percpu annotations.
The issue in dedup is at the point that we are deduplicating structures,
we have not yet deduplicated reference types like pointers. If multiple
copies of a pointer point at the same (deduplicated) integer as in this
case, we do not see them as identical. Special handling already exists
to deal with structures and arrays, so add pointer handling here too.
When processing a PREQ the code would always check whether we have a
mesh path locally and reply accordingly. However, when forwarding is
disabled then we should not reply with this information as we will not
forward data packets down that path.
Move the check for dot11MeshForwarding up in the function and skip the
mesh path lookup in that case. In the else block, set forward to false
so that the rest of the function becomes a no-op and the
dot11MeshForwarding check does not need to be duplicated.
This explains an effect observed in the Freifunk community where mesh
forwarding is disabled. In that case a mesh with three STAs and only bad
links in between them, individual STAs would occionally have indirect
mpath entries. This should not have happened.
The MSE102x doesn't provide any interrupt register, so the only way
to handle the level interrupt is to fetch the whole packet from
the MSE102x internal buffer via SPI. So in cases the interrupt
handler fails to do this, it should return IRQ_NONE. This allows
the core to disable the interrupt in case the issue persists
and prevent an interrupt storm.
The statistics are incremented with raw_cpu_inc() assuming it always
happens with bottom half disabled. Without per-CPU locking in
local_bh_disable() on PREEMPT_RT this is no longer true.
Use this_cpu_inc() on PREEMPT_RT for the increment to not worry about
preemption.
TL;DR: SGX page reclaim touches the page to copy its contents to
secondary storage. SGX instructions do not gracefully handle machine
checks. Despite this, the existing SGX code will try to reclaim pages
that it _knows_ are poisoned. Avoid even trying to reclaim poisoned pages.
The longer story:
Pages used by an enclave only get epc_page->poison set in
arch_memory_failure() but they currently stay on sgx_active_page_list until
sgx_encl_release(), with the SGX_EPC_PAGE_RECLAIMER_TRACKED flag untouched.
epc_page->poison is not checked in the reclaimer logic meaning that, if other
conditions are met, an attempt will be made to reclaim an EPC page that was
poisoned. This is bad because 1. we don't want that page to end up added
to another enclave and 2. it is likely to cause one core to shut down
and the kernel to panic.
Specifically, reclaiming uses microcode operations including "EWB" which
accesses the EPC page contents to encrypt and write them out to non-SGX
memory. Those operations cannot handle MCEs in their accesses other than
by putting the executing core into a special shutdown state (affecting
both threads with HT.) The kernel will subsequently panic on the
remaining cores seeing the core didn't enter MCE handler(s) in time.
Call sgx_unmark_page_reclaimable() to remove the affected EPC page from
sgx_active_page_list on memory error to stop it being considered for
reclaiming.
Testing epc_page->poison in sgx_reclaim_pages() would also work but I assume
it's better to add code in the less likely paths.
The affected EPC page is not added to &node->sgx_poison_page_list until
later in sgx_encl_release()->sgx_free_epc_page() when it is EREMOVEd.
Membership on other lists doesn't change to avoid changing any of the
lists' semantics except for sgx_active_page_list. There's a "TBD" comment
in arch_memory_failure() about pre-emptive actions, the goal here is not
to address everything that it may imply.
This also doesn't completely close the time window when a memory error
notification will be fatal (for a not previously poisoned EPC page) --
the MCE can happen after sgx_reclaim_pages() has selected its candidates
or even *inside* a microcode operation (actually easy to trigger due to
the amount of time spent in them.)
The spinlock in sgx_unmark_page_reclaimable() is safe because
memory_failure() runs in process context and no spinlocks are held,
explicitly noted in a mm/memory-failure.c comment.
Signed-off-by: Andrew Zaborowski <andrew.zaborowski@intel.com> Signed-off-by: Ingo Molnar <mingo@kernel.org> Acked-by: Dave Hansen <dave.hansen@linux.intel.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Tony Luck <tony.luck@intel.com> Cc: balrogg@gmail.com Cc: linux-sgx@vger.kernel.org Link: https://lore.kernel.org/r/20250508230429.456271-1-andrew.zaborowski@intel.com Signed-off-by: Sasha Levin <sashal@kernel.org>
tcp_rcv_rtt_update() goal is to maintain an estimation of the RTT
in tp->rcv_rtt_est.rtt_us, used by tcp_rcv_space_adjust()
When TCP TS are enabled, tcp_rcv_rtt_update() is using
EWMA to smooth the samples.
Change this to immediately latch the incoming value if it
is lower than tp->rcv_rtt_est.rtt_us, so that tcp_rcv_space_adjust()
does not overshoot tp->rcvq_space.space and sk->sk_rcvbuf.
Firmware requests 2 segments at first. The first segment is of 6799360
whose allocation fails due to dma remapping not available. The success
is returned to firmware. Then firmware asks for 22 smaller segments
instead of 2 big ones. Those get allocated successfully. At suspend/
hibernation time, these segments aren't freed as they will be reused
by firmware after resuming.
After resuming, the firmware asks for the 2 segments again with the
first segment of 6799360 size. Since chunk->vaddr is not NULL, the
type and size are compared with the previous type and size to know if
it can be reused or not. Unfortunately, it is detected that it cannot
be reused and this first smaller segment is freed. Then we continue to
allocate 6799360 size memory which fails and ath11k_qmi_free_target_mem_chunk()
is called which frees the second smaller segment as well. Later success
is returned to firmware which asks for 22 smaller segments again. But
as we had freed 2 segments already, we'll allocate the first 2 new
smaller segments again and reuse the remaining 20. Hence 20 small
segments are being reused instead of 22.
Add skip logic when vaddr is set, but size/type don't match. Use the
same skip and success logic as used when dma_alloc_coherent() fails.
By skipping, the possibility of resume failure due to kernel failing to
allocate memory for QMI can be avoided.
kernel: ath11k_pci 0000:03:00.0: failed to allocate dma memory for qmi (524288 B type 1)
ath11k_pci 0000:03:00.0: failed to allocate qmi target memory: -22
This patch synchronizes code that accesses from both user-space
and IRQ contexts. The `get_stats()` function can be called from both
context.
`dev->stats.tx_errors` and `dev->stats.collisions` are also updated
in the `tx_errors()` function. Therefore, these fields must also be
protected by synchronized.
There is no code that accessses `dev->stats.tx_errors` between the
previous and updated lines, so the updating point can be moved.
During init of the bus, the module checks that the bus is idle.
If one of the lines are stuck try to recover them first before failing.
Sometimes SDA and SCL are low if improper reset occurs (e.g., reboot).
Signed-off-by: Tali Perry <tali.perry1@gmail.com> Signed-off-by: Mohammed Elbadry <mohammed.0.elbadry@gmail.com> Reviewed-by: Mukesh Kumar Savaliya <quic_msavaliy@quicinc.com> Link: https://lore.kernel.org/r/20250328193252.1570811-1-mohammed.0.elbadry@gmail.com Signed-off-by: Andi Shyti <andi.shyti@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Currently, all SCMI devices with performance domains attempt to register
a cpufreq driver, even if their performance domains aren't used to
control the CPUs. The cpufreq framework only supports registering a
single driver, so only the first device will succeed. And if that device
isn't used for the CPUs, then cpufreq will scale the wrong domains.
To avoid this, return early from scmi_cpufreq_probe() if the probing
SCMI device isn't referenced by the CPU device phandles.
This keeps the existing assumption that all CPUs are controlled by a
single SCMI device.
Function __sctp_write_space() doesn't set poll key, which leads to
ep_poll_callback() waking up all waiters, not only these waiting
for the socket being writable. Set the key properly using
wake_up_interruptible_poll(), which is preferred over the sync
variant, as writers are not woken up before at least half of the
queue is available. Also, TCP does the same.
This fixes the 5G connectibity issue on LiteOn WN4519R module
see https://github.com/openwrt/mt76/issues/971
And may also fix the 5G issues on the XBox One Wireless Adapter
see https://github.com/openwrt/mt76/issues/200
I have looked at the FCC info related to the MT7632U chip as mentioned in here:
https://github.com/openwrt/mt76/issues/459
These confirm the chipset does not support 'ac' mode and hence VHT should be turned of.
Logic here always sets hdr->version to 2 if it is not a BE3 or Lancer chip,
even if it is BE2. Use 'else if' to prevent multiple assignments, setting
version 0 for BE2, version 1 for BE3 and Lancer, and version 2 for others.
Fixes potential incorrect version setting when BE2_chip and
BE3_chip/lancer_chip checks could both be true.
Replaced pm_runtime_put() with pm_runtime_put_sync_suspend() to ensure
the runtime suspend is invoked immediately when unregistering a slave.
This prevents a race condition where suspend was skipped when
unregistering and registering slave in quick succession.
For example, consider the rapid sequence of
`delete_device -> new_device -> delete_device -> new_device`.
In this sequence, it is observed that the dw_i2c_plat_runtime_suspend()
might not be invoked after `delete_device` operation.
This is because after `delete_device` operation, when the
pm_runtime_put() is about to trigger suspend, the following `new_device`
operation might race and cancel the suspend.
If that happens, during the `new_device` operation,
dw_i2c_plat_runtime_resume() is skipped (since there was no suspend), which
means `i_dev->init()`, i.e. i2c_dw_init_slave(), is skipped.
Since i2c_dw_init_slave() is skipped, i2c_dw_configure_fifo_slave() is
skipped too, which leaves `DW_IC_INTR_MASK` unconfigured. If we inspect
the interrupt mask register using devmem, it will show as zero.
Example shell script to reproduce the issue:
```
#!/bin/sh
bpf_map_lookup_percpu_elem() helper is also available for sleepable bpf
program. When BPF JIT is disabled or under 32-bit host,
bpf_map_lookup_percpu_elem() will not be inlined. Using it in a
sleepable bpf program will trigger the warning in
bpf_map_lookup_percpu_elem(), because the bpf program only holds
rcu_read_lock_trace lock. Therefore, add the missed check.
The tipc_aead_free() function currently uses kfree() to release the aead
structure. However, this structure contains sensitive information, such
as key's SALT value, which should be securely erased from memory to
prevent potential leakage.
To enhance security, replace kfree() with kfree_sensitive() when freeing
the aead structure. This change ensures that sensitive data is explicitly
cleared before memory deallocation, aligning with the approach used in
tipc_aead_init() and adhering to best practices for handling confidential
information.
Maximum OTP and EEPROM size for hearthstone PCI1xxxx devices are 8 Kb
and 64 Kb respectively. Adjust max size definitions and return correct
EEPROM length based on device. Also prevent out-of-bound read/write.
The ACPI specification requires that battery rate is always positive,
but the kernel ABI for POWER_SUPPLY_PROP_CURRENT_NOW
(Documentation/ABI/testing/sysfs-class-power) specifies that it should
be negative when a battery is discharging. When reporting CURRENT_NOW,
massage the value to match the documented ABI.
This only changes the sign of `current_now` and not `power_now` because
documentation doesn't describe any particular meaning for `power_now` so
leaving `power_now` unchanged is less likely to confuse userspace
unnecessarily, whereas becoming consistent with the documented ABI is
worth potentially confusing clients that read `current_now`.
pm_runtime_put_autosuspend() schedules a hrtimer to expire
at "dev->power.timer_expires". If the hrtimer's callback,
pm_suspend_timer_fn(), observes that the current time equals
"dev->power.timer_expires", it unexpectedly bails out instead of
proceeding with runtime suspend.
Additionally, as ->timer_expires is not cleared, all the future auto
suspend requests will not schedule hrtimer to perform auto suspend.
rpm_suspend():
if ((rpmflags & RPM_AUTO) &&...) {
if (!(dev->power.timer_expires && ...) { <-- this will fail.
hrtimer_start_range_ns(&dev->power.suspend_timer,...);
}
}
Fix this by as well checking if current time reaches the set expiration.
Co-developed-by: Patrick Daly <quic_pdaly@quicinc.com> Signed-off-by: Patrick Daly <quic_pdaly@quicinc.com> Signed-off-by: Charan Teja Kalla <quic_charante@quicinc.com> Link: https://patch.msgid.link/20250515064125.1211561-1-quic_charante@quicinc.com Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Multiple applications may access the battery gauge at the same time, so
the gauge may be busy and EBUSY will be returned. The driver will set a
flag to record the EBUSY state, and this flag will be kept until the next
periodic update. When this flag is set, bq27xxx_battery_get_property()
will just return ENODEV until the flag is updated.
Even if the gauge was busy during the last accessing attempt, returning
ENODEV is not ideal, and can cause confusion in the applications layer.
Instead, retry accessing the I2C to update the flag is as expected, for
the gauge typically recovers from busy state within a few milliseconds.
If still failed to access the gauge, the real error code would be returned
instead of ENODEV (as suggested by Pali Rohár).
I analyzed this memory leak in detail. I found that “Acpi-State” cache and
“Acpi-Parse” cache were merged because the size of cache objects was same
slab cache size.
I finally found “Acpi-Parse” cache and “Acpi-parse_ext” cache were leaked
using SLAB_NEVER_MERGE flag in kmem_cache_create() function.
When early abort is occurred due to invalid ACPI information, Linux kernel
terminates ACPI by calling acpi_terminate() function. The function calls
acpi_ut_delete_caches() function to delete local caches (acpi_gbl_namespace_
cache, state_cache, operand_cache, ps_node_cache, ps_node_ext_cache).
But the deletion codes in acpi_ut_delete_caches() function only delete
slab caches using kmem_cache_destroy() function, therefore the cache
objects should be flushed before acpi_ut_delete_caches() function.
"Acpi-Parse" cache and "Acpi-ParseExt" cache are used in an AML parse
function, acpi_ps_parse_loop(). The function should complete all ops
using acpi_ps_complete_final_op() when an error occurs due to invalid
AML codes.
However, the current implementation of acpi_ps_complete_final_op() does not
complete all ops when it meets some errors and this cause cache leak.
This cache leak has a security threat because an old kernel (<= 4.9) shows
memory locations of kernel functions in stack dump. Some malicious users
could use this information to neutralize kernel ASLR.
To fix ACPI cache leak for enhancing security, I made a patch to complete all
ops unconditionally for acpi_ps_complete_final_op() function.
I hope that this patch improves the security of Linux kernel.
The ISENSE/VSENSE blocks are only powered up when the amplifier
transitions from shutdown to active. This means that if those controls
are flipped on while the amplifier is already playing back audio, they
will have no effect.
Fix this by forcing a power cycle around transitions in those controls.
Reviewed-by: Neal Gompa <neal@gompa.dev> Signed-off-by: Hector Martin <marcan@marcan.st> Signed-off-by: James Calligeros <jcalligeros99@gmail.com> Link: https://patch.msgid.link/20250406-apple-codec-changes-v5-1-50a00ec850a3@gmail.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
ap_get_table_length() checks if tables are valid by
calling ap_is_valid_header(). The latter then calls
ACPI_VALIDATE_RSDP_SIG(Table->Signature).
ap_is_valid_header() accepts struct acpi_table_header as an argument, so
the signature size is always fixed to 4 bytes.
The problem is when the string comparison is between ACPI-defined table
signature and ACPI_SIG_RSDP. Common ACPI table header specifies the
Signature field to be 4 bytes long[1], with the exception of the RSDP
structure whose signature is 8 bytes long "RSD PTR " (including the
trailing blank character)[2]. Calling strncmp(sig, rsdp_sig, 8) would
then result in a sequence overread[3] as sig would be smaller (4 bytes)
than the specified bound (8 bytes).
As a workaround, pass the bound conditionally based on the size of the
signature being passed.
Adds the MMC_QUIRK_NO_UHS_DDR50_TUNING quirk and updates
mmc_execute_tuning() to return 0 if that quirk is set. This fixes an
issue on certain Swissbit SD cards that do not support DDR50 tuning
where tuning requests caused I/O errors to be thrown.
Right now, if the clocksource watchdog detects a clocksource skew, it might
perform a per CPU check, for example in the TSC case on x86. In other
words: supposing TSC is detected as unstable by the clocksource watchdog
running at CPU1, as part of marking TSC unstable the kernel will also run a
check of TSC readings on some CPUs to be sure it is synced between them
all.
But that check happens only on some CPUs, not all of them; this choice is
based on the parameter "verify_n_cpus" and in some random cpumask
calculation. So, the watchdog runs such per CPU checks on up to
"verify_n_cpus" random CPUs among all online CPUs, with the risk of
repeating CPUs (that aren't double checked) in the cpumask random
calculation.
But if "verify_n_cpus" > num_online_cpus(), it should skip the random
calculation and just go ahead and check the clocksource sync between
all online CPUs, without the risk of skipping some CPUs due to
duplicity in the random cpumask calculation.
Tests in a 4 CPU laptop with TSC skew detected led to some cases of the per
CPU verification skipping some CPU even with verify_n_cpus=8, due to the
duplicity on random cpumask generation. Skipping the randomization when the
number of online CPUs is smaller than verify_n_cpus, solves that.
Suggested-by: Thadeu Lima de Souza Cascardo <cascardo@igalia.com> Signed-off-by: Guilherme G. Piccoli <gpiccoli@igalia.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Link: https://lore.kernel.org/all/20250323173857.372390-1-gpiccoli@igalia.com Signed-off-by: Sasha Levin <sashal@kernel.org>
Similar to many other Lenovo models with AMD chips, the Lenovo
Yoga Pro 7 14ASP9 (product name 83HN) requires a specific quirk
to ensure internal mic detection. This patch adds a quirk fixing this.
I found an ACPI cache leak in ACPI early termination and boot continuing case.
When early termination occurs due to malicious ACPI table, Linux kernel
terminates ACPI function and continues to boot process. While kernel terminates
ACPI function, kmem_cache_destroy() reports Acpi-Operand cache leak.
I analyzed this memory leak in detail and found acpi_ds_obj_stack_pop_and_
delete() function miscalculated the top of the stack. acpi_ds_obj_stack_push()
function uses walk_state->operand_index for start position of the top, but
acpi_ds_obj_stack_pop_and_delete() function considers index 0 for it.
Therefore, this causes acpi operand memory leak.
This cache leak causes a security threat because an old kernel (<= 4.9) shows
memory locations of kernel functions in stack dump. Some malicious users
could use this information to neutralize kernel ASLR.
Fix incorrect value mask for register write. Register values are 8-bit,
not 9. If this function was called with a value > 0xFF and an even addr,
it would cause writing to the next register.
Fixes: f2a22e1e172f ("iio: adc: ad7606: Add support for software mode for ad7616") Signed-off-by: David Lechner <dlechner@baylibre.com> Reviewed-by: Angelo Dureghello <adureghello@baylibre.com> Link: https://patch.msgid.link/20250428-iio-adc-ad7606_spi-fix-write-value-mask-v1-1-a2d5e85a809f@baylibre.com Cc: <Stable@vger.kernel.org> Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
>From the documentation:
"offset to be added to <type>[Y]_raw prior toscaling by <type>[Y]_scale"
Offset should be applied before multiplying scale, so divide offset by
scale to make this correct.
Fixes: bc3eb0207fb5 ("iio: imu: inv_icm42600: add temperature sensor support") Signed-off-by: Sean Nyekjaer <sean@geanix.com> Acked-by: Jean-Baptiste Maneyrol <jean-baptiste.maneyrol@tdk.com> Link: https://patch.msgid.link/20250502-imu-v1-1-129b8391a4e3@geanix.com Cc: <Stable@vger.kernel.org> Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Mark the temperature element signed, data read from the TEMP_OUT register
is in two's complement format.
This will avoid the temperature being mishandled and miss displayed.
Fixes: a3e0b51884ee ("iio: accel: add support for FXLS8962AF/FXLS8964AF accelerometers") Suggested-by: Marcelo Schmitt <marcelo.schmitt1@gmail.com> Cc: stable@vger.kernel.org Reviewed-by: Marcelo Schmitt <marcelo.schmitt1@gmail.com> Signed-off-by: Sean Nyekjaer <sean@geanix.com> Link: https://patch.msgid.link/20250505-fxls-v4-2-a38652e21738@geanix.com Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The commit a4e772898f8b ("PCI: Add missing bridge lock to pci_bus_lock()")
made the lock function to call depend on dev->subordinate but left
pci_slot_unlock() unmodified creating locking asymmetry compared with
pci_slot_lock().
Because of the asymmetric lock handling, the same bridge device is unlocked
twice. First pci_bus_unlock() unlocks bus->self and then pci_slot_unlock()
will unconditionally unlock the same bridge device.
Move pci_dev_unlock() inside an else branch to match the logic in
pci_slot_lock().
Loongson PCIe Root Ports don't advertise an ACS capability, but they do not
allow peer-to-peer transactions between Root Ports. Add an ACS quirk so
each Root Port can be in a separate IOMMU group.
While cdns_pcie_ep_set_msix() writes the Table Size field correctly (N-1),
the calculation of the PBA offset is wrong because it calculates space for
(N-1) entries instead of N.
This results in the following QEMU error when using PCI passthrough on a
device which relies on the PCI endpoint subsystem:
failed to add PCI capability 0x11[0x50]@0xb0: table & pba overlap, or they don't fit in BARs, or don't align
Fix the calculation of PBA offset in the MSI-X capability.
[bhelgaas: more specific subject and commit log]
Fixes: 3ef5d16f50f8 ("PCI: cadence: Add MSI-X support to Endpoint driver") Signed-off-by: Niklas Cassel <cassel@kernel.org> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Wilfred Mallawa <wilfred.mallawa@wdc.com> Reviewed-by: Damien Le Moal <dlemoal@kernel.org> Cc: stable@vger.kernel.org Link: https://patch.msgid.link/20250514074313.283156-10-cassel@kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Interrupt and monitor pages should be in Hyper-V page size (4k bytes).
This can be different from the system page size.
This size is read and used by the user-mode program to determine the
mapped data region. An example of such user-mode program is the VMBus
driver in DPDK.
Cc: stable@vger.kernel.org Fixes: 95096f2fbd10 ("uio-hv-generic: new userspace i/o driver for VMBus") Signed-off-by: Long Li <longli@microsoft.com> Reviewed-by: Michael Kelley <mhklinux@outlook.com> Link: https://lore.kernel.org/r/1746492997-4599-3-git-send-email-longli@linuxonhyperv.com Signed-off-by: Wei Liu <wei.liu@kernel.org>
Message-ID: <1746492997-4599-3-git-send-email-longli@linuxonhyperv.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
cifs_reconnect can be called with a flag to mark the session as needing
reconnect too. When this is done, we expect the connections of all
channels to be reconnected too, which is not happening today.
Without doing this, we have seen bad things happen when primary and
secondary channels are connected to different servers (in case of cloud
services like Azure Files SMB).
This change would force all connections to reconnect as well, not just
the sessions and tcons.
Cc: <stable@vger.kernel.org> 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>
When rproc->state = RPROC_DETACHED is attached to remote processor
through rproc_attach(), if rproc_handle_resources() returns failure,
then the clean table should be released, otherwise the following
memory leak will occur.
When rproc->state = RPROC_DETACHED and rproc_attach() is used
to attach to the remote processor, if rproc_handle_resources()
returns a failure, the resources allocated by imx_rproc_prepare()
should be released, otherwise the following memory leak will occur.
Since almost the same thing is done in imx_rproc_prepare() and
rproc_resource_cleanup(), Function rproc_resource_cleanup() is able
to deal with empty lists so it is better to fix the "goto" statements
in rproc_attach(). replace the "unprepare_device" goto statement with
"clean_up_resources" and get rid of the "unprepare_device" label.
projects / thirdparty / kernel / stable.git / log
