'minItems' alone does not impose upper bound, unlike 'maxItems' which
implies lower bound. Add missing clock constraint so the list will have
exact number of items (clocks).
The dfa blob stream for the aa_dfa_unpack() function is expected to be aligned
on a 8 byte boundary.
The static nulldfa_src[] and stacksplitdfa_src[] arrays store the initial
apparmor dfa blob streams, but since they are declared as an array-of-chars
the compiler and linker will only ensure a "char" (1-byte) alignment.
Add an __aligned(8) annotation to the arrays to tell the linker to always
align them on a 8-byte boundary. This avoids runtime warnings at startup on
alignment-sensitive platforms like parisc such as:
Kernel: unaligned access to 0x7f2a584a in aa_dfa_unpack+0x124/0x788 (iir 0xca0109f)
Kernel: unaligned access to 0x7f2a584e in aa_dfa_unpack+0x210/0x788 (iir 0xca8109c)
Kernel: unaligned access to 0x7f2a586a in aa_dfa_unpack+0x278/0x788 (iir 0xcb01090)
Signed-off-by: Helge Deller <deller@gmx.de> Cc: stable@vger.kernel.org Fixes: 98b824ff8984 ("apparmor: refcount the pdb") Signed-off-by: John Johansen <john.johansen@canonical.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Enable internal bias pull-ups on the SoC-side I2C buses that do not have
external pull resistors populated on the SoM. This ensures proper
default line levels.
main_uart1 reserved for TIFS firmware traces is routed to the
onboard FT4232 via a FET switch which is connected to pin A21 and
B21 of the SoC and not E17 and C17. Fix it.
ufs-exynos driver configures the sysreg shareability as
cacheable for gs101 so we need to set the dma-coherent
property so the descriptors are also allocated cacheable.
This fixes the UFS stability issues we have seen with
the upstream UFS driver on gs101.
Fixes: 4c65d7054b4c ("arm64: dts: exynos: gs101: Add ufs and ufs-phy dt nodes") Cc: stable@vger.kernel.org Suggested-by: Will McVicker <willmcvicker@google.com> Signed-off-by: Peter Griffin <peter.griffin@linaro.org> Tested-by: Will McVicker <willmcvicker@google.com> Tested-by: André Draszik <andre.draszik@linaro.org> Reviewed-by: André Draszik <andre.draszik@linaro.org> Link: https://lore.kernel.org/r/20250314-ufs-dma-coherent-v1-1-bdf9f9be2919@linaro.org Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Switch Schmitt Trigger functions for PIN_INPUT* macros by default. This is
HW PoR configuration, the slew rate requirements without ST enabled are
pretty tough for these devices. We've noticed spurious GPIO interrupts even
with noise-free edges but not meeting slew rate requirements (3.3E+6 V/s
for 3.3v LVCMOS).
It's not obvious why one might want to disable the PoR-enabled ST on any
pin. Just enable it by default. As it's not possible to provide OR-able
macros to disable the ST, shall anyone require it, provide a set of
new macros with _NOST suffix.
Fixes: fe49f2d776f7 ("arm64: dts: ti: Use local header for pinctrl register values") Cc: stable@vger.kernel.org Signed-off-by: Alexander Sverdlin <alexander.sverdlin@siemens.com> Link: https://lore.kernel.org/r/20250701105437.3539924-1-alexander.sverdlin@siemens.com
[vigneshr@ti.com: Add Fixes tag] Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The NODATASUM message was printed twice by mistake and the NODATACOW was
missing from the 'unset' part. Fix the duplication and make the output
look the same.
After the fsconfig migration in 6.8, mount option info messages are no
longer displayed during mount operations because btrfs_emit_options() is
only called during remount, not during initial mount.
Fix this by calling btrfs_emit_options() in btrfs_fill_super() after
open_ctree() succeeds. Additionally, prevent log duplication by ensuring
btrfs_check_options() handles validation with warn-level and err-level
messages, while btrfs_emit_options() provides info-level messages.
Fix a wrong log message that appears when the "nobarrier" mount option
is unset. When "nobarrier" is unset, barrier is actually enabled.
However, the log incorrectly stated "turning off barriers".
Since commit 13bb483d32ab ("btrfs: zoned: activate metadata block group on
write time"), we activate a metadata block group at the write time. If the
zone capacity is small enough, we can allocate the entire region before the
first write. Then, we hit the btrfs_zoned_bg_is_full() in
btrfs_zone_activate() and the activation fails.
For a data block group, we activate it at the allocation time and we should
check the fullness condition in the caller side. Add, a WARN to check the
fullness condition.
For a metadata block group, we don't need the fullness check because we
activate it at the write time. Instead, activating it once it is written
should be invalid. Catch that with a WARN too.
Fixes: 13bb483d32ab ("btrfs: zoned: activate metadata block group on write time") CC: stable@vger.kernel.org # 6.6+ 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>
In a filesystem with a block size larger than 4KB, the hole length
calculation for a non-extent inode in ext4_ind_map_blocks() can easily
exceed INT_MAX. Then it could return a zero length hole and trigger the
following waring and infinite in the iomap infrastructure.
When the file system is frozen in preparation for taking an LVM
snapshot, the journal is checkpointed and if the orphan_file feature
is enabled, and the orphan file is empty, we clear the orphan_present
feature flag. But if there are pending inodes that need to be removed
the orphan_present feature flag can't be cleared.
The problem comes if the block device is read-only. In that case, we
can't process the orphan inode list, so it is skipped in
ext4_orphan_cleanup(). But then in ext4_mark_recovery_complete(),
this results in the ext4 error "Orphan file not empty on read-only fs"
firing and the file system mount is aborted.
Fix this by clearing the needs_recovery flag in the block device is
read-only. We do this after the call to ext4_load_and_init-journal()
since there are some error checks need to be done in case the journal
needs to be replayed and the block device is read-only, or if the
block device containing the externa journal is read-only, etc.
In some cases like small FSes with no meta_bg and where the resize
doesn't need extra gdt blocks as it can fit in the current one,
s_reserved_gdt_blocks is set as 0, which causes fsmap to emit a 0
length entry, which is incorrect.
With bigalloc enabled, the logic to report last extent has a bug since
we try to use cluster units instead of block units. This can cause an
issue where extra incorrect entries might be returned back to the
user. This was flagged by generic/365 with 64k bs and -O bigalloc.
** Details of issue **
The issue was noticed on 5G 64k blocksize FS with -O bigalloc which has
only 1 bg.
Only the first entry was expected to be returned but we get 2. This is
because:
ext4_getfsmap_datadev()
first_cluster, last_cluster = 0
...
info->gfi_last = true;
ext4_getfsmap_datadev_helper(sb, end_ag, last_cluster + 1, 0, info);
fsb = C2B(1) = 16
fslen = 0
...
/* Merge in any relevant extents from the meta_list */
list_for_each_entry_safe(p, tmp, &info->gfi_meta_list, fmr_list) {
...
// since fsb = 16, considers all metadata which starts before 16 blockno
iter 1: error = ext4_getfsmap_helper(sb, info, p); // p = sb (0,1), nop
info->gfi_next_fsblk = 1
iter 2: error = ext4_getfsmap_helper(sb, info, p); // p = gdt (1,2), nop
info->gfi_next_fsblk = 2
iter 3: error = ext4_getfsmap_helper(sb, info, p); // p = blk bitmap (2,3), nop
info->gfi_next_fsblk = 3
iter 4: error = ext4_getfsmap_helper(sb, info, p); // p = ino bitmap (18,19)
if (rec_blk > info->gfi_next_fsblk) { // (18 > 3)
// emits an extra entry ** BUG **
}
}
Fix this by directly calling ext4_getfsmap_datadev() with a dummy
record that has fmr_physical set to (end_fsb + 1) instead of
last_cluster + 1. By using the block instead of cluster we get the
correct behavior.
Replacing ext4_getfsmap_datadev_helper() with ext4_getfsmap_helper()
is okay since the gfi_lastfree and metadata checks in
ext4_getfsmap_datadev_helper() are anyways redundant when we only want
to emit the last allocated block of the range, as we have already
taken care of emitting metadata and any last free blocks.
The check for a fast symlink in the presence of only an
external xattr inode is incorrect. If a fast symlink does
not have an xattr block (i_file_acl == 0), but does have
an external xattr inode that increases inode i_blocks, then
the check for a fast symlink will incorrectly fail and
__ext4_iget()->ext4_ind_check_inode() will report the inode
is corrupt when it "validates" i_data[] on the next read:
(note that "block 7303014" = 0x6f6f66 = "foo" in LE order).
ext4_inode_is_fast_symlink() should check the superblock
EXT4_FEATURE_INCOMPAT_EA_INODE feature flag, not the inode
EXT4_EA_INODE_FL, since the latter is only set on the xattr
inode itself, and not on the inode that uses this xattr.
Cc: stable@vger.kernel.org Fixes: fc82228a5e38 ("ext4: support fast symlinks from ext3 file systems") Signed-off-by: Andreas Dilger <adilger@whamcloud.com> Reviewed-by: Li Dongyang <dongyangli@ddn.com> Reviewed-by: Alex Zhuravlev <bzzz@whamcloud.com> Reviewed-by: Oleg Drokin <green@whamcloud.com>
Reviewed-on: https://review.whamcloud.com/59879
Lustre-bug-id: https://jira.whamcloud.com/browse/LU-19121 Link: https://patch.msgid.link/20250717063709.757077-1-adilger@dilger.ca Signed-off-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Update the connection tracking logic to handle both IPv4 and IPv6
address families.
Cc: stable@vger.kernel.org Fixes: e6bb91939740 ("ksmbd: limit repeated connections from clients with the same IP") 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 ksmbd_conn_releasing(opinfo->conn) returns true,the refcount was not
decremented properly, causing a refcount leak that prevents the count from
reaching zero and the memory from being released.
Cc: stable@vger.kernel.org Signed-off-by: Ziyan Xu <ziyan@securitygossip.com> Signed-off-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The patch is wrong as it checks vc_origin against vc_screenbuf,
while in text mode it should compare against vga_vram_base.
As such it broke VGA text scrolling, which can be reproduced like this:
(1) boot a kernel that is configured to use text mode VGA-console
(2) type commands: ls -l /usr/bin | less -S
(3) scroll up/down with cursor-down/up keys
Reported-by: Jari Ruusu <jariruusu@protonmail.com> Cc: stable@vger.kernel.org Cc: Yi Yang <yiyang13@huawei.com> Cc: GONG Ruiqi <gongruiqi1@huawei.com> Signed-off-by: Helge Deller <deller@gmx.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
octeontx2 crypto driver allocates memory using kmalloc/kzalloc,
and uses this memory for dma (does dma_map_single()). It assumes
that kmalloc/kzalloc will return 128-byte aligned address. But
kmalloc/kzalloc returns 8-byte aligned address after below changes:
"9382bc44b5f5 arm64: allow kmalloc() caches aligned to the
smaller cache_line_size()
Memory allocated are used for following purpose:
- Input data or scatter list address - 8-Byte alignment
- Output data or gather list address - 8-Byte alignment
- Completion address - 32-Byte alignment.
This patch ensures all addresses are aligned as mentioned above.
octeontx2 crypto driver allocates memory using kmalloc/kzalloc,
and uses this memory for dma (does dma_map_single()). It assumes
that kmalloc/kzalloc will return 128-byte aligned address. But
kmalloc/kzalloc returns 8-byte aligned address after below changes:
"9382bc44b5f5 arm64: allow kmalloc() caches aligned to the
smaller cache_line_size()
Memory allocated are used for following purpose:
- Input data or scatter list address - 8-Byte alignment
- Output data or gather list address - 8-Byte alignment
- Completion address - 32-Byte alignment.
This patch ensures all addresses are aligned as mentioned above.
octeontx2 crypto driver allocates memory using kmalloc/kzalloc,
and uses this memory for dma (does dma_map_single()). It assumes
that kmalloc/kzalloc will return 128-byte aligned address. But
kmalloc/kzalloc returns 8-byte aligned address after below changes:
"9382bc44b5f5 arm64: allow kmalloc() caches aligned to the
smaller cache_line_size()"
Completion address should be 32-Byte alignment when loading
microcode.
Repeated loading and unloading of a device specific QAT driver, for
example qat_4xxx, in a tight loop can lead to a crash due to a
use-after-free scenario. This occurs when a power management (PM)
interrupt triggers just before the device-specific driver (e.g.,
qat_4xxx.ko) is unloaded, while the core driver (intel_qat.ko) remains
loaded.
Since the driver uses a shared workqueue (`qat_misc_wq`) across all
devices and owned by intel_qat.ko, a deferred routine from the
device-specific driver may still be pending in the queue. If this
routine executes after the driver is unloaded, it can dereference freed
memory, resulting in a page fault and kernel crash like the following:
BUG: unable to handle page fault for address: ffa000002e50a01c
#PF: supervisor read access in kernel mode
RIP: 0010:pm_bh_handler+0x1d2/0x250 [intel_qat]
Call Trace:
pm_bh_handler+0x1d2/0x250 [intel_qat]
process_one_work+0x171/0x340
worker_thread+0x277/0x3a0
kthread+0xf0/0x120
ret_from_fork+0x2d/0x50
To prevent this, flush the misc workqueue during device shutdown to
ensure that all pending work items are completed before the driver is
unloaded.
Note: This approach may slightly increase shutdown latency if the
workqueue contains jobs from other devices, but it ensures correctness
and stability.
Fixes: e5745f34113b ("crypto: qat - enable power management for QAT GEN4") Signed-off-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com> Cc: stable@vger.kernel.org Reviewed-by: Ahsan Atta <ahsan.atta@intel.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since the CAAM on these SoCs is managed by another ARM core, called the
SECO (Security Controller) on iMX8QM and Secure Enclave on iMX8ULP, which
also reserves access to register page 0 suspend operations cannot touch
this page.
This is similar to when running OPTEE, where OPTEE will reserve page 0.
Track this situation using a new state variable no_page0, reflecting if
page 0 is reserved elsewhere, either by other management cores in SoC or
by OPTEE.
Replace the optee_en check in suspend/resume with the new check.
optee_en cannot go away as it's needed elsewhere to gate OPTEE specific
situations.
The MIPS32r2 ChaCha code has never been buildable with the clang
assembler. First, clang doesn't support the 'rotl' pseudo-instruction:
error: unknown instruction, did you mean: rol, rotr?
Second, clang requires that both operands of the 'wsbh' instruction be
explicitly given:
error: too few operands for instruction
To fix this, align the code with the real instruction set by (1) using
the real instruction 'rotr' instead of the nonstandard pseudo-
instruction 'rotl', and (2) explicitly giving both operands to 'wsbh'.
To make removing the use of 'rotl' a bit easier, also remove the
unnecessary special-casing for big endian CPUs at
.Lchacha_mips_xor_bytes. The tail handling is actually
endian-independent since it processes one byte at a time. On big endian
CPUs the old code byte-swapped SAVED_X, then iterated through it in
reverse order. But the byteswap and reverse iteration canceled out.
Tested with chacha20poly1305-selftest in QEMU using "-M malta" with both
little endian and big endian mips32r2 kernels.
Fixes: 49aa7c00eddf ("crypto: mips/chacha - import 32r2 ChaCha code from Zinc") Cc: stable@vger.kernel.org Reported-by: kernel test robot <lkp@intel.com> Closes: https://lore.kernel.org/oe-kbuild-all/202505080409.EujEBwA0-lkp@intel.com/ Link: https://lore.kernel.org/r/20250619225535.679301-1-ebiggers@kernel.org Signed-off-by: Eric Biggers <ebiggers@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The maximum number of keycodes got bumped to 256 a very long time ago,
but the default keymaps were never adjusted to match. This is causing
the kernel to interpret keycodes above 127 as U+0000 if the shipped
generated keymap is used.
Fix this by mapping all keycodes above 127 to K_HOLE so the kernel
ignores them.
The contents of this patche were generated by rerunning `loadkeys
--mktable --unicode` and only including the changes to map keycodes
above 127 to K_HOLE.
We don't process Unicode characters if the virtual terminal is in raw
mode, so there's no reason why we shouldn't do the same for K_OFF
(especially since people would expect K_OFF to actually turn off all VT
key processing).
When a remote device sends a completion event to the host, it contains a
pointer to the consumed TRE. The host uses this pointer to process all of
the TREs between it and the host's local copy of the ring's read pointer.
This works when processing completion for chained transactions, but can
lead to nasty results if the device sends an event for a single-element
transaction with a read pointer that is multiple elements ahead of the
host's read pointer.
For instance, if the host accesses an event ring while the device is
updating it, the pointer inside of the event might still point to an old
TRE. If the host uses the channel's xfer_cb() to directly free the buffer
pointed to by the TRE, the buffer will be double-freed.
This behavior was observed on an ep that used upstream EP stack without
'commit 6f18d174b73d ("bus: mhi: ep: Update read pointer only after buffer
is written")'. Where the device updated the events ring pointer before
updating the event contents, so it left a window where the host was able to
access the stale data the event pointed to, before the device had the
chance to update them. The usual pattern was that the host received an
event pointing to a TRE that is not immediately after the last processed
one, so it got treated as if it was a chained transaction, processing all
of the TREs in between the two read pointers.
This commit aims to harden the host by ensuring transactions where the
event points to a TRE that isn't local_rp + 1 are chained.
Fixes: 1d3173a3bae7 ("bus: mhi: core: Add support for processing events from client device") Signed-off-by: Youssef Samir <quic_yabdulra@quicinc.com>
[mani: added stable tag and reworded commit message] Signed-off-by: Manivannan Sadhasivam <mani@kernel.org> Reviewed-by: Jeff Hugo <jeff.hugo@oss.qualcomm.com> Cc: stable@vger.kernel.org Link: https://patch.msgid.link/20250714163039.3438985-1-quic_yabdulra@quicinc.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
On big endian platform like PowerPC, the MHI bus (which is little endian)
does not start properly. The following example shows the error messages by
using QCN9274 WLAN device with ath12k driver:
ath12k_pci 0001:01:00.0: BAR 0: assigned [mem 0xc00000000-0xc001fffff 64bit]
ath12k_pci 0001:01:00.0: MSI vectors: 1
ath12k_pci 0001:01:00.0: Hardware name: qcn9274 hw2.0
ath12k_pci 0001:01:00.0: failed to set mhi state: POWER_ON(2)
ath12k_pci 0001:01:00.0: failed to start mhi: -110
ath12k_pci 0001:01:00.0: failed to power up :-110
ath12k_pci 0001:01:00.0: failed to create soc core: -110
ath12k_pci 0001:01:00.0: failed to init core: -110
ath12k_pci: probe of 0001:01:00.0 failed with error -110
The issue seems to be with the incorrect DMA address/size used for
transferring the firmware image over BHI. So fix it by converting the DMA
address and size of the BHI vector table to little endian format before
sending them to the device.
Fixes: 6cd330ae76ff ("bus: mhi: core: Add support for ringing channel/event ring doorbells") Signed-off-by: Alexander Wilhelm <alexander.wilhelm@westermo.com>
[mani: added stable tag and reworded commit message] Signed-off-by: Manivannan Sadhasivam <mani@kernel.org> Reviewed-by: Jeff Hugo <jeff.hugo@oss.qualcomm.com> Reviewed-by: Krishna Chaitanya Chundru <krishna.chundru@oss.qualcomm.com> Cc: stable@vger.kernel.org Link: https://patch.msgid.link/20250519145837.958153-1-alexander.wilhelm@westermo.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
After a recent change in clang to expose uninitialized warnings from
const variables [1], there is a warning in cxacru_heavy_init():
drivers/usb/atm/cxacru.c:1104:6: error: variable 'bp' is used uninitialized whenever 'if' condition is false [-Werror,-Wsometimes-uninitialized]
1104 | if (instance->modem_type->boot_rom_patch) {
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/usb/atm/cxacru.c:1113:39: note: uninitialized use occurs here
1113 | cxacru_upload_firmware(instance, fw, bp);
| ^~
drivers/usb/atm/cxacru.c:1104:2: note: remove the 'if' if its condition is always true
1104 | if (instance->modem_type->boot_rom_patch) {
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/usb/atm/cxacru.c:1095:32: note: initialize the variable 'bp' to silence this warning
1095 | const struct firmware *fw, *bp;
| ^
| = NULL
While the warning is technically correct that bp is conditionally passed
uninitialized to cxacru_upload_firmware(), it is ultimately a false
positive warning on the uninitialized use of bp because the same
condition that initializes bp, instance->modem_type->boot_rom_patch, is
the same one that gates the use of bp within cxacru_upload_firmware().
As this warning occurs in clang's frontend before inlining occurs, it
cannot know that these conditions are indentical to avoid the warning.
Manually inline cxacru_upload_firmware() into cxacru_heavy_init(), as
that is its only callsite, so that clang can see that bp is initialized
and used under the same condition, clearing up the warning without any
functional changes to the code (LLVM was already doing this inlining
later).
Move the cursor position rightward after rendering the character,
not before. This avoids complications that arise when the recursive
console_putc call has to wrap the line and/or scroll the display.
This also fixes the linewrap bug that crops off the rightmost column.
When the cursor is at the bottom of the display, a linefeed will not
move the cursor position further downward. Instead, the display scrolls
upward. Avoid the repeated add/subtract sequence by way of a single
subtraction at the initialization of console_struct_num_rows.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Cc: stable@vger.kernel.org Signed-off-by: Finn Thain <fthain@linux-m68k.org> Tested-by: Stan Johnson <userm57@yahoo.com> Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org> Link: https://lore.kernel.org/9d4e8c68a456d5f2bc254ac6f87a472d066ebd5e.1743115195.git.fthain@linux-m68k.org Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
DM targets must not split zone append and write operations using
dm_accept_partial_bio() as doing so is forbidden for zone append BIOs,
breaks zone append emulation using regular write BIOs and potentially
creates deadlock situations with queue freeze operations.
Modify dm_accept_partial_bio() to add missing BUG_ON() checks for all
these cases, that is, check that the BIO is a write or write zeroes
operation. This change packs all the zone related checks together under
a static_branch_unlikely(&zoned_enabled) and done only if the target is
a zoned device.
Fixes: f211268ed1f9 ("dm: Use the block layer zone append emulation") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <dlemoal@kernel.org> Reviewed-by: Mikulas Patocka <mpatocka@redhat.com> Link: https://lore.kernel.org/r/20250625093327.548866-6-dlemoal@kernel.org Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Read and write operations issued to a dm-crypt target may be split
according to the dm-crypt internal limits defined by the max_read_size
and max_write_size module parameters (default is 128 KB). The intent is
to improve processing time of large BIOs by splitting them into smaller
operations that can be parallelized on different CPUs.
For zoned dm-crypt targets, this BIO splitting is still done but without
the parallel execution to ensure that the issuing order of write
operations to the underlying devices remains sequential. However, the
splitting itself causes other problems:
1) Since dm-crypt relies on the block layer zone write plugging to
handle zone append emulation using regular write operations, the
reminder of a split write BIO will always be plugged into the target
zone write plugged. Once the on-going write BIO finishes, this
reminder BIO is unplugged and issued from the zone write plug work.
If this reminder BIO itself needs to be split, the reminder will be
re-issued and plugged again, but that causes a call to a
blk_queue_enter(), which may block if a queue freeze operation was
initiated. This results in a deadlock as DM submission still holds
BIOs that the queue freeze side is waiting for.
2) dm-crypt relies on the emulation done by the block layer using
regular write operations for processing zone append operations. This
still requires to properly return the written sector as the BIO
sector of the original BIO. However, this can be done correctly only
and only if there is a single clone BIO used for processing the
original zone append operation issued by the user. If the size of a
zone append operation is larger than dm-crypt max_write_size, then
the orginal BIO will be split and processed as a chain of regular
write operations. Such chaining result in an incorrect written sector
being returned to the zone append issuer using the original BIO
sector. This in turn results in file system data corruptions using
xfs or btrfs.
Fix this by modifying get_max_request_size() to always return the size
of the BIO to avoid it being split with dm_accpet_partial_bio() in
crypt_map(). get_max_request_size() is renamed to
get_max_request_sectors() to clarify the unit of the value returned
and its interface is changed to take a struct dm_target pointer and a
pointer to the struct bio being processed. In addition to this change,
to ensure that crypt_alloc_buffer() works correctly, set the dm-crypt
device max_hw_sectors limit to be at most
BIO_MAX_VECS << PAGE_SECTORS_SHIFT (1 MB with a 4KB page architecture).
This forces DM core to split write BIOs before passing them to
crypt_map(), and thus guaranteeing that dm-crypt can always accept an
entire write BIO without needing to split it.
This change does not have any effect on the read path of dm-crypt. Read
operations can still be split and the BIO fragments processed in
parallel. There is also no impact on the performance of the write path
given that all zone write BIOs were already processed inline instead of
in parallel.
This change also does not affect in any way regular dm-crypt block
devices.
Fixes: f211268ed1f9 ("dm: Use the block layer zone append emulation") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <dlemoal@kernel.org> Reviewed-by: Mikulas Patocka <mpatocka@redhat.com> Link: https://lore.kernel.org/r/20250625093327.548866-5-dlemoal@kernel.org Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For all practical purposes, there is no difference between the situation
in which a given device is not ignoring children and its active child
count is nonzero and the situation in which its runtime PM usage counter
is nonzero. However, pm_runtime_get_if_in_use() will only increment the
device's usage counter and return 1 in the latter case.
For consistency, make it do so in the former case either by adjusting
pm_runtime_get_conditional() and update the related kerneldoc comments
accordingly.
The freq_tables[] array has num_possible_cpus() elements so, to avoid an
out of bounds access, this loop should be capped at "< nb_cpus" instead
of "<= nb_cpus". The freq_tables[] array is allocated in
armada_8k_cpufreq_init().
Cc: stable@vger.kernel.org Fixes: f525a670533d ("cpufreq: ap806: add cpufreq driver for Armada 8K") Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Improve the description of the possible default SATA link power
management policies and add the missing description for policy 5.
No functional changes.
Fixes: a5ec5a7bfd1f ("ata: ahci: Support state with min power but Partial low power state") Cc: stable@vger.kernel.org Signed-off-by: Damien Le Moal <dlemoal@kernel.org> Reviewed-by: Hannes Reinecke <hare@suse.de> Reviewed-by: Niklas Cassel <cassel@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When the PSLVERR_RESP_EN parameter is set to 1, the device generates
an error response if an attempt is made to read an empty RBR (Receive
Buffer Register) while the FIFO is enabled.
In serial8250_do_startup(), calling serial_port_out(port, UART_LCR,
UART_LCR_WLEN8) triggers dw8250_check_lcr(), which invokes
dw8250_force_idle() and serial8250_clear_and_reinit_fifos(). The latter
function enables the FIFO via serial_out(p, UART_FCR, p->fcr).
Execution proceeds to the serial_port_in(port, UART_RX).
This satisfies the PSLVERR trigger condition.
When another CPU (e.g., using printk()) is accessing the UART (UART
is busy), the current CPU fails the check (value & ~UART_LCR_SPAR) ==
(lcr & ~UART_LCR_SPAR) in dw8250_check_lcr(), causing it to enter
dw8250_force_idle().
Put serial_port_out(port, UART_LCR, UART_LCR_WLEN8) under the port->lock
to fix this issue.
If CONFIG_ACPI_SPCR_TABLE is disabled, acpi_parse_spcr()
currently returns 0, which may incorrectly suggest that
SPCR parsing was successful. This patch changes the behavior
to return -ENODEV to clearly indicate that SPCR support
is not available.
This prepares the codebase for future changes that depend
on acpi_parse_spcr() failure detection, such as suppressing
misleading console messages.
The Supported Link Speeds Vector in the Link Capabilities 2 Register
indicates the *supported* link speeds. The Max Link Speed field in the
Link Capabilities Register indicates the *maximum* of those speeds.
pcie_get_supported_speeds() neglects to honor the Max Link Speed field and
will thus incorrectly deem higher speeds as supported. Fix it.
One user-visible issue addressed here is an incorrect value in the sysfs
attribute "max_link_speed".
But the main motivation is a boot hang reported by Niklas: Intel JHL7540
"Titan Ridge 2018" Thunderbolt controllers supports 2.5-8 GT/s speeds,
but indicate 2.5 GT/s as maximum. Ilpo recalls seeing this on more
devices. It can be explained by the controller's Downstream Ports
supporting 8 GT/s if an Endpoint is attached, but limiting to 2.5 GT/s
if the port interfaces to a PCIe Adapter, in accordance with USB4 v2
sec 11.2.1:
"This section defines the functionality of an Internal PCIe Port that
interfaces to a PCIe Adapter. [...]
The Logical sub-block shall update the PCIe configuration registers
with the following characteristics: [...]
Max Link Speed field in the Link Capabilities Register set to 0001b
(data rate of 2.5 GT/s only).
Note: These settings do not represent actual throughput. Throughput
is implementation specific and based on the USB4 Fabric performance."
The present commit is not sufficient on its own to fix Niklas' boot hang,
but it is a prerequisite: A subsequent commit will fix the boot hang by
enabling bandwidth control only if more than one speed is supported.
The GENMASK() macro used herein specifies 0 as lowest bit, even though
the Supported Link Speeds Vector ends at bit 1. This is done on purpose
to avoid a GENMASK(0, 1) macro if Max Link Speed is zero. That macro
would be invalid as the lowest bit is greater than the highest bit.
Ilpo has witnessed a zero Max Link Speed on Root Complex Integrated
Endpoints in particular, so it does occur in practice. (The Link
Capabilities Register is optional on RCiEPs per PCIe r6.2 sec 7.5.3.)
Fixes: d2bd39c0456b ("PCI: Store all PCIe Supported Link Speeds") Closes: https://lore.kernel.org/r/70829798889c6d779ca0f6cd3260a765780d1369.camel@kernel.org Link: https://lore.kernel.org/r/fe03941e3e1cc42fb9bf4395e302bff53ee2198b.1734428762.git.lukas@wunner.de Reported-by: Niklas Schnelle <niks@kernel.org> Tested-by: Niklas Schnelle <niks@kernel.org> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 2df7168717b7 ("dm: Always split write BIOs to zoned device
limits") updates the device-mapper driver to perform splits for the
write BIOs. However, it did not address the cases where DM targets do
not emulate zone append, such as in the cases of dm-linear or dm-flakey.
For these targets, when the write BIOs span across zone boundaries, they
trigger WARN_ON_ONCE(bio_straddles_zones(bio)) in
blk_zone_wplug_handle_write(). This results in I/O errors. The errors
are reproduced by running blktests test case zbd/004 using zoned
dm-linear or dm-flakey devices.
To avoid the I/O errors, handle the write BIOs regardless whether DM
targets emulate zone append or not, so that all write BIOs are split at
zone boundaries. For that purpose, drop the check for zone append
emulation in dm_zone_bio_needs_split(). Its argument 'md' is no longer
used then drop it also.
RCU re-initializes the deferred QS irq work everytime before attempting
to queue it. However there are situations where the irq work is
attempted to be queued even though it is already queued. In that case
re-initializing messes-up with the irq work queue that is about to be
handled.
The chances for that to happen are higher when the architecture doesn't
support self-IPIs and irq work are then all lazy, such as with the
following sequence:
1) rcu_read_unlock() is called when IRQs are disabled and there is a
grace period involving blocked tasks on the node. The irq work
is then initialized and queued.
2) The related tasks are unblocked and the CPU quiescent state
is reported. rdp->defer_qs_iw_pending is reset to DEFER_QS_IDLE,
allowing the irq work to be requeued in the future (note the previous
one hasn't fired yet).
3) A new grace period starts and the node has blocked tasks.
4) rcu_read_unlock() is called when IRQs are disabled again. The irq work
is re-initialized (but it's queued! and its node is cleared) and
requeued. Which means it's requeued to itself.
5) The irq work finally fires with the tick. But since it was requeued
to itself, it loops and hangs.
Fix this with initializing the irq work only once before the CPU boots.
Fixes: b41642c87716 ("rcu: Fix rcu_read_unlock() deadloop due to IRQ work") Reported-by: kernel test robot <oliver.sang@intel.com> Closes: https://lore.kernel.org/oe-lkp/202508071303.c1134cce-lkp@intel.com Signed-off-by: Frederic Weisbecker <frederic@kernel.org> Reviewed-by: Joel Fernandes <joelagnelf@nvidia.com> Signed-off-by: Neeraj Upadhyay (AMD) <neeraj.upadhyay@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[Why & How]
Not letting DCN301 to clear after surface/stream update results
in artifacts when switching between active overlay planes. The issue
is known and has been solved initially. See below:
(https://gitlab.freedesktop.org/drm/amd/-/issues/3441)
Fixes: f354556e29f4 ("drm/amd/display: limit clear_update_flags t dcn32 and above") Reviewed-by: Mario Limonciello <mario.limonciello@amd.com> Signed-off-by: Ivan Lipski <ivan.lipski@amd.com> Tested-by: Daniel Wheeler <daniel.wheeler@amd.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The old SET_SYSTEM_SLEEP_PM_OPS() macro leads to a warning about an
unused function:
| drivers/firmware/arm_scmi/scmi_power_control.c:363:12: error:
| 'scmi_system_power_resume' defined but not used [-Werror=unused-function]
| static int scmi_system_power_resume(struct device *dev)
The proper way to do this these days is to use SYSTEM_SLEEP_PM_OPS()
and pm_sleep_ptr().
Fixes: 9a0658d3991e ("firmware: arm_scmi: power_control: Ensure SCMI_SYSPOWER_IDLE is set early during resume") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Acked-by: Peng Fan <peng.fan@nxp.com>
Message-Id: <20250709070107.1388512-1-arnd@kernel.org> Signed-off-by: Sudeep Holla <sudeep.holla@arm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A port link power management (LPM) policy can be controlled using the
link_power_management_policy sysfs host attribute. However, this
attribute exists also for hosts that do not support LPM and in such
case, attempting to change the LPM policy for the host (port) will fail
with -EOPNOTSUPP.
Introduce the new sysfs link_power_management_supported host attribute
to indicate to the user if a the port and the devices connected to the
port for the host support LPM, which implies that the
link_power_management_policy attribute can be used.
Since checking that a port and its devices support LPM is common between
the new ata_scsi_lpm_supported_show() function and the existing
ata_scsi_lpm_store() function, the new helper ata_scsi_lpm_supported()
is introduced.
Fixes: 413e800cadbf ("ata: libata-sata: Disallow changing LPM state if not supported") Reported-by: Borah, Chaitanya Kumar <chaitanya.kumar.borah@intel.com> Reported-by: kernel test robot <oliver.sang@intel.com> Closes: https://lore.kernel.org/oe-lkp/202507251014.a5becc3b-lkp@intel.com Signed-off-by: Damien Le Moal <dlemoal@kernel.org> Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Starting with Rust 1.88.0 (released 2025-06-26), `rustdoc` complains
about a target modifier mismatch in configurations where `-Zfixed-x18`
is passed:
error: mixing `-Zfixed-x18` will cause an ABI mismatch in crate `rust_out`
|
= help: the `-Zfixed-x18` flag modifies the ABI so Rust crates compiled with different values of this flag cannot be used together safely
= note: unset `-Zfixed-x18` in this crate is incompatible with `-Zfixed-x18=` in dependency `core`
= help: set `-Zfixed-x18=` in this crate or unset `-Zfixed-x18` in `core`
= help: if you are sure this will not cause problems, you may use `-Cunsafe-allow-abi-mismatch=fixed-x18` to silence this error
The reason is that `rustdoc` was not passing the target modifiers when
configuring the session options, and thus it would report a mismatch
that did not exist as soon as a target modifier is used in a dependency.
We did not notice it in the kernel until now because `-Zfixed-x18` has
been a target modifier only since 1.88.0 (and it is the only one we use
so far).
The issue has been reported upstream [1] and a fix has been submitted
[2], including a test similar to the kernel case.
[ This is now fixed upstream (thanks Guillaume for the quick review),
so it will be fixed in Rust 1.90.0 (expected 2025-09-18).
- Miguel ]
Meanwhile, conditionally pass `-Cunsafe-allow-abi-mismatch=fixed-x18`
to workaround the issue on our side.
`rustdoc` can get confused when generating documentation into a folder
that contains generated files from other `rustdoc` versions.
For instance, running something like:
rustup default 1.78.0
make LLVM=1 rustdoc
rustup default 1.88.0
make LLVM=1 rustdoc
may generate errors like:
error: couldn't generate documentation: invalid template: last line expected to start with a comment
|
= note: failed to create or modify "./Documentation/output/rust/rustdoc/src-files.js"
Thus just always clean the output folder before generating the
documentation -- we are anyway regenerating it every time the `rustdoc`
target gets called, at least for the time being.
Cc: stable@vger.kernel.org # Needed in 6.12.y and later (Rust is pinned in older LTSs). Reported-by: Daniel Almeida <daniel.almeida@collabora.com> Closes: https://rust-for-linux.zulipchat.com/#narrow/channel/288089/topic/x/near/527201113 Reviewed-by: Tamir Duberstein <tamird@kernel.org> Link: https://lore.kernel.org/r/20250726133435.2460085-1-ojeda@kernel.org Signed-off-by: Miguel Ojeda <ojeda@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
According to the "GPIO Expander Map / Table" section of the J722S EVM
Schematic within the Evaluation Module Design Files package [0], the
GPIO Pin P05 located on the GPIO Expander 1 (I2C0/0x23) has to be pulled
down to select the Type-C interface. Since commit under Fixes claims to
enable the Type-C interface, update the property within "p05-hog" from
"output-high" to "output-low", thereby switching from the Type-A
interface to the Type-C interface.
[0]: https://www.ti.com/lit/zip/sprr495
Cc: stable@vger.kernel.org Fixes: 485705df5d5f ("arm64: dts: ti: k3-j722s: Enable PCIe and USB support on J722S-EVM") Signed-off-by: Siddharth Vadapalli <s-vadapalli@ti.com> Link: https://lore.kernel.org/r/20250623100657.4082031-1-s-vadapalli@ti.com Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
J722S SOC has two usb controllers USB0 and USB1. USB0 is brought out on
the EVM as a stacked USB connector which has one Type-A and one Type-C
port. These Type-A and Type-C ports are connected to MUX so only
one of them can be enabled at a time.
Commit under Fixes, tries to enable the USB0 instance of USB to
interface with the Type-C port via the USB hub, by configuring the
USB2.0_MUX_SEL to GPIO_ACTIVE_HIGH. But it is observed on J722S-EVM
that Type-A port is enabled instead of Type-C port.
Fix this by setting USB2.0_MUX_SEL to GPIO_ACTIVE_LOW to enable Type-C
port.
Fixes: 485705df5d5f ("arm64: dts: ti: k3-j722s: Enable PCIe and USB support on J722S-EVM") Signed-off-by: Hrushikesh Salunke <h-salunke@ti.com> Reviewed-by: Roger Quadros <rogerq@kernel.org> Link: https://lore.kernel.org/r/20250116125726.2549489-1-h-salunke@ti.com Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Stable-dep-of: 65ba2a6e77e9 ("arm64: dts: ti: k3-j722s-evm: Fix USB gpio-hog level for Type-C") Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
pci_bridge_d3_possible() is called from both pcie_portdrv_probe() and
pcie_portdrv_remove() to determine whether runtime power management shall
be enabled (on probe) or disabled (on remove) on a PCIe port.
The underlying assumption is that pci_bridge_d3_possible() always returns
the same value, else a runtime PM reference imbalance would occur. That
assumption is not given if the PCIe port is inaccessible on remove due to
hot-unplug: pci_bridge_d3_possible() calls pciehp_is_native(), which
accesses Config Space to determine whether the port is Hot-Plug Capable.
An inaccessible port returns "all ones", which is converted to "all
zeroes" by pcie_capability_read_dword(). Hence the port no longer seems
Hot-Plug Capable on remove even though it was on probe.
The resulting runtime PM ref imbalance causes warning messages such as:
Avoid the Config Space access (and thus the runtime PM ref imbalance) by
caching the Hot-Plug Capable bit in struct pci_dev.
The struct already contains an "is_hotplug_bridge" flag, which however is
not only set on Hot-Plug Capable PCIe ports, but also Conventional PCI
Hot-Plug bridges and ACPI slots. The flag identifies bridges which are
allocated additional MMIO and bus number resources to allow for hierarchy
expansion.
The kernel is somewhat sloppily using "is_hotplug_bridge" in a number of
places to identify Hot-Plug Capable PCIe ports, even though the flag
encompasses other devices. Subsequent commits replace these occurrences
with the new flag to clearly delineate Hot-Plug Capable PCIe ports from
other kinds of hotplug bridges.
Document the existing "is_hotplug_bridge" and the new "is_pciehp" flag
and document the (non-obvious) requirement that pci_bridge_d3_possible()
always returns the same value across the entire lifetime of a bridge,
including its hot-removal.
Currently, pci_bridge_d3_possible() encodes a variety of decision factors
when deciding whether a given bridge can be put into D3. A particular one
of note is for "recent enough PCIe ports." Per Rafael [0]:
"There were hardware issues related to PM on x86 platforms predating
the introduction of Connected Standby in Windows. For instance,
programming a port into D3hot by writing to its PMCSR might cause the
PCIe link behind it to go down and the only way to revive it was to
power cycle the Root Complex. And similar."
Thus, this function contains a DMI-based check for post-2015 BIOS.
The above factors (Windows, x86) don't really apply to non-x86 systems, and
also, many such systems don't have BIOS or DMI. However, we'd like to be
able to suspend bridges on non-x86 systems too.
Restrict the "recent enough" check to x86. If we find further
incompatibilities, it probably makes sense to expand on the deny-list
approach (i.e., bridge_d3_blacklist or similar).
The PCIe bandwidth controller added by a subsequent commit will require
selecting PCIe Link Speeds that are lower than the Maximum Link Speed.
The struct pci_bus only stores max_bus_speed. Even if PCIe r6.1 sec 8.2.1
currently disallows gaps in supported Link Speeds, the Implementation Note
in PCIe r6.1 sec 7.5.3.18, recommends determining supported Link Speeds
using the Supported Link Speeds Vector in the Link Capabilities 2 Register
(when available) to "avoid software being confused if a future
specification defines Links that do not require support for all slower
speeds."
Reuse code in pcie_get_speed_cap() to add pcie_get_supported_speeds() to
query the Supported Link Speeds Vector of a PCIe device. The value is taken
directly from the Supported Link Speeds Vector or synthesized from the Max
Link Speed in the Link Capabilities Register when the Link Capabilities 2
Register is not available.
The Supported Link Speeds Vector in the Link Capabilities Register 2
corresponds to the bus below on Root Ports and Downstream Ports, whereas it
corresponds to the bus above on Upstream Ports and Endpoints (PCIe r6.1 sec
7.5.3.18):
Supported Link Speeds Vector - This field indicates the supported Link
speed(s) of the associated Port.
Add supported_speeds into the struct pci_dev that caches the
Supported Link Speeds Vector.
supported_speeds contains a set of Link Speeds only in the case where PCIe
Link Speed can be determined. Root Complex Integrated Endpoints do not have
a well-defined Link Speed because they do not implement either of the Link
Capabilities Registers, which is allowed by PCIe r6.1 sec 7.5.3 (the same
limitation applies to determining cur_bus_speed and max_bus_speed that are
PCI_SPEED_UNKNOWN in such case). This is of no concern from PCIe bandwidth
controller point of view because such devices are not attached into a PCIe
Root Port that could be controlled.
The supported_speeds field keeps the extra reserved zero at the least
significant bit to match the Link Capabilities 2 Register layout.
An attempt was made to store supported_speeds field into the struct pci_bus
as an intersection of both ends of the Link, however, the subordinate
struct pci_bus is not available early enough. The Target Speed quirk (in
pcie_failed_link_retrain()) can run either during initial scan or later,
requiring it to use the API provided by the PCIe bandwidth controller to
set the Target Link Speed in order to co-exist with the bandwidth
controller. When the Target Speed quirk is calling the bandwidth controller
during initial scan, the struct pci_bus is not yet initialized. As such,
storing supported_speeds into the struct pci_bus is not viable.
After commit 5c70eb5c593d ("net: better track kernel sockets lifetime"),
kernel sockets now use net_passive reference counting. However, commit 95d2b9f693ff ("Revert "smb: client: fix TCP timers deadlock after rmmod"")
restored the manual socket refcount manipulation without adapting to this
new mechanism, causing a memory leak.
The issue can be reproduced by[1]:
1. Creating a network namespace
2. Mounting and Unmounting CIFS within the namespace
3. Deleting the namespace
Some memory leaks may appear after a period of time following step 3.
Root cause:
When creating kernel sockets, sk_alloc() calls net_passive_inc() for
sockets with sk_net_refcnt=0. The CIFS code manually converts kernel
sockets to user sockets by setting sk_net_refcnt=1, but doesn't call
the corresponding net_passive_dec(). This creates an imbalance in the
net_passive counter, which prevents the network namespace from being
destroyed when its last user reference is dropped. As a result, the
entire namespace and all its associated resources remain allocated.
Timeline of patches leading to this issue:
- commit ef7134c7fc48 ("smb: client: Fix use-after-free of network
namespace.") in v6.12 fixed the original netns UAF by manually
managing socket refcounts
- commit e9f2517a3e18 ("smb: client: fix TCP timers deadlock after
rmmod") in v6.13 attempted to use kernel sockets but introduced
TCP timer issues
- commit 5c70eb5c593d ("net: better track kernel sockets lifetime")
in v6.14-rc5 introduced the net_passive mechanism with
sk_net_refcnt_upgrade() for proper socket conversion
- commit 95d2b9f693ff ("Revert "smb: client: fix TCP timers deadlock
after rmmod"") in v6.15-rc3 reverted to manual refcount management
without adapting to the new net_passive changes
Fix this by using sk_net_refcnt_upgrade() which properly handles the
net_passive counter when converting kernel sockets to user sockets.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=220343 Fixes: 95d2b9f693ff ("Revert "smb: client: fix TCP timers deadlock after rmmod"") Cc: stable@vger.kernel.org Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com> Reviewed-by: Enzo Matsumiya <ematsumiya@suse.de> Signed-off-by: Wang Zhaolong <wangzhaolong@huaweicloud.com> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
While kernel sockets are dismantled during pernet_operations->exit(),
their freeing can be delayed by any tx packets still held in qdisc
or device queues, due to skb_set_owner_w() prior calls.
This then trigger the following warning from ref_tracker_dir_exit() [1]
To fix this, make sure that kernel sockets own a reference on net->passive.
Add sk_net_refcnt_upgrade() helper, used whenever a kernel socket
is converted to a refcounted one.
The charge-control subsystem in the ChromeOS EC is not strictly tied to
its USB-PD subsystem.
Since commit 7613bc0d116a ("mfd: cros_ec: Don't load charger with UCSI")
the presence of EC_FEATURE_UCSI_PPM would inhibit the probing of the
charge-control driver.
Furthermore recent versions of the EC firmware in Framework laptops
hard-disable EC_FEATURE_USB_PD to avoid probing cros-usbpd-charger,
which then also breaks cros-charge-control.
Currently, the battery timer is set up for all devices using hid-apple,
irrespective of whether they actually have a battery or not.
APPLE_RDESC_BATTERY is a quirk that indicates the device has a battery
and needs the battery timer. This patch checks for this quirk before
setting up the timer, ensuring that only devices with a battery will
have the timer set up.
Size of ring buffer, as defined in uio_hv_generic driver, is no longer
fixed to 16 KB. This creates a problem in fcopy, since this size was
hardcoded. With the change in place to make ring sysfs node actually
reflect the size of underlying ring buffer, it is safe to get the size
of ring sysfs file and use it for ring buffer size in fcopy daemon.
Fix the issue of disparity in ring buffer size, by making it dynamic
in fcopy uio daemon.
Cc: stable@vger.kernel.org Fixes: 0315fef2aff9 ("uio_hv_generic: Align ring size to system page") Signed-off-by: Naman Jain <namjain@linux.microsoft.com> Reviewed-by: Saurabh Sengar <ssengar@linux.microsoft.com> Reviewed-by: Long Li <longli@microsoft.com> Link: https://lore.kernel.org/r/20250711060846.9168-1-namjain@linux.microsoft.com Signed-off-by: Wei Liu <wei.liu@kernel.org>
Message-ID: <20250711060846.9168-1-namjain@linux.microsoft.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When userspace sets supported rates for a new station via
NL80211_CMD_NEW_STATION, it might send a list that's empty
or contains only invalid values. Currently, we process these
values in sta_link_apply_parameters() without checking the result of
ieee80211_parse_bitrates(), which can lead to an empty rates bitmap.
A similar issue was addressed for NL80211_CMD_SET_BSS in commit ce04abc3fcc6 ("wifi: mac80211: check basic rates validity").
This patch applies the same approach in sta_link_apply_parameters()
for NL80211_CMD_NEW_STATION, ensuring there is at least one valid
rate by inspecting the result of ieee80211_parse_bitrates().
Found by Linux Verification Center (linuxtesting.org) with Syzkaller.
Fixes: b95eb7f0eee4 ("wifi: cfg80211/mac80211: separate link params from station params") Signed-off-by: Mikhail Lobanov <m.lobanov@rosa.ru> Link: https://patch.msgid.link/20250317103139.17625-1-m.lobanov@rosa.ru Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Hanne-Lotta Mäenpää <hannelotta@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently, the battery timer is set up for all devices using
hid-magicmouse, irrespective of whether they actually need it or not.
The current implementation requires the battery timer for Magic Mouse 2
and Magic Trackpad 2 when connected via USB only. Add checks to ensure
that the battery timer is only set up when they are connected via USB.
Ever since commit c2ff29e99a76 ("siw: Inline do_tcp_sendpages()"),
we have been doing this:
static int siw_tcp_sendpages(struct socket *s, struct page **page, int offset,
size_t size)
[...]
/* Calculate the number of bytes we need to push, for this page
* specifically */
size_t bytes = min_t(size_t, PAGE_SIZE - offset, size);
/* If we can't splice it, then copy it in, as normal */
if (!sendpage_ok(page[i]))
msg.msg_flags &= ~MSG_SPLICE_PAGES;
/* Set the bvec pointing to the page, with len $bytes */
bvec_set_page(&bvec, page[i], bytes, offset);
/* Set the iter to $size, aka the size of the whole sendpages (!!!) */
iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
try_page_again:
lock_sock(sk);
/* Sendmsg with $size size (!!!) */
rv = tcp_sendmsg_locked(sk, &msg, size);
This means we've been sending oversized iov_iters and tcp_sendmsg calls
for a while. This has a been a benign bug because sendpage_ok() always
returned true. With the recent slab allocator changes being slowly
introduced into next (that disallow sendpage on large kmalloc
allocations), we have recently hit out-of-bounds crashes, due to slight
differences in iov_iter behavior between the MSG_SPLICE_PAGES and
"regular" copy paths:
(MSG_SPLICE_PAGES)
skb_splice_from_iter
iov_iter_extract_pages
iov_iter_extract_bvec_pages
uses i->nr_segs to correctly stop in its tracks before OoB'ing everywhere
skb_splice_from_iter gets a "short" read
(!MSG_SPLICE_PAGES)
skb_copy_to_page_nocache copy=iov_iter_count
[...]
copy_from_iter
/* this doesn't help */
if (unlikely(iter->count < len))
len = iter->count;
iterate_bvec
... and we run off the bvecs
Fix this by properly setting the iov_iter's byte count, plus sending the
correct byte count to tcp_sendmsg_locked.
Link: https://patch.msgid.link/r/20250729120348.495568-1-pfalcato@suse.de Cc: stable@vger.kernel.org Fixes: c2ff29e99a76 ("siw: Inline do_tcp_sendpages()") Reported-by: kernel test robot <oliver.sang@intel.com> Closes: https://lore.kernel.org/oe-lkp/202507220801.50a7210-lkp@intel.com Reviewed-by: David Howells <dhowells@redhat.com> Signed-off-by: Pedro Falcato <pfalcato@suse.de> Acked-by: Bernard Metzler <bernard.metzler@linux.dev> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
While nolibc-test does test syscalls, it doesn't test as much the rest
of the macros, and a wrong spelling of FD_SETBITMASK in commit feaf75658783a broke programs using either FD_SET() or FD_CLR() without
being noticed. Let's fix these macros.
Commit 6f1466123d73 ("media: s5p-mfc: Add YV12 and I420 multiplanar
format support") added support for the new formats to s5p-mfc driver,
what in turn required some internal calls to the v4l2_format_info()
function while setting up formats. This in turn broke support for the
"old" tiled NV12MT* formats, which are not recognized by this function.
Fix this by adding those variants of NV12M pixel format to
v4l2_format_info() function database.
Fixes: 6f1466123d73 ("media: s5p-mfc: Add YV12 and I420 multiplanar format support") Cc: stable@vger.kernel.org Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com> Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently, the driver performs a length check of the metadata buffer
before the actual metadata size is known and before the metadata is
decided to be copied. This results in valid metadata buffers being
incorrectly marked as invalid.
Move the length check to occur after the metadata size is determined and
is decided to be copied.
Cc: stable@vger.kernel.org Fixes: 088ead255245 ("media: uvcvideo: Add a metadata device node") Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Reviewed-by: Hans de Goede <hansg@kernel.org> Signed-off-by: Ricardo Ribalda <ribalda@chromium.org> Link: https://lore.kernel.org/r/20250707-uvc-meta-v8-1-ed17f8b1218b@chromium.org Signed-off-by: Hans de Goede <hansg@kernel.org> Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Currently, The event_seq_changed() handler processes a variable number
of properties sent by the firmware. The number of properties is indicated
by the firmware and used to iterate over the payload. However, the
payload size is not being validated against the actual message length.
This can lead to out-of-bounds memory access if the firmware provides a
property count that exceeds the data available in the payload. Such a
condition can result in kernel crashes or potential information leaks if
memory beyond the buffer is accessed.
Fix this by properly validating the remaining size of the payload before
each property access and updating bounds accordingly as properties are
parsed.
This ensures that property parsing is safely bounded within the received
message buffer and protects against malformed or malicious firmware
behavior.
The buffer length check before calling uvc_parse_format() only ensured
that the buffer has at least 3 bytes (buflen > 2), buf the function
accesses buffer[3], requiring at least 4 bytes.
This can lead to an out-of-bounds read if the buffer has exactly 3 bytes.
Fix it by checking that the buffer has at least 4 bytes in
uvc_parse_format().
Signed-off-by: Youngjun Lee <yjjuny.lee@samsung.com> Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Fixes: c0efd232929c ("V4L/DVB (8145a): USB Video Class driver") Cc: stable@vger.kernel.org Reviewed-by: Ricardo Ribalda <ribalda@chromium.org> Link: https://lore.kernel.org/r/20250610124107.37360-1-yjjuny.lee@samsung.com Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com> Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When netpoll is enabled, calling pr_warn_once() while holding
kmemleak_lock in mem_pool_alloc() can cause a deadlock due to lock
inversion with the netconsole subsystem. This occurs because
pr_warn_once() may trigger netpoll, which eventually leads to
__alloc_skb() and back into kmemleak code, attempting to reacquire
kmemleak_lock.
Fix this by setting a flag and issuing the pr_warn_once() after
kmemleak_lock is released.
Link: https://lkml.kernel.org/r/20250731-kmemleak_lock-v1-1-728fd470198f@debian.org Fixes: c5665868183f ("mm: kmemleak: use the memory pool for early allocations") Signed-off-by: Breno Leitao <leitao@debian.org> Reported-by: Jakub Kicinski <kuba@kernel.org> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A soft lockup warning was observed on a relative small system x86-64
system with 16 GB of memory when running a debug kernel with kmemleak
enabled.
watchdog: BUG: soft lockup - CPU#8 stuck for 33s! [kworker/8:1:134]
The test system was running a workload with hot unplug happening in
parallel. Then kemleak decided to disable itself due to its inability to
allocate more kmemleak objects. The debug kernel has its
CONFIG_DEBUG_KMEMLEAK_MEM_POOL_SIZE set to 40,000.
The soft lockup happened in kmemleak_do_cleanup() when the existing
kmemleak objects were being removed and deleted one-by-one in a loop via a
workqueue. In this particular case, there are at least 40,000 objects
that need to be processed and given the slowness of a debug kernel and the
fact that a raw_spinlock has to be acquired and released in
__delete_object(), it could take a while to properly handle all these
objects.
As kmemleak has been disabled in this case, the object removal and
deletion process can be further optimized as locking isn't really needed.
However, it is probably not worth the effort to optimize for such an edge
case that should rarely happen. So the simple solution is to call
cond_resched() at periodic interval in the iteration loop to avoid soft
lockup.
Link: https://lkml.kernel.org/r/20250728190248.605750-1-longman@redhat.com Signed-off-by: Waiman Long <longman@redhat.com> Acked-by: Catalin Marinas <catalin.marinas@arm.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Memory hot remove unmaps and tears down various kernel page table regions
as required. The ptdump code can race with concurrent modifications of
the kernel page tables. When leaf entries are modified concurrently, the
dump code may log stale or inconsistent information for a VA range, but
this is otherwise not harmful.
But when intermediate levels of kernel page table are freed, the dump code
will continue to use memory that has been freed and potentially
reallocated for another purpose. In such cases, the ptdump code may
dereference bogus addresses, leading to a number of potential problems.
To avoid the above mentioned race condition, platforms such as arm64,
riscv and s390 take memory hotplug lock, while dumping kernel page table
via the sysfs interface /sys/kernel/debug/kernel_page_tables.
Similar race condition exists while checking for pages that might have
been marked W+X via /sys/kernel/debug/kernel_page_tables/check_wx_pages
which in turn calls ptdump_check_wx(). Instead of solving this race
condition again, let's just move the memory hotplug lock inside generic
ptdump_check_wx() which will benefit both the scenarios.
Drop get_online_mems() and put_online_mems() combination from all existing
platform ptdump code paths.
Link: https://lkml.kernel.org/r/20250620052427.2092093-1-anshuman.khandual@arm.com Fixes: bbd6ec605c0f ("arm64/mm: Enable memory hot remove") Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com> Acked-by: David Hildenbrand <david@redhat.com> Reviewed-by: Dev Jain <dev.jain@arm.com> Acked-by: Alexander Gordeev <agordeev@linux.ibm.com> [s390] Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Will Deacon <will@kernel.org> Cc: Ryan Roberts <ryan.roberts@arm.com> Cc: Paul Walmsley <paul.walmsley@sifive.com> Cc: Palmer Dabbelt <palmer@dabbelt.com> Cc: Alexander Gordeev <agordeev@linux.ibm.com> Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com> Cc: Heiko Carstens <hca@linux.ibm.com> Cc: Vasily Gorbik <gor@linux.ibm.com> Cc: Christian Borntraeger <borntraeger@linux.ibm.com> Cc: Sven Schnelle <svens@linux.ibm.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The slab allocator observes the task's NUMA policy in various places
such as allocating slab pages. Large kmalloc() allocations used to do
that too, until an unintended change by c4cab557521a ("mm/slab_common:
cleanup kmalloc_large()") resulted in ignoring mempolicy and just
preferring the local node. Restore the NUMA policy support.
Fixes: c4cab557521a ("mm/slab_common: cleanup kmalloc_large()") Cc: <stable@vger.kernel.org> Acked-by: Christoph Lameter (Ampere) <cl@gentwo.org> Acked-by: Roman Gushchin <roman.gushchin@linux.dev> Reviewed-by: Harry Yoo <harry.yoo@oracle.com> Signed-off-by: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The existing code move the VF NIC to new namespace when NETDEV_REGISTER is
received on netvsc NIC. During deletion of the namespace,
default_device_exit_batch() >> default_device_exit_net() is called. When
netvsc NIC is moved back and registered to the default namespace, it
automatically brings VF NIC back to the default namespace. This will cause
the default_device_exit_net() >> for_each_netdev_safe loop unable to detect
the list end, and hit NULL ptr:
Shuang reported sch_ets test-case [1] crashing in ets_class_qlen_notify()
after recent changes from Lion [2]. The problem is: in ets_qdisc_change()
we purge unused DWRR queues; the value of 'q->nbands' is the new one, and
the cleanup should be done with the old one. The problem is here since my
first attempts to fix ets_qdisc_change(), but it surfaced again after the
recent qdisc len accounting fixes. Fix it purging idle DWRR queues before
assigning a new value of 'q->nbands', so that all purge operations find a
consistent configuration:
- old 'q->nbands' because it's needed by ets_class_find()
- old 'q->nstrict' because it's needed by ets_class_is_strict()
This issue triggers when a userspace program does an ioctl
FBIOPUT_CON2FBMAP by passing console number and frame buffer number.
Ideally this maps console to frame buffer and updates the screen if
console is visible.
As part of mapping it has to do resize of console according to frame
buffer info. if this resize fails and returns from vc_do_resize() and
continues further. At this point console and new frame buffer are mapped
and sets display vars. Despite failure still it continue to proceed
updating the screen at later stages where vc_data is related to previous
frame buffer and frame buffer info and display vars are mapped to new
frame buffer and eventully leading to out-of-bounds write in
fast_imageblit(). This bheviour is excepted only when fg_console is
equal to requested console which is a visible console and updates screen
with invalid struct references in fbcon_putcs().
When UFFDIO_MOVE encounters a migration PMD entry, it proceeds with
obtaining a folio and accessing it even though the entry is swp_entry_t.
Add the missing check and let split_huge_pmd() handle migration entries.
While at it also remove unnecessary folio check.
[surenb@google.com: remove extra folio check, per David] Link: https://lkml.kernel.org/r/20250807200418.1963585-1-surenb@google.com Link: https://lkml.kernel.org/r/20250806220022.926763-1-surenb@google.com Fixes: adef440691ba ("userfaultfd: UFFDIO_MOVE uABI") Signed-off-by: Suren Baghdasaryan <surenb@google.com> Reported-by: syzbot+b446dbe27035ef6bd6c2@syzkaller.appspotmail.com Closes: https://lore.kernel.org/all/68794b5c.a70a0220.693ce.0050.GAE@google.com/ Reviewed-by: Peter Xu <peterx@redhat.com> Acked-by: David Hildenbrand <david@redhat.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Lokesh Gidra <lokeshgidra@google.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[BUG]
There is an internal report that balance triggered transaction abort,
with the following call trace:
item 85 key (594509824 169 0) itemoff 12599 itemsize 33
extent refs 1 gen 197740 flags 2
ref#0: tree block backref root 7
item 86 key (594558976 169 0) itemoff 12566 itemsize 33
extent refs 1 gen 197522 flags 2
ref#0: tree block backref root 7
...
BTRFS error (device loop0): extent item not found for insert, bytenr 594526208 num_bytes 16384 parent 449921024 root_objectid 934 owner 1 offset 0
BTRFS error (device loop0): failed to run delayed ref for logical 594526208 num_bytes 16384 type 182 action 1 ref_mod 1: -117
------------[ cut here ]------------
BTRFS: Transaction aborted (error -117)
WARNING: CPU: 1 PID: 6963 at ../fs/btrfs/extent-tree.c:2168 btrfs_run_delayed_refs+0xfa/0x110 [btrfs]
And btrfs check doesn't report anything wrong related to the extent
tree.
[CAUSE]
The cause is a little complex, firstly the extent tree indeed doesn't
have the backref for 594526208.
The extent tree only have the following two backrefs around that bytenr
on-disk:
item 65 key (594509824 METADATA_ITEM 0) itemoff 13880 itemsize 33
refs 1 gen 197740 flags TREE_BLOCK
tree block skinny level 0
(176 0x7) tree block backref root CSUM_TREE
item 66 key (594558976 METADATA_ITEM 0) itemoff 13847 itemsize 33
refs 1 gen 197522 flags TREE_BLOCK
tree block skinny level 0
(176 0x7) tree block backref root CSUM_TREE
But the such missing backref item is not an corruption on disk, as the
offending delayed ref belongs to subvolume 934, and that subvolume is
being dropped:
And that offending tree block 594526208 is inside the dropped range of
that subvolume. That explains why there is no backref item for that
bytenr and why btrfs check is not reporting anything wrong.
But this also shows another problem, as btrfs will do all the orphan
subvolume cleanup at a read-write mount.
So half-dropped subvolume should not exist after an RW mount, and
balance itself is also exclusive to subvolume cleanup, meaning we
shouldn't hit a subvolume half-dropped during relocation.
The root cause is, there is no orphan item for this subvolume.
In fact there are 5 subvolumes from around 2021 that have the same
problem.
It looks like the original report has some older kernels running, and
caused those zombie subvolumes.
Thankfully upstream commit 8d488a8c7ba2 ("btrfs: fix subvolume/snapshot
deletion not triggered on mount") has long fixed the bug.
[ENHANCEMENT]
For repairing such old fs, btrfs-progs will be enhanced.
Considering how delayed the problem will show up (at run delayed ref
time) and at that time we have to abort transaction already, it is too
late.
Instead here we reject any half-dropped subvolume for reloc tree at the
earliest time, preventing confusion and extra time wasted on debugging
similar bugs.
CC: stable@vger.kernel.org # 5.15+ 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>
__qgroup_excl_accounting() uses the qgroup iterator machinery to
update the account of one qgroups usage for all its parent hierarchy,
when we either add or remove a relation and have only exclusive usage.
However, there is a small bug there: we loop with an extra iteration
temporary qgroup called `cur` but never actually refer to that in the
body of the loop. As a result, we redundantly account the same usage to
the first qgroup in the list.
We call btrfs_zone_finish_one_bg() to zone finish one block group and make
room to activate another block group. Currently, we can choose a metadata
block group as a target. But, as we reserve an active metadata block group,
we no longer want to select a metadata block group. So, skip it in the
loop.
CC: stable@vger.kernel.org # 6.6+ Reviewed-by: Damien Le Moal <dlemoal@kernel.org> Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: Naohiro Aota <naohiro.aota@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>
Currently we only log an error message if we can't find the block group
for a log tree extent buffer when unaccounting it (while freeing a log
tree). A missing block group means something is seriously wrong and we
end up leaking space from the metadata space info. So return -ENOENT in
case we don't find the block group.
CC: stable@vger.kernel.org # 6.12+ Reviewed-by: Boris Burkov <boris@bur.io> Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.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>
If we log a new inode (not persisted in a past transaction) that has 0
links and extents, then log another inode with an higher inode number, we
end up with failing to replay the log tree with -EINVAL. The steps for
this are:
1) create new file A
2) write some data to file A
3) open an fd on file A
4) unlink file A
5) fsync file A using the previously open fd
6) create file B (has higher inode number than file A)
7) fsync file B
8) power fail before current transaction commits
Now when attempting to mount the fs, the log replay will fail with
-ENOENT at replay_one_extent() when attempting to replay the first
extent of file A. The failure comes when trying to open the inode for
file A in the subvolume tree, since it doesn't exist.
Before commit 5f61b961599a ("btrfs: fix inode lookup error handling
during log replay"), the returned error was -EIO instead of -ENOENT,
since we converted any errors when attempting to read an inode during
log replay to -EIO.
The reason for this is that the log replay procedure fails to ignore
the current inode when we are at the stage LOG_WALK_REPLAY_ALL, our
current inode has 0 links and last inode we processed in the previous
stage has a non 0 link count. In other words, the issue is that at
replay_one_extent() we only update wc->ignore_cur_inode if the current
replay stage is LOG_WALK_REPLAY_INODES.
Fix this by updating wc->ignore_cur_inode whenever we find an inode item
regardless of the current replay stage. This is a simple solution and easy
to backport, but later we can do other alternatives like avoid logging
extents or inode items other than the inode item for inodes with a link
count of 0.
The problem with the wc->ignore_cur_inode logic has been around since
commit f2d72f42d5fa ("Btrfs: fix warning when replaying log after fsync
of a tmpfile") but it only became frequent to hit since the more recent
commit 5e85262e542d ("btrfs: fix fsync of files with no hard links not
persisting deletion"), because we stopped skipping inodes with a link
count of 0 when logging, while before the problem would only be triggered
if trying to replay a log tree created with an older kernel which has a
logged inode with 0 links.
If we failed to insert the tree mod log operation, we are not removing the
dirty status from the allocated and dirtied extent buffer before we free
it. Removing the dirty status is needed for several reasons such as to
adjust the fs_info->dirty_metadata_bytes counter and remove the dirty
status from the respective folios. So add the missing call to
btrfs_clear_buffer_dirty().
Fixes: f61aa7ba08ab ("btrfs: do not BUG_ON() on tree mod log failure at insert_new_root()") CC: stable@vger.kernel.org # 6.6+ Reviewed-by: Boris Burkov <boris@bur.io> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
During log replay, at add_inode_ref(), if we have an extref item that
contains multiple extrefs and one of them points to a directory that does
not exist in the subvolume tree, we are supposed to ignore it and process
the remaining extrefs encoded in the extref item, since each extref can
point to a different parent inode. However when that happens we just
return from the function and ignore the remaining extrefs.
The problem has been around since extrefs were introduced, in commit f186373fef00 ("btrfs: extended inode refs"), but it's hard to hit in
practice because getting extref items encoding multiple extref requires
getting a hash collision when computing the offset of the extref's
key. The offset if computed like this:
When quotas are disabled qgroup ioctls are supposed to return -ENOTCONN,
but the qgroup create ioctl stopped doing that when it races with a quota
disable operation, returning 0 instead. This change of behaviour happened
in commit 6ed05643ddb1 ("btrfs: create qgroup earlier in snapshot
creation").
The issue happens as follows:
1) Task A enters btrfs_ioctl_qgroup_create(), qgroups are enabled and so
qgroup_enabled() returns true since fs_info->quota_root is not NULL;
2) Task B enters btrfs_ioctl_quota_ctl() -> btrfs_quota_disable() and
disables qgroups, so now fs_info->quota_root is NULL;
3) Task A enters btrfs_create_qgroup() and calls btrfs_qgroup_mode(),
which returns BTRFS_QGROUP_MODE_DISABLED since quotas are disabled,
and then btrfs_create_qgroup() returns 0 to the caller, which makes
the ioctl return 0 instead of -ENOTCONN.
The check for fs_info->quota_root and returning -ENOTCONN if it's NULL
is made only after the call btrfs_qgroup_mode().
Fix this by moving the check for disabled quotas with btrfs_qgroup_mode()
into transaction.c:create_pending_snapshot(), so that we don't abort the
transaction if btrfs_create_qgroup() returns -ENOTCONN and quotas are
disabled.
Fixes: 6ed05643ddb1 ("btrfs: create qgroup earlier in snapshot creation") CC: stable@vger.kernel.org # 6.12+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.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>
[TEST FAILURE WITH EXPERIMENTAL FEATURES]
When running test case generic/508, the test case will fail with the new
btrfs shutdown support:
generic/508 - output mismatch (see /home/adam/xfstests/results//generic/508.out.bad)
# --- tests/generic/508.out 2022-05-11 11:25:30.806666664 +0930
# +++ /home/adam/xfstests/results//generic/508.out.bad 2025-07-02 14:53:22.401824212 +0930
# @@ -1,2 +1,6 @@
# QA output created by 508
# Silence is golden
# +Before:
# +After : stat.btime = Thu Jan 1 09:30:00 1970
# +Before:
# +After : stat.btime = Wed Jul 2 14:53:22 2025
# ...
# (Run 'diff -u /home/adam/xfstests/tests/generic/508.out /home/adam/xfstests/results//generic/508.out.bad' to see the entire diff)
Ran: generic/508
Failures: generic/508
Failed 1 of 1 tests
Please note that the test case requires shutdown support, thus the test
case will be skipped using the current upstream kernel, as it doesn't
have shutdown ioctl support.
[CAUSE]
The direct cause the 0 time stamp in the log tree:
If the ssd_spread mount option is enabled, then we run the so called
clustered allocator for data block groups. In practice, this results in
creating a btrfs_free_cluster which caches a block_group and borrows its
free extents for allocation.
Since the introduction of allocation size classes in 6.1, there has been
a bug in the interaction between that feature and ssd_spread.
find_free_extent() has a number of nested loops. The loop going over the
allocation stages, stored in ffe_ctl->loop and managed by
find_free_extent_update_loop(), the loop over the raid levels, and the
loop over all the block_groups in a space_info. The size class feature
relies on the block_group loop to ensure it gets a chance to see a
block_group of a given size class. However, the clustered allocator
uses the cached cluster block_group and breaks that loop. Each call to
do_allocation() will really just go back to the same cached block_group.
Normally, this is OK, as the allocation either succeeds and we don't
want to loop any more or it fails, and we clear the cluster and return
its space to the block_group.
But with size classes, the allocation can succeed, then later fail,
outside of do_allocation() due to size class mismatch. That latter
failure is not properly handled due to the highly complex multi loop
logic. The result is a painful loop where we continue to allocate the
same num_bytes from the cluster in a tight loop until it fails and
releases the cluster and lets us try a new block_group. But by then, we
have skipped great swaths of the available block_groups and are likely
to fail to allocate, looping the outer loop. In pathological cases like
the reproducer below, the cached block_group is often the very last one,
in which case we don't perform this tight bg loop but instead rip
through the ffe stages to LOOP_CHUNK_ALLOC and allocate a chunk, which
is now the last one, and we enter the tight inner loop until an
allocation failure. Then allocation succeeds on the final block_group
and if the next allocation is a size mismatch, the exact same thing
happens again.
Triggering this is as easy as mounting with -o ssd_spread and then
running:
if you do the two writes + sync in a loop, you can force btrfs to spin
an excessive amount on semi-successful clustered allocations, before
ultimately failing and advancing to the stage where we force a chunk
allocation. This results in 2G of data allocated per iteration, despite
only using ~20M of data. By using a small size classed extent, the inner
loop takes longer and we can spin for longer.
The simplest, shortest term fix to unbreak this is to make the clustered
allocator size_class aware in the dumbest way, where it fails on size
class mismatch. This may hinder the operation of the clustered
allocator, but better hindered than completely broken and terribly
overallocating.
Further re-design improvements are also in the works.
Fixes: 52bb7a2166af ("btrfs: introduce size class to block group allocator") CC: stable@vger.kernel.org # 6.1+ Reported-by: David Sterba <dsterba@suse.com> Reviewed-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Boris Burkov <boris@bur.io> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
During log replay, at add_inode_ref(), we return -ENOENT if our current
inode isn't found on the subvolume tree or if a parent directory isn't
found. The error comes from btrfs_iget_logging() <- btrfs_iget() <-
btrfs_read_locked_inode().
The single caller of add_inode_ref(), replay_one_buffer(), ignores an
-ENOENT error because it expects that error to mean only that a parent
directory wasn't found and that is ok.
Before commit 5f61b961599a ("btrfs: fix inode lookup error handling during
log replay") we were converting any error when getting a parent directory
to -ENOENT and any error when getting the current inode to -EIO, so our
caller would fail log replay in case we can't find the current inode.
After that commit however in case the current inode is not found we return
-ENOENT to the caller and therefore it ignores the critical fact that the
current inode was not found in the subvolume tree.
Fix this by converting -ENOENT to 0 when we don't find a parent directory,
returning -ENOENT when we don't find the current inode and making the
caller, replay_one_buffer(), not ignore -ENOENT anymore.
Fixes: 5f61b961599a ("btrfs: fix inode lookup error handling during log replay") CC: stable@vger.kernel.org # 6.16 Reviewed-by: Boris Burkov <boris@bur.io> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Before waiting for the rescan worker to finish and flushing reservations,
we clear the BTRFS_FS_QUOTA_ENABLED flag from fs_info. If we fail flushing
reservations we leave with the flag not set which is not correct since
quotas are still enabled - we must set back the flag on error paths, such
as when we fail to start a transaction, except for error paths that abort
a transaction. The reservation flushing happens very early before we do
any operation that actually disables quotas and before we start a
transaction, so set back BTRFS_FS_QUOTA_ENABLED if it fails.
Fixes: af0e2aab3b70 ("btrfs: qgroup: flush reservations during quota disable") CC: stable@vger.kernel.org # 6.12+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.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>
There are some reports of "unable to find chunk map for logical 2147483648
length 16384" error message appears in dmesg. This means some IOs are
occurring after a block group is removed.
When a metadata tree node is cleaned on a zoned setup, we keep that node
still dirty and write it out not to create a write hole. However, this can
make a block group's used bytes == 0 while there is a dirty region left.
Such an unused block group is moved into the unused_bg list and processed
for removal. When the removal succeeds, the block group is removed from the
transaction->dirty_bgs list, so the unused dirty nodes in the block group
are not sent at the transaction commit time. It will be written at some
later time e.g, sync or umount, and causes "unable to find chunk map"
errors.
This can happen relatively easy on SMR whose zone size is 256MB. However,
calling do_zone_finish() on such block group returns -EAGAIN and keep that
block group intact, which is why the issue is hidden until now.
Fixes: afba2bc036b0 ("btrfs: zoned: implement active zone tracking") CC: stable@vger.kernel.org # 6.1+ 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>
If we failed walking a log tree during replay, we have a missing
transaction abort to prevent committing a transaction where we didn't
fully replay all the changes from a log tree and therefore can leave the
respective subvolume tree in some inconsistent state. So add the missing
transaction abort.
CC: stable@vger.kernel.org # 6.1+ Reviewed-by: Qu Wenruo <wqu@suse.com> Signed-off-by: Filipe Manana <fdmanana@suse.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 deciding if a zoned filesystem is reaching the threshold to reclaim
data block groups, look at the size of the filesystem not to potentially
total available size of all drives in the filesystem.
Especially if a filesystem was created with mkfs' -b option, constraining
it to only a portion of the block device, the numbers won't match and
potentially garbage collection is kicking in too late.
Fixes: 3687fcb0752a ("btrfs: zoned: make auto-reclaim less aggressive") CC: stable@vger.kernel.org # 6.1+ Reviewed-by: Damien Le Moal <dlemoal@kernel.org> Tested-by: Damien Le Moal <dlemoal@kernel.org> Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
projects / thirdparty / kernel / stable.git / log
