If "udp_cmsg_send()" returned 0 (i.e. only UDP cmsg),
"connected" should not be set to 0. Otherwise it stops
the connected socket from using the cached route.
Fixes: 2e8de8576343 ("udp: add gso segment cmsg") Signed-off-by: Yick Xie <yick.xie@gmail.com> Cc: stable@vger.kernel.org Reviewed-by: Willem de Bruijn <willemb@google.com> Link: https://lore.kernel.org/r/20240418170610.867084-1-yick.xie@gmail.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
With deferred IO enabled, a page fault happens when data is written to the
framebuffer device. Then driver determines which page is being updated by
calculating the offset of the written virtual address within the virtual
memory area, and uses this offset to get the updated page within the
internal buffer. This page is later copied to hardware (thus the name
"deferred IO").
This offset calculation is only correct if the virtual memory area is
mapped to the beginning of the internal buffer. Otherwise this is wrong.
For example, if users do:
mmap(ptr, 4096, PROT_WRITE, MAP_FIXED | MAP_SHARED, fd, 0xff000);
Then the virtual memory area will mapped at offset 0xff000 within the
internal buffer. This offset 0xff000 is not accounted for, and wrong page
is updated.
Correct the calculation by using vmf->pgoff instead. With this change, the
variable "offset" will no longer hold the exact offset value, but it is
rounded down to multiples of PAGE_SIZE. But this is still correct, because
this variable is only used to calculate the page offset.
Reported-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com> Closes: https://lore.kernel.org/linux-fbdev/271372d6-e665-4e7f-b088-dee5f4ab341a@oracle.com Fixes: 56c134f7f1b5 ("fbdev: Track deferred-I/O pages in pageref struct") Cc: <stable@vger.kernel.org> Signed-off-by: Nam Cao <namcao@linutronix.de> Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de> Tested-by: Harshit Mogalapalli <harshit.m.mogalapalli@oracle.com> Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de> Link: https://patchwork.freedesktop.org/patch/msgid/20240423115053.4490-1-namcao@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If stack_depot_save_flags() allocates memory it always drops
__GFP_NOLOCKDEP flag. So when KASAN tries to track __GFP_NOLOCKDEP
allocation we may end up with lockdep splat like bellow:
======================================================
WARNING: possible circular locking dependency detected
6.9.0-rc3+ #49 Not tainted
------------------------------------------------------
kswapd0/149 is trying to acquire lock: ffff88811346a920
(&xfs_nondir_ilock_class){++++}-{4:4}, at: xfs_reclaim_inode+0x3ac/0x590
[xfs]
but task is already holding lock: ffffffff8bb33100 (fs_reclaim){+.+.}-{0:0}, at:
balance_pgdat+0x5d9/0xad0
b44_free_rings() accesses b44::rx_buffers (and ::tx_buffers)
unconditionally, but b44::rx_buffers is only valid when the
device is up (they get allocated in b44_open(), and deallocated
again in b44_close()), any other time these are just a NULL pointers.
So if you try to change the pause params while the network interface
is disabled/administratively down, everything explodes (which likely
netifd tries to do).
Link: https://github.com/openwrt/openwrt/issues/13789 Fixes: 1da177e4c3f4 (Linux-2.6.12-rc2) Cc: stable@vger.kernel.org Reported-by: Peter Münster <pm@a16n.net> Suggested-by: Jonas Gorski <jonas.gorski@gmail.com> Signed-off-by: Vaclav Svoboda <svoboda@neng.cz> Tested-by: Peter Münster <pm@a16n.net> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Signed-off-by: Peter Münster <pm@a16n.net> Reviewed-by: Michael Chan <michael.chan@broadcom.com> Link: https://lore.kernel.org/r/87y192oolj.fsf@a16n.net Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 2f4a4d63a193 ("ACPI: CPPC: Use access_width over bit_width for system
memory accesses") modified cpc_read()/cpc_write() to use access_width to
read CPC registers.
However, for PCC registers the access width field in the ACPI register
macro specifies the PCC subspace ID. For non-zero PCC subspace ID it is
incorrectly treated as access width. This causes errors when reading
from PCC registers in the CPPC driver.
For PCC registers, base the size of read/write on the bit width field.
The debug message in cpc_read()/cpc_write() is updated to print relevant
information for the address space type used to read the register.
Fixes: 2f4a4d63a193 ("ACPI: CPPC: Use access_width over bit_width for system memory accesses") Signed-off-by: Vanshidhar Konda <vanshikonda@os.amperecomputing.com> Tested-by: Jarred White <jarredwhite@linux.microsoft.com> Reviewed-by: Jarred White <jarredwhite@linux.microsoft.com> Reviewed-by: Easwar Hariharan <eahariha@linux.microsoft.com> Cc: 5.15+ <stable@vger.kernel.org> # 5.15+ Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 2f4a4d63a193 ("ACPI: CPPC: Use access_width over bit_width for
system memory accesses") neglected to properly wrap the bit_offset shift
when it comes to applying the mask. This may cause incorrect values to be
read and may cause the cpufreq module not be loaded.
[ 11.059751] cpu_capacity: CPU0 missing/invalid highest performance.
[ 11.066005] cpu_capacity: partial information: fallback to 1024 for all CPUs
Also, corrected the bitmask generation in GENMASK (extra bit being added).
Fixes: 2f4a4d63a193 ("ACPI: CPPC: Use access_width over bit_width for system memory accesses") Signed-off-by: Jarred White <jarredwhite@linux.microsoft.com> Cc: 5.15+ <stable@vger.kernel.org> # 5.15+ Reviewed-by: Vanshidhar Konda <vanshikonda@os.amperecomputing.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To align with ACPI 6.3+, since bit_width can be any 8-bit value, it
cannot be depended on to be always on a clean 8b boundary. This was
uncovered on the Cobalt 100 platform.
Instead, use access_width to determine the size and use the offset and
width to shift and mask the bits to read/write out. Make sure to add a
check for system memory since pcc redefines the access_width to
subspace id.
If access_width is not set, then fall back to using bit_width.
Signed-off-by: Jarred White <jarredwhite@linux.microsoft.com> Reviewed-by: Easwar Hariharan <eahariha@linux.microsoft.com> Cc: 5.15+ <stable@vger.kernel.org> # 5.15+
[ rjw: Subject and changelog edits, comment adjustments ] Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The error handling path in its_vpe_irq_domain_alloc() causes a double free
when its_vpe_init() fails after successfully allocating at least one
interrupt. This happens because its_vpe_irq_domain_free() frees the
interrupts along with the area bitmap and the vprop_page and
its_vpe_irq_domain_alloc() subsequently frees the area bitmap and the
vprop_page again.
Fix this by unconditionally invoking its_vpe_irq_domain_free() which
handles all cases correctly and by removing the bitmap/vprop_page freeing
from its_vpe_irq_domain_alloc().
This avoids a potential conflict with firmwares with the newer
HDP flush mechanism.
Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Q7_THRM# pin is connected to a diode on the module which is used
as a level shifter, and the pin have a pull-down enabled by
default. We need to configure it to internal pull-up, other-
wise whenever the pin is configured as INPUT and we try to
control it externally the value will always remain zero.
While adding the GIC ITS MSI support, it was found that the msi-map entries
needed to be swapped to receive MSIs from the endpoint.
But later it was identified that the swapping was needed due to a bug in
the Qualcomm PCIe controller driver. And since the bug is now fixed with
commit bf79e33cdd89 ("PCI: qcom: Enable BDF to SID translation properly"),
let's fix the msi-map entries also to reflect the actual mapping in the
hardware.
Cc: stable@vger.kernel.org # 6.3: bf79e33cdd89 ("PCI: qcom: Enable BDF to SID translation properly") Fixes: ff384ab56f16 ("arm64: dts: qcom: sm8450: Use GIC-ITS for PCIe0 and PCIe1") Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org> Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org> Link: https://lore.kernel.org/r/20240318-pci-bdf-sid-fix-v1-1-acca6c5d9cf1@linaro.org Signed-off-by: Bjorn Andersson <andersson@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add the missing PCIe CX performance level votes to avoid relying on
other drivers (e.g. USB or UFS) to maintain the nominal performance
level required for Gen3 speeds.
In order to fix perf's callchain parse error for LoongArch, we implement
perf_arch_fetch_caller_regs() which fills several necessary registers
used for callchain unwinding, including sp, fp, and era. This is similar
to the following commits.
Rename x86's to CPU_MITIGATIONS, define it in generic code, and force it
on for all architectures exception x86. A recent commit to turn
mitigations off by default if SPECULATION_MITIGATIONS=n kinda sorta
missed that "cpu_mitigations" is completely generic, whereas
SPECULATION_MITIGATIONS is x86-specific.
Rename x86's SPECULATIVE_MITIGATIONS instead of keeping both and have it
select CPU_MITIGATIONS, as having two configs for the same thing is
unnecessary and confusing. This will also allow x86 to use the knob to
manage mitigations that aren't strictly related to speculative
execution.
Use another Kconfig to communicate to common code that CPU_MITIGATIONS
is already defined instead of having x86's menu depend on the common
CPU_MITIGATIONS. This allows keeping a single point of contact for all
of x86's mitigations, and it's not clear that other architectures *want*
to allow disabling mitigations at compile-time.
Fixes: f337a6a21e2f ("x86/cpu: Actually turn off mitigations by default for SPECULATION_MITIGATIONS=n") Closes: https://lkml.kernel.org/r/20240413115324.53303a68%40canb.auug.org.au Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> Reported-by: Michael Ellerman <mpe@ellerman.id.au> Reported-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Acked-by: Borislav Petkov (AMD) <bp@alien8.de> Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20240420000556.2645001-2-seanjc@google.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Bytes 40-65535 of 65536 are uninitialized
Memory access of size 65536 starts at ffff888045a40000
This happens, because we're copying a 'struct btrfs_data_container' back
to user-space. This btrfs_data_container is allocated in
'init_data_container()' via kvmalloc(), which does not zero-fill the
memory.
Fix this by using kvzalloc() which zeroes out the memory on allocation.
CC: stable@vger.kernel.org # 4.14+ Reported-by: <syzbot+510a1abbb8116eeb341d@syzkaller.appspotmail.com> Reviewed-by: Qu Wenruo <wqu@suse.com> Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Johannes Thumshirn <Johannes.thumshirn@wdc.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When btrfs scrub finds an error, it reads mirrors to find correct data. If
all the errors are fixed, sctx->error_bitmap is cleared for the stripe
range. However, in the zoned mode, it runs relocation to repair scrub
errors when the bitmap is *not* empty, which is a flipped condition.
Also, it runs the relocation even if the scrub is read-only. This was
missed by a fix in commit 1f2030ff6e49 ("btrfs: scrub: respect the
read-only flag during repair").
The repair is only necessary when there is a repaired sector and should be
done on read-write scrub. So, tweak the condition for both regular and
zoned case.
Fixes: 54765392a1b9 ("btrfs: scrub: introduce helper to queue a stripe for scrub") Fixes: 1f2030ff6e49 ("btrfs: scrub: respect the read-only flag during repair") CC: stable@vger.kernel.org # 6.6+ Reviewed-by: Qu Wenruo <wqu@suse.com> Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[BUG]
During my extent_map cleanup/refactor, with extra sanity checks,
extent-map-tests::test_case_7() would not pass the checks.
The problem is, after btrfs_drop_extent_map_range(), the resulted
extent_map has a @block_start way too large.
Meanwhile my btrfs_file_extent_item based members are returning a
correct @disk_bytenr/@offset combination.
The extent map layout looks like this:
0 16K 32K 48K
| PINNED | | Regular |
The regular em at [32K, 48K) also has 32K @block_start.
Then drop range [0, 36K), which should shrink the regular one to be
[36K, 48K).
However the @block_start is incorrect, we expect 32K + 4K, but got 52K.
[CAUSE]
Inside btrfs_drop_extent_map_range() function, if we hit an extent_map
that covers the target range but is still beyond it, we need to split
that extent map into half:
|<-- drop range -->|
|<----- existing extent_map --->|
And if the extent map is not compressed, we need to forward
extent_map::block_start by the difference between the end of drop range
and the extent map start.
However in that particular case, the difference is calculated using
(start + len - em->start).
The problem is @start can be modified if the drop range covers any
pinned extent.
This leads to wrong calculation, and would be caught by my later
extent_map sanity checks, which checks the em::block_start against
btrfs_file_extent_item::disk_bytenr + btrfs_file_extent_item::offset.
This is a regression caused by commit c962098ca4af ("btrfs: fix
incorrect splitting in btrfs_drop_extent_map_range"), which removed the
@len update for pinned extents.
[FIX]
Fix it by avoiding using @start completely, and use @end - em->start
instead, which @end is exclusive bytenr number.
And update the test case to verify the @block_start to prevent such
problem from happening.
Thankfully this is not going to lead to any data corruption, as IO path
does not utilize btrfs_drop_extent_map_range() with @skip_pinned set.
So this fix is only here for the sake of consistency/correctness.
CC: stable@vger.kernel.org # 6.5+ Fixes: c962098ca4af ("btrfs: fix incorrect splitting in btrfs_drop_extent_map_range") Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Qu Wenruo <wqu@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In commit b4ccace878f4 ("btrfs: refactor submit_compressed_extents()"), if
an async extent compressed but failed to find enough space, we changed
from falling back to an uncompressed write to just failing the write
altogether. The principle was that if there's not enough space to write
the compressed version of the data, there can't possibly be enough space
to write the larger, uncompressed version of the data.
However, this isn't necessarily true: due to fragmentation, there could
be enough discontiguous free blocks to write the uncompressed version,
but not enough contiguous free blocks to write the smaller but
unsplittable compressed version.
This has occurred to an internal workload which relied on write()'s
return value indicating there was space. While rare, it has happened a
few times.
Thus, in order to prevent early ENOSPC, re-add a fallback to
uncompressed writing.
The flag I2C_HID_READ_PENDING is used to serialize I2C operations.
However, this is not necessary, because I2C core already has its own
locking for that.
More importantly, this flag can cause a lock-up: if the flag is set in
i2c_hid_xfer() and an interrupt happens, the interrupt handler
(i2c_hid_irq) will check this flag and return immediately without doing
anything, then the interrupt handler will be invoked again in an
infinite loop.
Since interrupt handler is an RT task, it takes over the CPU and the
flag-clearing task never gets scheduled, thus we have a lock-up.
Delete this unnecessary flag.
Reported-and-tested-by: Eva Kurchatova <nyandarknessgirl@gmail.com> Closes: https://lore.kernel.org/r/CA+eeCSPUDpUg76ZO8dszSbAGn+UHjcyv8F1J-CUPVARAzEtW9w@mail.gmail.com Fixes: 4a200c3b9a40 ("HID: i2c-hid: introduce HID over i2c specification implementation") Cc: <stable@vger.kernel.org> Signed-off-by: Nam Cao <namcao@linutronix.de> Signed-off-by: Jiri Kosina <jkosina@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Generic sdhci code registers LED device and uses host->runtime_suspended
flag to protect access to it. The sdhci-msm driver doesn't set this flag,
which causes a crash when LED is accessed while controller is runtime
suspended. Fix this by setting the flag correctly.
Cc: stable@vger.kernel.org Fixes: 67e6db113c90 ("mmc: sdhci-msm: Add pm_runtime and system PM support") Signed-off-by: Mantas Pucka <mantas@8devices.com> Acked-by: Adrian Hunter <adrian.hunter@intel.com> Link: https://lore.kernel.org/r/20240321-sdhci-mmc-suspend-v1-1-fbc555a64400@8devices.com Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Should be an issue in hugetlb but triggered in an userfault context, where
it goes into the unlikely path where two threads modifying the resv map
together. Mike has a fix in that path for resv uncharge but it looks like
the locking criteria was overlooked: hugetlb_cgroup_uncharge_folio_rsvd()
will update the cgroup pointer, so it requires to be called with the lock
held.
Link: https://lkml.kernel.org/r/20240417211836.2742593-3-peterx@redhat.com Fixes: 79aa925bf239 ("hugetlb_cgroup: fix reservation accounting") Signed-off-by: Peter Xu <peterx@redhat.com> Reported-by: syzbot+4b8077a5fccc61c385a1@syzkaller.appspotmail.com Reviewed-by: Mina Almasry <almasrymina@google.com> Cc: David Hildenbrand <david@redhat.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
While migrating to exec_ops in commit a82990c8a409 ("mtd: rawnand: qcom:
Add read/read_start ops in exec_op path"), OP_RESET_DEVICE command handling
got broken unintentionally. Right now for the OP_RESET_DEVICE command,
qcom_misc_cmd_type_exec() will simply return 0 without handling it. Even,
if that gets fixed, an unnecessary FLASH_STATUS read descriptor command is
being added in the middle and that seems to be causing the command to fail
on IPQ806x devices.
So let's fix the above two issues to make OP_RESET_DEVICE command working
again.
Qualcomm ROME controllers can be registered from the Bluetooth line
discipline and in this case the HCI UART serdev pointer is NULL.
Add the missing sanity check to prevent a NULL-pointer dereference when
setup() is called for a non-serdev controller.
Fixes: e9b3e5b8c657 ("Bluetooth: hci_qca: only assign wakeup with serial port support") Cc: stable@vger.kernel.org # 6.2 Cc: Zhengping Jiang <jiangzp@google.com> Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Qualcomm ROME controllers can be registered from the Bluetooth line
discipline and in this case the HCI UART serdev pointer is NULL.
Add the missing sanity check to prevent a NULL-pointer dereference when
wakeup() is called for a non-serdev controller during suspend.
Just return true for now to restore the original behaviour and address
the crash with pre-6.2 kernels, which do not have commit e9b3e5b8c657
("Bluetooth: hci_qca: only assign wakeup with serial port support") that
causes the crash to happen already at setup() time.
Fixes: c1a74160eaf1 ("Bluetooth: hci_qca: Add device_may_wakeup support") Cc: stable@vger.kernel.org # 5.13 Signed-off-by: Johan Hovold <johan+linaro@kernel.org> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add the support ID(0x0bda, 0x4853) to usb_device_id table for
Realtek RTL8852BE.
Without this change the device utilizes an obsolete version of
the firmware that is encoded in it rather than the updated Realtek
firmware and config files from the firmware directory. The latter
files implement many new features.
After an innocuous optimization change in LLVM main (19.0.0), x86_64
allmodconfig (which enables CONFIG_KCSAN / -fsanitize=thread) fails to
build due to the checks in check_copy_size():
In file included from net/bluetooth/sco.c:27:
In file included from include/linux/module.h:13:
In file included from include/linux/stat.h:19:
In file included from include/linux/time.h:60:
In file included from include/linux/time32.h:13:
In file included from include/linux/timex.h:67:
In file included from arch/x86/include/asm/timex.h:6:
In file included from arch/x86/include/asm/tsc.h:10:
In file included from arch/x86/include/asm/msr.h:15:
In file included from include/linux/percpu.h:7:
In file included from include/linux/smp.h:118:
include/linux/thread_info.h:244:4: error: call to '__bad_copy_from'
declared with 'error' attribute: copy source size is too small
244 | __bad_copy_from();
| ^
The same exact error occurs in l2cap_sock.c. The copy_to_user()
statements that are failing come from l2cap_sock_getsockopt_old() and
sco_sock_getsockopt_old(). This does not occur with GCC with or without
KCSAN or Clang without KCSAN enabled.
len is defined as an 'int' because it is assigned from
'__user int *optlen'. However, it is clamped against the result of
sizeof(), which has a type of 'size_t' ('unsigned long' for 64-bit
platforms). This is done with min_t() because min() requires compatible
types, which results in both len and the result of sizeof() being casted
to 'unsigned int', meaning len changes signs and the result of sizeof()
is truncated. From there, len is passed to copy_to_user(), which has a
third parameter type of 'unsigned long', so it is widened and changes
signs again. This excessive casting in combination with the KCSAN
instrumentation causes LLVM to fail to eliminate the __bad_copy_from()
call, failing the build.
The official recommendation from LLVM developers is to consistently use
long types for all size variables to avoid the unnecessary casting in
the first place. Change the type of len to size_t in both
l2cap_sock_getsockopt_old() and sco_sock_getsockopt_old(). This clears
up the error while allowing min_t() to be replaced with min(), resulting
in simpler code with no casts and fewer implicit conversions. While len
is a different type than optlen now, it should result in no functional
change because the result of sizeof() will clamp all values of optlen in
the same manner as before.
Remove argument `params` from the `module` macro example, because the
macro does not currently support module parameters since it was not sent
with the initial merge.
If one attempts to build an essentially empty file somewhere in the
kernel tree, it leads to a build error because the compiler does not
recognize the `new_uninit` unstable feature:
The reason is that we pass `-Zcrate-attr='feature(new_uninit)'` (together
with `-Zallow-features=new_uninit`) to let non-`rust/` code use that
unstable feature.
However, the compiler only recognizes the feature if the `alloc` crate
is resolved (the feature is an `alloc` one). `--extern alloc`, which we
pass, is not enough to resolve the crate.
Introducing a reference like `use alloc;` or `extern crate alloc;`
solves the issue, thus this is not seen in normal files. For instance,
`use`ing the `kernel` prelude introduces such a reference, since `alloc`
is used inside.
While normal use of the build system is not impacted by this, it can still
be fairly confusing for kernel developers [1], thus use the unstable
`force` option of `--extern` [2] (added in Rust 1.71 [3]) to force the
compiler to resolve `alloc`.
This new unstable feature is only needed meanwhile we use the other
unstable feature, since then we will not need `-Zcrate-attr`.
When KUnit tests are enabled, under very big kernel configurations
(e.g. `allyesconfig`), we can trigger a `rustdoc` ICE [1]:
RUSTDOC TK rust/kernel/lib.rs
error: the compiler unexpectedly panicked. this is a bug.
The reason is that this build step has a duplicated `@rustc_cfg` argument,
which contains the kernel configuration, and thus a lot of arguments. The
factor 2 happens to be enough to reach the ICE.
Thus remove the unneeded `@rustc_cfg`. By doing so, we clean up the
command and workaround the ICE.
The ICE has been fixed in the upcoming Rust 1.79 [2].
On RISC-V and arm64, and presumably x86, if CFI_CLANG is enabled,
loading a rust module will trigger a kernel panic. Support for
sanitisers, including kcfi (CFI_CLANG), is in the works, but for now
they're nightly-only options in rustc. Make RUST depend on !CFI_CLANG
to prevent configuring a kernel without symmetrical support for kfi.
[ Matthew Maurer writes [1]:
This patch is fine by me - the last patch needed for KCFI to be
functional in Rust just landed upstream last night, so we should
revisit this (in the form of enabling it) once we move to
`rustc-1.79.0` or later.
Ramon de C Valle also gave feedback [2] on the status of KCFI for
Rust and created a tracking issue [3] in upstream Rust. - Miguel ]
In Rust, producing an invalid value of any type is immediate undefined
behavior (UB); this includes via zeroing memory. Therefore, since an
uninhabited type has no valid values, producing any values at all for it is
UB.
The Rust standard library type `core::convert::Infallible` is uninhabited,
by virtue of having been declared as an enum with no cases, which always
produces uninhabited types in Rust.
The current kernel code allows this UB to be triggered, for example by code
like `Box::<core::convert::Infallible>::init(kernel::init::zeroed())`.
Thus, remove the implementation of `Zeroable` for `Infallible`, thereby
avoiding the unsoundness (potential for future UB).
Cc: stable@vger.kernel.org Fixes: 38cde0bd7b67 ("rust: init: add `Zeroable` trait and `init::zeroed` function") Closes: https://github.com/Rust-for-Linux/pinned-init/pull/13 Signed-off-by: Laine Taffin Altman <alexanderaltman@me.com> Reviewed-by: Alice Ryhl <aliceryhl@google.com> Reviewed-by: Boqun Feng <boqun.feng@gmail.com> Reviewed-by: Benno Lossin <benno.lossin@proton.me> Link: https://lore.kernel.org/r/CA160A4E-561E-4918-837E-3DCEBA74F808@me.com
[ Reformatted the comment slightly. ] Signed-off-by: Miguel Ojeda <ojeda@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This was originally part of commit 4b9a68f2e59a0 ("rust: add support for
static synchronisation primitives") from the old Rust branch, which used
module constructors to initialize globals containing various
synchronisation primitives with pin-init. That commit has never been
upstreamed, but the `select CONSTRUCTORS` statement ended up being
included in the patch that initially added Rust support to the Linux
Kernel.
We are not using module constructors, so let's remove the select.
cpu_feature_enabled(X86_FEATURE_OSPKE) does not necessarily reflect
whether CR4.PKE is set on the CPU. In particular, they may differ on
non-BSP CPUs before setup_pku() is executed. In this scenario, RDPKRU
will #UD causing the system to hang.
Fix by checking CR4 for PKE enablement which is always correct for the
current CPU.
The scenario happens by inserting a WARN* before setup_pku() in
identiy_cpu() or some other diagnostic which would lead to calling
__show_regs().
Thorvald reported a WARNING [1]. And the root cause is below race:
CPU 1 CPU 2
fork hugetlbfs_fallocate
dup_mmap hugetlbfs_punch_hole
i_mmap_lock_write(mapping);
vma_interval_tree_insert_after -- Child vma is visible through i_mmap tree.
i_mmap_unlock_write(mapping);
hugetlb_dup_vma_private -- Clear vma_lock outside i_mmap_rwsem!
i_mmap_lock_write(mapping);
hugetlb_vmdelete_list
vma_interval_tree_foreach
hugetlb_vma_trylock_write -- Vma_lock is cleared.
tmp->vm_ops->open -- Alloc new vma_lock outside i_mmap_rwsem!
hugetlb_vma_unlock_write -- Vma_lock is assigned!!!
i_mmap_unlock_write(mapping);
hugetlb_dup_vma_private() and hugetlb_vm_op_open() are called outside
i_mmap_rwsem lock while vma lock can be used in the same time. Fix this
by deferring linking file vma until vma is fully initialized. Those vmas
should be initialized first before they can be used.
Link: https://lkml.kernel.org/r/20240410091441.3539905-1-linmiaohe@huawei.com Fixes: 8d9bfb260814 ("hugetlb: add vma based lock for pmd sharing") Signed-off-by: Miaohe Lin <linmiaohe@huawei.com> Reported-by: Thorvald Natvig <thorvald@google.com> Closes: https://lore.kernel.org/linux-mm/20240129161735.6gmjsswx62o4pbja@revolver/T/ [1] Reviewed-by: Jane Chu <jane.chu@oracle.com> Cc: Christian Brauner <brauner@kernel.org> Cc: Heiko Carstens <hca@linux.ibm.com> Cc: Kent Overstreet <kent.overstreet@linux.dev> Cc: Liam R. Howlett <Liam.Howlett@oracle.com> Cc: Mateusz Guzik <mjguzik@gmail.com> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Miaohe Lin <linmiaohe@huawei.com> Cc: Muchun Song <muchun.song@linux.dev> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Peng Zhang <zhangpeng.00@bytedance.com> Cc: Tycho Andersen <tandersen@netflix.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The Bionic version of pthread_create used on Android calls the prctl
function to give the stack and thread local storage a useful name. This
will cause the KILL_THREAD test to fail as it will kill the thread as
soon as it is created.
Currently the user_notification_addfd test checks what the next expected
file descriptor will be by incrementing a variable nextfd. This does not
account for file descriptors that may already be open before the test is
started and will cause the test to fail if any exist.
Replace nextfd++ with a function get_next_fd which will check and return
the next available file descriptor.
Syskiller has produced an out of bounds access in fill_meta_index().
That out of bounds access is ultimately caused because the inode
has an inode number with the invalid value of zero, which was not checked.
The reason this causes the out of bounds access is due to following
sequence of events:
1. Fill_meta_index() is called to allocate (via empty_meta_index())
and fill a metadata index. It however suffers a data read error
and aborts, invalidating the newly returned empty metadata index.
It does this by setting the inode number of the index to zero,
which means unused (zero is not a valid inode number).
2. When fill_meta_index() is subsequently called again on another
read operation, locate_meta_index() returns the previous index
because it matches the inode number of 0. Because this index
has been returned it is expected to have been filled, and because
it hasn't been, an out of bounds access is performed.
This patch adds a sanity check which checks that the inode number
is not zero when the inode is created and returns -EINVAL if it is.
Convert to using the new inode timestamp accessor functions.
Signed-off-by: Jeff Layton <jlayton@kernel.org> Link: https://lore.kernel.org/r/20231004185347.80880-68-jlayton@kernel.org Signed-off-by: Christian Brauner <brauner@kernel.org>
Stable-dep-of: 9253c54e01b6 ("Squashfs: check the inode number is not the invalid value of zero") Signed-off-by: Sasha Levin <sashal@kernel.org>
We only pool write combined and uncached allocations because they
require extra overhead on allocation and release.
If we also pool cached NUMA it not only means some extra unnecessary
overhead, but also that under memory pressure it can happen that
pages from the wrong NUMA node enters the pool and are re-used
over and over again.
This can lead to performance reduction after running into memory
pressure.
v2: restructure and cleanup the code a bit from the internal hack to
test this.
Add shared stats. Useful for seeing shared memory.
v2: take dma-buf into account as well
v3: use the new gem helper
Link: https://lore.kernel.org/all/20231207180225.439482-1-alexander.deucher@amd.com/ Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Cc: Rob Clark <robdclark@gmail.com> Reviewed-by: Christian König <christian.keonig@amd.com> Signed-off-by: Christian König <christian.koenig@amd.com>
Stable-dep-of: a6ff969fe9cb ("drm/amdgpu: fix visible VRAM handling during faults") Signed-off-by: Sasha Levin <sashal@kernel.org>
Darrick reports that in some cases where pread() would fail with -EIO and
mmap()+access would generate a SIGBUS signal, MADV_POPULATE_READ /
MADV_POPULATE_WRITE will keep retrying forever and not fail with -EFAULT.
While the madvise() call can be interrupted by a signal, this is not the
desired behavior. MADV_POPULATE_READ / MADV_POPULATE_WRITE should behave
like page faults in that case: fail and not retry forever.
A reproducer can be found at [1].
The reason is that __get_user_pages(), as called by
faultin_vma_page_range(), will not handle VM_FAULT_RETRY in a proper way:
it will simply return 0 when VM_FAULT_RETRY happened, making
madvise_populate()->faultin_vma_page_range() retry again and again, never
setting FOLL_TRIED->FAULT_FLAG_TRIED for __get_user_pages().
__get_user_pages_locked() does what we want, but duplicating that logic in
faultin_vma_page_range() feels wrong.
So let's use __get_user_pages_locked() instead, that will detect
VM_FAULT_RETRY and set FOLL_TRIED when retrying, making the fault handler
return VM_FAULT_SIGBUS (VM_FAULT_ERROR) at some point, propagating -EFAULT
from faultin_page() to __get_user_pages(), all the way to
madvise_populate().
But, there is an issue: __get_user_pages_locked() will end up re-taking
the MM lock and then __get_user_pages() will do another VMA lookup. In
the meantime, the VMA layout could have changed and we'd fail with
different error codes than we'd want to.
As __get_user_pages() will currently do a new VMA lookup either way, let
it do the VMA handling in a different way, controlled by a new
FOLL_MADV_POPULATE flag, effectively moving these checks from
madvise_populate() + faultin_page_range() in there.
With this change, Darricks reproducer properly fails with -EFAULT, as
documented for MADV_POPULATE_READ / MADV_POPULATE_WRITE.
Set the enable bits for general purpose counters in IA32_PERF_GLOBAL_CTRL
when refreshing the PMU to emulate the MSR's architecturally defined
post-RESET behavior. Per Intel's SDM:
IA32_PERF_GLOBAL_CTRL: Sets bits n-1:0 and clears the upper bits.
and
Where "n" is the number of general-purpose counters available in the processor.
AMD also documents this behavior for PerfMonV2 CPUs in one of AMD's many
PPRs.
Do not set any PERF_GLOBAL_CTRL bits if there are no general purpose
counters, although a literal reading of the SDM would require the CPU to
set either bits 63:0 or 31:0. The intent of the behavior is to globally
enable all GP counters; honor the intent, if not the letter of the law.
Leaving PERF_GLOBAL_CTRL '0' effectively breaks PMU usage in guests that
haven't been updated to work with PMUs that support PERF_GLOBAL_CTRL.
This bug was recently exposed when KVM added supported for AMD's
PerfMonV2, i.e. when KVM started exposing a vPMU with PERF_GLOBAL_CTRL to
guest software that only knew how to program v1 PMUs (that don't support
PERF_GLOBAL_CTRL).
Failure to emulate the post-RESET behavior results in such guests
unknowingly leaving all general purpose counters globally disabled (the
entire reason the post-RESET value sets the GP counter enable bits is to
maintain backwards compatibility).
The bug has likely gone unnoticed because PERF_GLOBAL_CTRL has been
supported on Intel CPUs for as long as KVM has existed, i.e. hardly anyone
is running guest software that isn't aware of PERF_GLOBAL_CTRL on Intel
PMUs. And because up until v6.0, KVM _did_ emulate the behavior for Intel
CPUs, although the old behavior was likely dumb luck.
Because (a) that old code was also broken in its own way (the history of
this code is a comedy of errors), and (b) PERF_GLOBAL_CTRL was documented
as having a value of '0' post-RESET in all SDMs before March 2023.
Initial vPMU support in commit f5132b01386b ("KVM: Expose a version 2
architectural PMU to a guests") *almost* got it right (again likely by
dumb luck), but for some reason only set the bits if the guest PMU was
advertised as v1:
Commit f19a0c2c2e6a ("KVM: PMU emulation: GLOBAL_CTRL MSR should be
enabled on reset") then tried to remedy that goof, presumably because
guest PMUs were leaving PERF_GLOBAL_CTRL '0', i.e. weren't enabling
counters.
That was KVM's behavior up until commit c49467a45fe0 ("KVM: x86/pmu:
Don't overwrite the pmu->global_ctrl when refreshing") removed
*everything*. However, it did so based on the behavior defined by the
SDM , which at the time stated that "Global Perf Counter Controls" is
'0' at Power-Up and RESET.
But then the March 2023 SDM (325462-079US), stealthily changed its
"IA-32 and Intel 64 Processor States Following Power-up, Reset, or INIT"
table to say:
IA32_PERF_GLOBAL_CTRL: Sets bits n-1:0 and clears the upper bits.
Note, kvm_pmu_refresh() can be invoked multiple times, i.e. it's not a
"pure" RESET flow. But it can only be called prior to the first KVM_RUN,
i.e. the guest will only ever observe the final value.
Note #2, KVM has always cleared global_ctrl during refresh (see commit f5132b01386b ("KVM: Expose a version 2 architectural PMU to a guests")),
i.e. there is no danger of breaking existing setups by clobbering a value
set by userspace.
Reported-by: Babu Moger <babu.moger@amd.com> Cc: Sandipan Das <sandipan.das@amd.com> Cc: Like Xu <like.xu.linux@gmail.com> Cc: Mingwei Zhang <mizhang@google.com> Cc: Dapeng Mi <dapeng1.mi@linux.intel.com> Cc: stable@vger.kernel.org Reviewed-by: Dapeng Mi <dapeng1.mi@linux.intel.com> Tested-by: Dapeng Mi <dapeng1.mi@linux.intel.com> Link: https://lore.kernel.org/r/20240309013641.1413400-2-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Move the purging of common PMU metadata from intel_pmu_refresh() to
kvm_pmu_refresh(), and invoke the vendor refresh() hook if and only if
the VM is supposed to have a vPMU.
KVM already denies access to the PMU based on kvm->arch.enable_pmu, as
get_gp_pmc_amd() returns NULL for all PMCs in that case, i.e. KVM already
violates AMD's architecture by not virtualizing a PMU (kernels have long
since learned to not panic when the PMU is unavailable). But configuring
the PMU as if it were enabled causes unwanted side effects, e.g. calls to
kvm_pmu_trigger_event() waste an absurd number of cycles due to the
all_valid_pmc_idx bitmap being non-zero.
Fixes: b1d66dad65dc ("KVM: x86/svm: Add module param to control PMU virtualization") Reported-by: Konstantin Khorenko <khorenko@virtuozzo.com> Closes: https://lore.kernel.org/all/20231109180646.2963718-2-khorenko@virtuozzo.com Link: https://lore.kernel.org/r/20231110022857.1273836-2-seanjc@google.com Signed-off-by: Sean Christopherson <seanjc@google.com>
Stable-dep-of: de120e1d692d ("KVM: x86/pmu: Set enable bits for GP counters in PERF_GLOBAL_CTRL at "RESET"") Signed-off-by: Sasha Levin <sashal@kernel.org>
tls_sk_poll is called without locking the socket, and needs to read
strp->msg_ready (via tls_strp_msg_ready). Convert msg_ready to a bool
and use READ_ONCE/WRITE_ONCE where needed. The remaining reads are
only performed when the socket is locked.
The CPTS, by design, captures the messageType (Sync, Delay_Req, etc.)
field from the second nibble of the PTP header which is defined in the
PTPv2 (1588-2008) specification. In the PTPv1 (1588-2002) specification
the first two bytes of the PTP header are defined as the versionType
which is always 0x0001. This means that any PTPv1 packets that are
tagged for TX timestamping by the CPTS will have their messageType set
to 0x0 which corresponds to a Sync message type. This causes issues
when a PTPv1 stack is expecting a Delay_Req (messageType: 0x1)
timestamp that never appears.
Fix this by checking if the ptp_class of the timestamped TX packet is
PTP_CLASS_V1 and then matching the PTP sequence ID to the stored
sequence ID in the skb->cb data structure. If the sequence IDs match
and the packet is of type PTPv1 then there is a chance that the
messageType has been incorrectly stored by the CPTS so overwrite the
messageType stored by the CPTS with the messageType from the skb->cb
data structure. This allows the PTPv1 stack to receive TX timestamps
for Delay_Req packets which are necessary to lock onto a PTP Leader.
Signed-off-by: Jason Reeder <jreeder@ti.com> Signed-off-by: Ravi Gunasekaran <r-gunasekaran@ti.com> Tested-by: Ed Trexel <ed.trexel@hp.com> Fixes: f6bd59526ca5 ("net: ethernet: ti: introduce am654 common platform time sync driver") Link: https://lore.kernel.org/r/20240424071626.32558-1-r-gunasekaran@ti.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
9f74a3dfcf83 ("ice: Fix VF Reset paths when interface in a failed over
aggregate"), the ice driver has acquired the LAG mutex in ice_reset_vf().
The commit placed this lock acquisition just prior to the acquisition of
the VF configuration lock.
If ice_reset_vf() acquires the configuration lock via the ICE_VF_RESET_LOCK
flag, this could deadlock with ice_vc_cfg_qs_msg() because it always
acquires the locks in the order of the VF configuration lock and then the
LAG mutex.
Lockdep reports this violation almost immediately on creating and then
removing 2 VF:
======================================================
WARNING: possible circular locking dependency detected
6.8.0-rc6 #54 Tainted: G W O
------------------------------------------------------
kworker/60:3/6771 is trying to acquire lock: ff40d43e099380a0 (&vf->cfg_lock){+.+.}-{3:3}, at: ice_reset_vf+0x22f/0x4d0 [ice]
but task is already holding lock: ff40d43ea1961210 (&pf->lag_mutex){+.+.}-{3:3}, at: ice_reset_vf+0xb7/0x4d0 [ice]
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&pf->lag_mutex);
lock(&vf->cfg_lock);
lock(&pf->lag_mutex);
lock(&vf->cfg_lock);
To avoid deadlock, we must acquire the LAG mutex only after acquiring the
VF configuration lock. Fix the ice_reset_vf() to acquire the LAG mutex only
after we either acquire or check that the VF configuration lock is held.
Fixes: 9f74a3dfcf83 ("ice: Fix VF Reset paths when interface in a failed over aggregate") Signed-off-by: Jacob Keller <jacob.e.keller@intel.com> Reviewed-by: Dave Ertman <david.m.ertman@intel.com> Reviewed-by: Mateusz Polchlopek <mateusz.polchlopek@intel.com> Tested-by: Przemek Kitszel <przemyslaw.kitszel@intel.com> Tested-by: Rafal Romanowski <rafal.romanowski@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Link: https://lore.kernel.org/r/20240423182723.740401-5-anthony.l.nguyen@intel.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Same number of TCs doesn't imply that underlying TC configs are
same. The config could be different due to difference in number
of queues in each TC. Add utility function to determine if TC
configs are same.
Fixes: d5b33d024496 ("i40evf: add ndo_setup_tc callback to i40evf") Signed-off-by: Sudheer Mogilappagari <sudheer.mogilappagari@intel.com> Tested-by: Mineri Bhange <minerix.bhange@intel.com> (A Contingent Worker at Intel) Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Link: https://lore.kernel.org/r/20240423182723.740401-4-anthony.l.nguyen@intel.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
If the MFS is set below the default (0x2600), a warning message is
reported like the following :
MFS for port 1 has been set below the default: 600
This message is a bit confusing as the number shown here (600) is in
fact an hexa number: 0x600 = 1536
Without any explicit "0x" prefix, this message is read like the MFS is
set to 600 bytes.
MFS, as per MTUs, are usually expressed in decimal base.
This commit reports both current and default MFS values in decimal
so it's less confusing for end-users.
A typical warning message looks like the following :
MFS for port 1 (1536) has been set below the default (9728)
Signed-off-by: Erwan Velu <e.velu@criteo.com> Reviewed-by: Simon Horman <horms@kernel.org> Tested-by: Tony Brelinski <tony.brelinski@intel.com> Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com> Fixes: 3a2c6ced90e1 ("i40e: Add a check to see if MFS is set") Link: https://lore.kernel.org/r/20240423182723.740401-3-anthony.l.nguyen@intel.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Issue reported by customer during SRIOV testing, call trace:
When both i40e and the i40iw driver are loaded, a warning
in check_flush_dependency is being triggered. This seems
to be because of the i40e driver workqueue is allocated with
the WQ_MEM_RECLAIM flag, and the i40iw one is not.
Similar error was encountered on ice too and it was fixed by
removing the flag. Do the same for i40e too.
The rx_chn->irq[] array is unsigned int but it should be signed for the
error handling to work. Also if k3_udma_glue_rx_get_irq() returns zero
then we should return -ENXIO instead of success.
The DP83869 driver sets the MII bit (needed for PHY to work in MII mode)
only if the op-mode is either DP83869_100M_MEDIA_CONVERT or
DP83869_RGMII_100_BASE.
Some drivers i.e. ICSSG support MII mode with op-mode as
DP83869_RGMII_COPPER_ETHERNET for which the MII bit is not set in dp83869
driver. As a result MII mode on ICSSG doesn't work and below log is seen.
TI DP83869 300b2400.mdio:0f: selected op-mode is not valid with MII mode
icssg-prueth icssg1-eth: couldn't connect to phy ethernet-phy@0
icssg-prueth icssg1-eth: can't phy connect port MII0
Fix this by setting MII bit for DP83869_RGMII_COPPER_ETHERNET op-mode as
well.
Fixes: 94e86ef1b801 ("net: phy: dp83869: support mii mode when rgmii strap cfg is used") Signed-off-by: MD Danish Anwar <danishanwar@ti.com> Reviewed-by: Ravi Gunasekaran <r-gunasekaran@ti.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Fixes: d54725cd11a5 ("netfilter: nf_tables: support for multiple devices per netdev hook") Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
I added OOM and netpoll discard counters, naively assuming that
the cpr pointer is pointing to a common completion ring.
Turns out that is usually *a* completion ring but not *the*
completion ring which bnapi->cp_ring points to. bnapi->cp_ring
is where the stats are read from, so we end up reporting 0
thru ethtool -S and qstat even though the drop events have happened.
Make 100% sure we're recording statistics in the correct structure.
Fixes: 907fd4a294db ("bnxt: count discards due to memory allocation errors") Reviewed-by: Michael Chan <michael.chan@broadcom.com> Link: https://lore.kernel.org/r/20240424002148.3937059-1-kuba@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The rehash delayed work is rescheduled with a delay if the number of
credits at end of the work is not negative as supposedly it means that
the migration ended. Otherwise, it is rescheduled immediately.
After "mlxsw: spectrum_acl_tcam: Fix possible use-after-free during
rehash" the above is no longer accurate as a non-negative number of
credits is no longer indicative of the migration being done. It can also
happen if the work encountered an error in which case the migration will
resume the next time the work is scheduled.
The significance of the above is that it is possible for the work to be
pending and associated with hints that were allocated when the migration
started. This leads to the hints being leaked [1] when the work is
canceled while pending as part of ACL region dismantle.
Fix by freeing the hints if hints are associated with a work that was
canceled while pending.
Blame the original commit since the reliance on not having a pending
work associated with hints is fragile.
Both the function that migrates all the chunks within a region and the
function that migrates all the entries within a chunk call
list_first_entry() on the respective lists without checking that the
lists are not empty. This is incorrect usage of the API, which leads to
the following warning [1].
Fix by returning if the lists are empty as there is nothing to migrate
in this case.
As previously explained, the rehash delayed work migrates filters from
one region to another. This is done by iterating over all chunks (all
the filters with the same priority) in the region and in each chunk
iterating over all the filters.
When the work runs out of credits it stores the current chunk and entry
as markers in the per-work context so that it would know where to resume
the migration from the next time the work is scheduled.
Upon error, the chunk marker is reset to NULL, but without resetting the
entry markers despite being relative to it. This can result in migration
being resumed from an entry that does not belong to the chunk being
migrated. In turn, this will eventually lead to a chunk being iterated
over as if it is an entry. Because of how the two structures happen to
be defined, this does not lead to KASAN splats, but to warnings such as
[1].
Fix by creating a helper that resets all the markers and call it from
all the places the currently only reset the chunk marker. For good
measures also call it when starting a completely new rehash. Add a
warning to avoid future cases.
The rehash delayed work migrates filters from one region to another.
This is done by iterating over all chunks (all the filters with the same
priority) in the region and in each chunk iterating over all the
filters.
If the migration fails, the code tries to migrate the filters back to
the old region. However, the rollback itself can also fail in which case
another migration will be erroneously performed. Besides the fact that
this ping pong is not a very good idea, it also creates a problem.
Each virtual chunk references two chunks: The currently used one
('vchunk->chunk') and a backup ('vchunk->chunk2'). During migration the
first holds the chunk we want to migrate filters to and the second holds
the chunk we are migrating filters from.
The code currently assumes - but does not verify - that the backup chunk
does not exist (NULL) if the currently used chunk does not reference the
target region. This assumption breaks when we are trying to rollback a
rollback, resulting in the backup chunk being overwritten and leaked
[1].
Fix by not rolling back a failed rollback and add a warning to avoid
future cases.
Fixes: 843500518509 ("mlxsw: spectrum_acl: Do rollback as another call to mlxsw_sp_acl_tcam_vchunk_migrate_all()") Signed-off-by: Ido Schimmel <idosch@nvidia.com> Tested-by: Alexander Zubkov <green@qrator.net> Reviewed-by: Petr Machata <petrm@nvidia.com> Signed-off-by: Petr Machata <petrm@nvidia.com> Reviewed-by: Simon Horman <horms@kernel.org> Link: https://lore.kernel.org/r/d5edd4f4503934186ae5cfe268503b16345b4e0f.1713797103.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In the rare cases when the device resources are exhausted it is likely
that the rehash delayed work will fail. An error message will be printed
whenever this happens which can be overwhelming considering the fact
that the work is per-region and that there can be hundreds of regions.
Fix by rate limiting the error message.
Fixes: e5e7962ee5c2 ("mlxsw: spectrum_acl: Implement region migration according to hints") Signed-off-by: Ido Schimmel <idosch@nvidia.com> Tested-by: Alexander Zubkov <green@qrator.net> Reviewed-by: Petr Machata <petrm@nvidia.com> Signed-off-by: Petr Machata <petrm@nvidia.com> Reviewed-by: Simon Horman <horms@kernel.org> Link: https://lore.kernel.org/r/c510763b2ebd25e7990d80183feff91cde593145.1713797103.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The rehash delayed work migrates filters from one region to another
according to the number of available credits.
The migrated from region is destroyed at the end of the work if the
number of credits is non-negative as the assumption is that this is
indicative of migration being complete. This assumption is incorrect as
a non-negative number of credits can also be the result of a failed
migration.
The destruction of a region that still has filters referencing it can
result in a use-after-free [1].
Fix by not destroying the region if migration failed.
[1]
BUG: KASAN: slab-use-after-free in mlxsw_sp_acl_ctcam_region_entry_remove+0x21d/0x230
Read of size 8 at addr ffff8881735319e8 by task kworker/0:31/3858
The rule activity update delayed work periodically traverses the list of
configured rules and queries their activity from the device.
As part of this task it accesses the entry pointed by 'ventry->entry',
but this entry can be changed concurrently by the rehash delayed work,
leading to a use-after-free [1].
Fix by closing the race and perform the activity query under the
'vregion->lock' mutex.
[1]
BUG: KASAN: slab-use-after-free in mlxsw_sp_acl_tcam_flower_rule_activity_get+0x121/0x140
Read of size 8 at addr ffff8881054ed808 by task kworker/0:18/181
The purpose of the rehash delayed work is to reduce the number of masks
(eRPs) used by an ACL region as the eRP bank is a global and limited
resource.
This is done in three steps:
1. Creating a new set of masks and a new ACL region which will use the
new masks and to which the existing filters will be migrated to. The
new region is assigned to 'vregion->region' and the region from which
the filters are migrated from is assigned to 'vregion->region2'.
2. Migrating all the filters from the old region to the new region.
3. Destroying the old region and setting 'vregion->region2' to NULL.
Only the second steps is performed under the 'vregion->lock' mutex
although its comments says that among other things it "Protects
consistency of region, region2 pointers".
This is problematic as the first step can race with filter insertion
from user space that uses 'vregion->region', but under the mutex.
Fix by holding the mutex across the entirety of the delayed work and not
only during the second step.
Fixes: 2bffc5322fd8 ("mlxsw: spectrum_acl: Don't take mutex in mlxsw_sp_acl_tcam_vregion_rehash_work()") Signed-off-by: Ido Schimmel <idosch@nvidia.com> Tested-by: Alexander Zubkov <green@qrator.net> Reviewed-by: Petr Machata <petrm@nvidia.com> Signed-off-by: Petr Machata <petrm@nvidia.com> Reviewed-by: Simon Horman <horms@kernel.org> Link: https://lore.kernel.org/r/1ec1d54edf2bad0a369e6b4fa030aba64e1f124b.1713797103.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Region identifiers can be allocated both when user space tries to insert
a new tc filter and when filters are migrated from one region to another
as part of the rehash delayed work.
There is no lock protecting the bitmap from which these identifiers are
allocated from, which is racy and leads to bad parameter errors from the
device's firmware.
Fix by converting the bitmap to IDA which handles its own locking. For
consistency, do the same for the group identifiers that are part of the
same structure.
Fixes: 2bffc5322fd8 ("mlxsw: spectrum_acl: Don't take mutex in mlxsw_sp_acl_tcam_vregion_rehash_work()") Reported-by: Amit Cohen <amcohen@nvidia.com> Signed-off-by: Ido Schimmel <idosch@nvidia.com> Tested-by: Alexander Zubkov <green@qrator.net> Reviewed-by: Petr Machata <petrm@nvidia.com> Signed-off-by: Petr Machata <petrm@nvidia.com> Reviewed-by: Simon Horman <horms@kernel.org> Link: https://lore.kernel.org/r/ce494b7940cadfe84f3e18da7785b51ef5f776e3.1713797103.git.petrm@nvidia.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
mlxsw driver uses 'unsigned int' for reference counters in several
structures. Instead, use refcount_t type which allows us to catch overflow
and underflow issues. Change the type of the counters and use the
appropriate API.
Signed-off-by: Amit Cohen <amcohen@nvidia.com> Reviewed-by: Ido Schimmel <idosch@nvidia.com> Signed-off-by: Ido Schimmel <idosch@nvidia.com> Signed-off-by: Petr Machata <petrm@nvidia.com> Reviewed-by: Simon Horman <horms@kernel.org> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Stable-dep-of: 627f9c1bb882 ("mlxsw: spectrum_acl_tcam: Fix race in region ID allocation") Signed-off-by: Sasha Levin <sashal@kernel.org>
Since kfree_rcu, which is called in the hlist_for_each_entry_rcu traversal
of ovs_ct_limit_exit, is not part of the RCU read critical section, it
is possible that the RCU grace period will pass during the traversal and
the key will be free.
To prevent this, it should be changed to hlist_for_each_entry_safe.
Fixes: 11efd5cb04a1 ("openvswitch: Support conntrack zone limit") Signed-off-by: Hyunwoo Kim <v4bel@theori.io> Reviewed-by: Eric Dumazet <edumazet@google.com> Reviewed-by: Aaron Conole <aconole@redhat.com> Link: https://lore.kernel.org/r/ZiYvzQN/Ry5oeFQW@v4bel-B760M-AORUS-ELITE-AX Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
It was observed in the wild that pairs of consecutive packets would leave
the IPVS with the same wrong checksum, and the issue only went away when
disabling GSO.
IPVS needs to avoid computing the SCTP checksum when using GSO.
Any return value from gpiod_get_optional() other than a pointer to a
GPIO descriptor or a NULL-pointer is an error and the driver should
abort probing. That being said: commit 56d074d26c58 ("Bluetooth: hci_qca:
don't use IS_ERR_OR_NULL() with gpiod_get_optional()") no longer sets
power_ctrl_enabled on NULL-pointer returned by
devm_gpiod_get_optional(). Restore this behavior but bail-out on errors.
While at it: also bail-out on error returned when trying to get the
"swctrl" GPIO.
Reported-by: Wren Turkal <wt@penguintechs.org> Reported-by: Zijun Hu <quic_zijuhu@quicinc.com> Closes: https://lore.kernel.org/linux-bluetooth/1713449192-25926-2-git-send-email-quic_zijuhu@quicinc.com/ Fixes: 56d074d26c58 ("Bluetooth: hci_qca: don't use IS_ERR_OR_NULL() with gpiod_get_optional()") Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> Tested-by: Wren Turkal <wt@penguintechs.org> Reported-by: Wren Turkal <wt@penguintechs.org> Reported-by: Zijun Hu <quic_zijuhu@quicinc.com> Reviewed-by: Krzysztof Kozlowski<krzysztof.kozlowski@linaro.org> Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Since the d883a4669a1de be introduced in v6.4, bluetooth daemon
got the following failed message of MGMT_OP_REMOVE_ADV_MONITOR
command when controller is power-off:
bluetoothd[20976]:
src/adapter.c:reset_adv_monitors_complete() Failed to reset Adv
Monitors: Failed>
Normally this situation is happened when the bluetoothd deamon
be started manually after system booting. Which means that
bluetoothd received MGMT_EV_INDEX_ADDED event after kernel
runs hci_power_off().
Base on doc/mgmt-api.txt, the MGMT_OP_REMOVE_ADV_MONITOR command
can be used when the controller is not powered. This patch changes
the code in remove_adv_monitor() to use hci_cmd_sync_submit()
instead of hci_cmd_sync_queue().
Fixes: d883a4669a1de ("Bluetooth: hci_sync: Only allow hci_cmd_sync_queue if running") Cc: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Cc: Manish Mandlik <mmandlik@google.com> Cc: Archie Pusaka <apusaka@chromium.org> Cc: Miao-chen Chou <mcchou@chromium.org> Signed-off-by: Chun-Yi Lee <jlee@suse.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
These commands don't require the adapter to be up and running so don't
use hci_cmd_sync_queue which would check that flag, instead use
hci_cmd_sync_submit which would ensure mgmt_class_complete is set
properly regardless if any command was actually run or not.
Link: https://github.com/bluez/bluez/issues/809 Fixes: d883a4669a1d ("Bluetooth: hci_sync: Only allow hci_cmd_sync_queue if running") Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The code shall always check if HCI_QUIRK_BROKEN_READ_ENC_KEY_SIZE has
been set before attempting to use HCI_OP_READ_ENC_KEY_SIZE.
Fixes: c569242cd492 ("Bluetooth: hci_event: set the conn encrypted before conn establishes") Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
btusb_coredump_qca() uses __hci_cmd_sync() to send a vendor-specific
command to trigger firmware coredump, but the command does not
have any event as its sync response, so it is not suitable to use
__hci_cmd_sync(), fixed by using __hci_cmd_send().
Fixes: 20981ce2d5a5 ("Bluetooth: btusb: Add WCN6855 devcoredump support") Signed-off-by: Zijun Hu <quic_zijuhu@quicinc.com> Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The controller has several register bits describing access control
information for a given GPIO pin. When SCR_SEC_[R|W]EN is unset, it
means we have full read/write access to all the registers for given GPIO
pin. When SCR_SEC[R|W]EN is set, it means we need to further check the
accompanying SCR_SEC_G1[R|W] bit to determine read/write access to all
the registers for given GPIO pin.
This check was previously declaring that a GPIO pin was accessible
only if either of the following conditions were met:
- SCR_SEC_REN + SCR_SEC_WEN both set
or
- SCR_SEC_REN + SCR_SEC_WEN both set and
SCR_SEC_G1R + SCR_SEC_G1W both set
Update the check to properly handle cases where only one of
SCR_SEC_REN or SCR_SEC_WEN is set.
Only blink if the link is up on a LED which is programmed to also
indicate link-status.
Otherwise, if both LEDs are in use to indicate different speeds, the
resulting blinking being inverted on LEDs which aren't switched on at
a specific speed is quite counter-intuitive.
Also make sure that state left behind by reset or the bootloader is
recognized correctly including the half-duplex and full-duplex bits as
well as the (unsupported by Linux netdev trigger semantics) link-down
bit.
Fixes: c66937b0f8db ("net: phy: mediatek-ge-soc: support PHY LEDs") Signed-off-by: Daniel Golle <daniel@makrotopia.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Since call_rcu, which is called in the hlist_for_each_entry_rcu traversal
of gtp_dellink, is not part of the RCU read critical section, it
is possible that the RCU grace period will pass during the traversal and
the key will be free.
To prevent this, it should be changed to hlist_for_each_entry_safe.
Fixes: 94dc550a5062 ("gtp: fix an use-after-free in ipv4_pdp_find()") Signed-off-by: Hyunwoo Kim <v4bel@theori.io> Reviewed-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
Some usb drivers try to set small skb->truesize and break
core networking stacks.
In this patch, I removed one of the skb->truesize overide.
I also replaced one skb_clone() by an allocation of a fresh
and small skb, to get minimally sized skbs, like we did
in commit 1e2c61172342 ("net: cdc_ncm: reduce skb truesize
in rx path")
Fixes: f8ebb3ac881b ("net: usb: ax88179_178a: Fix packet receiving") Reported-by: shironeko <shironeko@tesaguri.club> Closes: https://lore.kernel.org/netdev/c110f41a0d2776b525930f213ca9715c@tesaguri.club/ Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Jose Alonso <joalonsof@gmail.com> Reviewed-by: Simon Horman <horms@kernel.org> Link: https://lore.kernel.org/r/20240421193828.1966195-1-edumazet@google.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Fixes: 02b24941619f ("ipv4: use dst hint for ipv4 list receive") Reported-by: syzbot <syzkaller@googlegroups.com> Signed-off-by: Eric Dumazet <edumazet@google.com> Acked-by: Paolo Abeni <pabeni@redhat.com> Link: https://lore.kernel.org/r/20240421184326.1704930-1-edumazet@google.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Jonathan Heathcote reported a regression caused by blamed commit
on aarch64 architecture.
x86 happens to have irq-safe __this_cpu_add_return()
and __this_cpu_sub(), but this is not generic.
I think my confusion came from "struct sock" argument,
because these helpers are called with a locked socket.
But the memory accounting is per-proto (and per-cpu after
the blamed commit). We might cleanup these helpers later
to directly accept a "struct proto *proto" argument.
Switch to this_cpu_add_return() and this_cpu_xchg()
operations, and get rid of preempt_disable()/preempt_enable() pairs.
Fast path becomes a bit faster as a result :)
Many thanks to Jonathan Heathcote for his awesome report and
investigations.
Fixes: 3cd3399dd7a8 ("net: implement per-cpu reserves for memory_allocated") Reported-by: Jonathan Heathcote <jonathan.heathcote@bbc.co.uk> Closes: https://lore.kernel.org/netdev/VI1PR01MB42407D7947B2EA448F1E04EFD10D2@VI1PR01MB4240.eurprd01.prod.exchangelabs.com/ Signed-off-by: Eric Dumazet <edumazet@google.com> Acked-by: Soheil Hassas Yeganeh <soheil@google.com> Reviewed-by: Shakeel Butt <shakeel.butt@linux.dev> Link: https://lore.kernel.org/r/20240421175248.1692552-1-edumazet@google.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
"Implement a per-cpu cache of +1/-1 MB, to reduce number
of changes to sk->sk_prot->memory_allocated, which
would otherwise be cause of false sharing."
sk_prot->memory_allocated points to global atomic variable:
atomic_long_t tcp_memory_allocated ____cacheline_aligned_in_smp;
If increasing the per-cpu cache size from 1MB to e.g. 16MB,
changes to sk->sk_prot->memory_allocated can be further reduced.
Performance may be improved on system with many cores.
Signed-off-by: Adam Li <adamli@os.amperecomputing.com> Reviewed-by: Christoph Lameter (Ampere) <cl@linux.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Stable-dep-of: 3584718cf2ec ("net: fix sk_memory_allocated_{add|sub} vs softirqs") Signed-off-by: Sasha Levin <sashal@kernel.org>
NLMSG_DONE contains an error code, it has to be extracted.
Prior to this change all dumps will end in success,
and in case of failure the result is silently truncated.
Fixes: e4b48ed460d3 ("tools: ynl: add a completely generic client") Signed-off-by: Jakub Kicinski <kuba@kernel.org> Reviewed-by: Donald Hunter <donald.hunter@gmail.com> Link: https://lore.kernel.org/r/20240420020827.3288615-1-kuba@kernel.org Signed-off-by: Paolo Abeni <pabeni@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The dev_tracker is added to ax25_cb in ax25_bind(). When the
ax25 device is detaching, the dev_tracker of ax25_cb should be
deallocated in ax25_kill_by_device() instead of the dev_tracker
of ax25_dev. The log reported by ref_tracker is shown below:
During module probe, regulator 'vin' and 'vdd-io' are used and enabled,
but the vdd-io regulator overwrites the 'vin' regulator pointer. During
remove, only the vdd-io is disabled, as the vin regulator pointer is not
available anymore. When regulator_put() is called during resource
cleanup a kernel warning is given, as the regulator is still enabled.
Store the two regulators in separate pointers and disable both the
regulators on module remove.
With the recent PHYLINK changes requiring supported_interfaces to be set,
MV88E6250 family switches like the 88E6020 fail to probe - cmode is
never initialized on these devices, so mv88e6250_phylink_get_caps() does
not set any supported_interfaces flags.
Instead of a cmode, on 88E6250 we have a read-only port mode value that
encodes similar information. There is no reason to bother mapping port
mode to the cmodes of other switch models; instead we introduce a
mv88e6250_setup_supported_interfaces() that is called directly from
mv88e6250_phylink_get_caps().
Fixes: de5c9bf40c45 ("net: phylink: require supported_interfaces to be filled") Signed-off-by: Matthias Schiffer <matthias.schiffer@ew.tq-group.com> Link: https://lore.kernel.org/r/20240417103737.166651-1-matthias.schiffer@ew.tq-group.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
A recent change to cxl_mem_get_records_log() [1] highlighted a subtle
nuance of looping calls to cxl_internal_send_cmd(), i.e. that
cxl_internal_send_cmd() modifies the 'size_out' member of the @mbox_cmd
argument. That mechanism is useful for communicating underflow, but it
is unwanted when reusing @mbox_cmd for a subsequent submission. It turns
out that cxl_xfer_log() avoids this scenario by always redefining
@mbox_cmd each iteration.
Update cxl_mem_get_records_log() and cxl_mem_get_poison() to follow the
same style as cxl_xfer_log(), i.e. re-define @mbox_cmd each iteration.
The cxl_mem_get_records_log() change is just a style fixup, but the
cxl_mem_get_poison() change is a potential fix, per Alison [2]:
Poison list retrieval can hit this case if the MORE flag is set and
a follow on read of the list delivers more records than the previous
read. ie. device gives one record, sets the _MORE flag, then gives 5.
Not an urgent fix since this behavior has not been seen in the wild,
but worth tracking as a fix.
We do not support two simultaneous recoveries so check for reset
flag, BNXT_STATE_IN_FW_RESET, and do not proceed with AER further.
When the pci channel state is pci_channel_io_frozen, the PCIe link
can not be trusted so we disable the traffic immediately and stop
BAR access by calling bnxt_fw_fatal_close(). BAR access after
AER fatal error can cause an NMI.
Fixes: f75d9a0aa967 ("bnxt_en: Re-write PCI BARs after PCI fatal error.") Signed-off-by: Vikas Gupta <vikas.gupta@broadcom.com> Signed-off-by: Michael Chan <michael.chan@broadcom.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sashal@kernel.org>
projects / thirdparty / kernel / stable.git / log
