[ Upstream commit 773704c1ef ]
w/ below testcase, it will cause inconsistence in between SIT and SSA.
create_null_blk 512 2 1024 1024
mkfs.f2fs -m /dev/nullb0
mount /dev/nullb0 /mnt/f2fs/
touch /mnt/f2fs/file
f2fs_io pinfile set /mnt/f2fs/file
fallocate -l 4GiB /mnt/f2fs/file
F2FS-fs (nullb0): Inconsistent segment (0) type [1, 0] in SSA and SIT
CPU: 5 UID: 0 PID: 2398 Comm: fallocate Tainted: G O 6.13.0-rc1 #84
Tainted: [O]=OOT_MODULE
Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006
Call Trace:
<TASK>
dump_stack_lvl+0xb3/0xd0
dump_stack+0x14/0x20
f2fs_handle_critical_error+0x18c/0x220 [f2fs]
f2fs_stop_checkpoint+0x38/0x50 [f2fs]
do_garbage_collect+0x674/0x6e0 [f2fs]
f2fs_gc_range+0x12b/0x230 [f2fs]
f2fs_allocate_pinning_section+0x5c/0x150 [f2fs]
f2fs_expand_inode_data+0x1cc/0x3c0 [f2fs]
f2fs_fallocate+0x3c3/0x410 [f2fs]
vfs_fallocate+0x15f/0x4b0
__x64_sys_fallocate+0x4a/0x80
x64_sys_call+0x15e8/0x1b80
do_syscall_64+0x68/0x130
entry_SYSCALL_64_after_hwframe+0x67/0x6f
RIP: 0033:0x7f9dba5197ca
F2FS-fs (nullb0): Stopped filesystem due to reason: 4
The reason is f2fs_gc_range() may try to migrate block in curseg, however,
its SSA block is not uptodate due to the last summary block data is still
in cache of curseg.
In this patch, we add a condition in f2fs_gc_range() to check whether
section is opened or not, and skip block migration for opened section.
Fixes: 9703d69d9d ("f2fs: support file pinning for zoned devices")
Reviewed-by: Daeho Jeong <daehojeong@google.com>
Cc: Daeho Jeong <daehojeong@google.com>
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
[ Minor conflict resolved. ]
Signed-off-by: Li hongliang <1468888505@139.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 1bba3900ca ]
In the preparation stage of CPU online, if the corresponding
the rdp's->nocb_cb_kthread does not exist, will be created,
there is a situation where the rdp's rcuop kthreads creation fails,
and then de-offload this CPU's rdp, does not assign this CPU's
rdp->nocb_cb_kthread pointer, but this rdp's->nocb_gp_rdp and
rdp's->rdp_gp->nocb_gp_kthread is still valid.
This will cause the subsequent re-offload operation of this offline
CPU, which will pass the conditional check and the kthread_unpark()
will access invalid rdp's->nocb_cb_kthread pointer.
This commit therefore use rdp's->nocb_gp_kthread instead of
rdp_gp's->nocb_gp_kthread for safety check.
Signed-off-by: Zqiang <qiang.zhang1211@gmail.com>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Neeraj Upadhyay (AMD) <neeraj.upadhyay@kernel.org>
[ Minor conflict resolved. ]
Signed-off-by: Jianqiang kang <jianqkang@sina.cn>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit d55c571e43 ]
This script
#!/usr/bin/bash
echo 0 > /proc/sys/kernel/randomize_va_space
echo 'void main(void) {}' > TEST.c
# -fcf-protection to ensure that the 1st endbr32 insn can't be emulated
gcc -m32 -fcf-protection=branch TEST.c -o test
bpftrace -e 'uprobe:./test:main {}' -c ./test
"hangs", the probed ./test task enters an endless loop.
The problem is that with randomize_va_space == 0
get_unmapped_area(TASK_SIZE - PAGE_SIZE) called by xol_add_vma() can not
just return the "addr == TASK_SIZE - PAGE_SIZE" hint, this addr is used
by the stack vma.
arch_get_unmapped_area_topdown() doesn't take TIF_ADDR32 into account and
in_32bit_syscall() is false, this leads to info.high_limit > TASK_SIZE.
vm_unmapped_area() happily returns the high address > TASK_SIZE and then
get_unmapped_area() returns -ENOMEM after the "if (addr > TASK_SIZE - len)"
check.
handle_swbp() doesn't report this failure (probably it should) and silently
restarts the probed insn. Endless loop.
I think that the right fix should change the x86 get_unmapped_area() paths
to rely on TIF_ADDR32 rather than in_32bit_syscall(). Note also that if
CONFIG_X86_X32_ABI=y, in_x32_syscall() falsely returns true in this case
because ->orig_ax = -1.
But we need a simple fix for -stable, so this patch just sets TS_COMPAT if
the probed task is 32-bit to make in_ia32_syscall() true.
Fixes: 1b028f784e ("x86/mm: Introduce mmap_compat_base() for 32-bit mmap()")
Reported-by: Paulo Andrade <pandrade@redhat.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/all/aV5uldEvV7pb4RA8@redhat.com/
Cc: stable@vger.kernel.org
Link: https://patch.msgid.link/aWO7Fdxn39piQnxu@redhat.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 52b330799e ]
Exynos Virtual Display driver performs memory alloc/free operations
without lock protection, which easily causes concurrency problem.
For example, use-after-free can occur in race scenario like this:
```
CPU0 CPU1 CPU2
---- ---- ----
vidi_connection_ioctl()
if (vidi->connection) // true
drm_edid = drm_edid_alloc(); // alloc drm_edid
...
ctx->raw_edid = drm_edid;
...
drm_mode_getconnector()
drm_helper_probe_single_connector_modes()
vidi_get_modes()
if (ctx->raw_edid) // true
drm_edid_dup(ctx->raw_edid);
if (!drm_edid) // false
...
vidi_connection_ioctl()
if (vidi->connection) // false
drm_edid_free(ctx->raw_edid); // free drm_edid
...
drm_edid_alloc(drm_edid->edid)
kmemdup(edid); // UAF!!
...
```
To prevent these vulns, at least in vidi_context, member variables related
to memory alloc/free should be protected with ctx->lock.
Cc: <stable@vger.kernel.org>
Signed-off-by: Jeongjun Park <aha310510@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit d4c98c077c ]
In vidi_connection_ioctl(), vidi->edid(user pointer) is directly
dereferenced in the kernel.
This allows arbitrary kernel memory access from the user space, so instead
of directly accessing the user pointer in the kernel, we should modify it
to copy edid to kernel memory using copy_from_user() and use it.
Cc: <stable@vger.kernel.org>
Signed-off-by: Jeongjun Park <aha310510@gmail.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit d3968a0d85 ]
vidi_connection_ioctl() retrieves the driver_data from drm_dev->dev to
obtain a struct vidi_context pointer. However, drm_dev->dev is the
exynos-drm master device, and the driver_data contained therein is not
the vidi component device, but a completely different device.
This can lead to various bugs, ranging from null pointer dereferences and
garbage value accesses to, in unlucky cases, out-of-bounds errors,
use-after-free errors, and more.
To resolve this issue, we need to store/delete the vidi device pointer in
exynos_drm_private->vidi_dev during bind/unbind, and then read this
exynos_drm_private->vidi_dev within ioctl() to obtain the correct
struct vidi_context pointer.
Cc: <stable@vger.kernel.org>
Signed-off-by: Jeongjun Park <aha310510@gmail.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 5df0d6addb upstream.
If RAS header read from EEPROM is corrupted, it could result in trying
to allocate huge memory for reading the records. Add some validation to
header fields.
Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
[ RAS_TABLE_VER_V3 is not supported in v6.6.y. ]
Signed-off-by: Alva Lan <alvalan9@foxmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 9d73b107a6 upstream.
This commit is necessary for DC to function well with chips
that use the legacy power management code, ie. SI and KV.
Communicate display information from DC to the legacy PM code.
Currently DC uses pm_display_cfg to communicate power management
requirements from the display code to the DPM code.
However, the legacy (non-DC) code path used different fields
and therefore could not take into account anything from DC.
Change the legacy display code to fill the same pm_display_cfg
struct as DC and use the same in the legacy DPM code.
To ease review and reduce churn, this commit does not yet
delete the now unneeded code, that is done in the next commit.
v2:
Rebase.
Fix single_display in amdgpu_dpm_pick_power_state.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit b515dcb0dc upstream.
This commit adds the pixel_clock field to the display config
struct so that power management (DPM) can use it.
We currently don't have a proper bandwidth calculation on old
GPUs with DCE 6-10 because dce_calcs only supports DCE 11+.
So the power management (DPM) on these GPUs may need to make
ad-hoc decisions for display based on the pixel clock.
Also rename sym_clock to pixel_clock in dm_pp_single_disp_config
to avoid confusion with other code where the sym_clock refers to
the DisplayPort symbol clock.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 06e219f6a7 upstream.
Problem description
-------------------
DSA has a mumbo-jumbo of reference handling of the conduit net device
and its kobject which, sadly, is just wrong and doesn't make sense.
There are two distinct problems.
1. The OF path, which uses of_find_net_device_by_node(), never releases
the elevated refcount on the conduit's kobject. Nominally, the OF and
non-OF paths should result in objects having identical reference
counts taken, and it is already suspicious that
dsa_dev_to_net_device() has a put_device() call which is missing in
dsa_port_parse_of(), but we can actually even verify that an issue
exists. With CONFIG_DEBUG_KOBJECT_RELEASE=y, if we run this command
"before" and "after" applying this patch:
(unbind the conduit driver for net device eno2)
echo 0000:00:00.2 > /sys/bus/pci/drivers/fsl_enetc/unbind
we see these lines in the output diff which appear only with the patch
applied:
kobject: 'eno2' (ffff002009a3a6b8): kobject_release, parent 0000000000000000 (delayed 1000)
kobject: '109' (ffff0020099d59a0): kobject_release, parent 0000000000000000 (delayed 1000)
2. After we find the conduit interface one way (OF) or another (non-OF),
it can get unregistered at any time, and DSA remains with a long-lived,
but in this case stale, cpu_dp->conduit pointer. Holding the net
device's underlying kobject isn't actually of much help, it just
prevents it from being freed (but we never need that kobject
directly). What helps us to prevent the net device from being
unregistered is the parallel netdev reference mechanism (dev_hold()
and dev_put()).
Actually we actually use that netdev tracker mechanism implicitly on
user ports since commit 2f1e8ea726 ("net: dsa: link interfaces with
the DSA master to get rid of lockdep warnings"), via netdev_upper_dev_link().
But time still passes at DSA switch probe time between the initial
of_find_net_device_by_node() code and the user port creation time, time
during which the conduit could unregister itself and DSA wouldn't know
about it.
So we have to run of_find_net_device_by_node() under rtnl_lock() to
prevent that from happening, and release the lock only with the netdev
tracker having acquired the reference.
Do we need to keep the reference until dsa_unregister_switch() /
dsa_switch_shutdown()?
1: Maybe yes. A switch device will still be registered even if all user
ports failed to probe, see commit 86f8b1c01a ("net: dsa: Do not
make user port errors fatal"), and the cpu_dp->conduit pointers
remain valid. I haven't audited all call paths to see whether they
will actually use the conduit in lack of any user port, but if they
do, it seems safer to not rely on user ports for that reference.
2. Definitely yes. We support changing the conduit which a user port is
associated to, and we can get into a situation where we've moved all
user ports away from a conduit, thus no longer hold any reference to
it via the net device tracker. But we shouldn't let it go nonetheless
- see the next change in relation to dsa_tree_find_first_conduit()
and LAG conduits which disappear.
We have to be prepared to return to the physical conduit, so the CPU
port must explicitly keep another reference to it. This is also to
say: the user ports and their CPU ports may not always keep a
reference to the same conduit net device, and both are needed.
As for the conduit's kobject for the /sys/class/net/ entry, we don't
care about it, we can release it as soon as we hold the net device
object itself.
History and blame attribution
-----------------------------
The code has been refactored so many times, it is very difficult to
follow and properly attribute a blame, but I'll try to make a short
history which I hope to be correct.
We have two distinct probing paths:
- one for OF, introduced in 2016 in commit 83c0afaec7 ("net: dsa: Add
new binding implementation")
- one for non-OF, introduced in 2017 in commit 71e0bbde0d ("net: dsa:
Add support for platform data")
These are both complete rewrites of the original probing paths (which
used struct dsa_switch_driver and other weird stuff, instead of regular
devices on their respective buses for register access, like MDIO, SPI,
I2C etc):
- one for OF, introduced in 2013 in commit 5e95329b70 ("dsa: add
device tree bindings to register DSA switches")
- one for non-OF, introduced in 2008 in commit 91da11f870 ("net:
Distributed Switch Architecture protocol support")
except for tiny bits and pieces like dsa_dev_to_net_device() which were
seemingly carried over since the original commit, and used to this day.
The point is that the original probing paths received a fix in 2015 in
the form of commit 679fb46c57 ("net: dsa: Add missing master netdev
dev_put() calls"), but the fix never made it into the "new" (dsa2)
probing paths that can still be traced to today, and the fixed probing
path was later deleted in 2019 in commit 93e86b3bc8 ("net: dsa: Remove
legacy probing support").
That is to say, the new probing paths were never quite correct in this
area.
The existence of the legacy probing support which was deleted in 2019
explains why dsa_dev_to_net_device() returns a conduit with elevated
refcount (because it was supposed to be released during
dsa_remove_dst()). After the removal of the legacy code, the only user
of dsa_dev_to_net_device() calls dev_put(conduit) immediately after this
function returns. This pattern makes no sense today, and can only be
interpreted historically to understand why dev_hold() was there in the
first place.
Change details
--------------
Today we have a better netdev tracking infrastructure which we should
use. Logically netdev_hold() belongs in common code
(dsa_port_parse_cpu(), where dp->conduit is assigned), but there is a
tradeoff to be made with the rtnl_lock() section which would become a
bit too long if we did that - dsa_port_parse_cpu() also calls
request_module(). So we duplicate a bit of logic in order for the
callers of dsa_port_parse_cpu() to be the ones responsible of holding
the conduit reference and releasing it on error. This shortens the
rtnl_lock() section significantly.
In the dsa_switch_probe() error path, dsa_switch_release_ports() will be
called in a number of situations, one being where dsa_port_parse_cpu()
maybe didn't get the chance to run at all (a different port failed
earlier, etc). So we have to test for the conduit being NULL prior to
calling netdev_put().
There have still been so many transformations to the code since the
blamed commits (rename master -> conduit, commit 0650bf52b3 ("net:
dsa: be compatible with masters which unregister on shutdown")), that it
only makes sense to fix the code using the best methods available today
and see how it can be backported to stable later. I suspect the fix
cannot even be backported to kernels which lack dsa_switch_shutdown(),
and I suspect this is also maybe why the long-lived conduit reference
didn't make it into the new DSA probing paths at the time (problems
during shutdown).
Because dsa_dev_to_net_device() has a single call site and has to be
changed anyway, the logic was just absorbed into the non-OF
dsa_port_parse().
Tested on the ocelot/felix switch and on dsa_loop, both on the NXP
LS1028A with CONFIG_DEBUG_KOBJECT_RELEASE=y.
Reported-by: Ma Ke <make24@iscas.ac.cn>
Closes: https://lore.kernel.org/netdev/20251214131204.4684-1-make24@iscas.ac.cn/
Fixes: 83c0afaec7 ("net: dsa: Add new binding implementation")
Fixes: 71e0bbde0d ("net: dsa: Add support for platform data")
Reviewed-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Link: https://patch.msgid.link/20251215150236.3931670-1-vladimir.oltean@nxp.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
[ backport: "conduit" -> "master" in code, kept original commit message ]
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 6279846b9b upstream.
Syzbot reported a kernel warning due to a range invariant violation on
the following BPF program.
0: call bpf_get_netns_cookie
1: if r0 == 0 goto <exit>
2: if r0 & Oxffffffff goto <exit>
The issue is on the path where we fall through both jumps.
That path is unreachable at runtime: after insn 1, we know r0 != 0, but
with the sign extension on the jset, we would only fallthrough insn 2
if r0 == 0. Unfortunately, is_branch_taken() isn't currently able to
figure this out, so the verifier walks all branches. The verifier then
refines the register bounds using the second condition and we end
up with inconsistent bounds on this unreachable path:
1: if r0 == 0 goto <exit>
r0: u64=[0x1, 0xffffffffffffffff] var_off=(0, 0xffffffffffffffff)
2: if r0 & 0xffffffff goto <exit>
r0 before reg_bounds_sync: u64=[0x1, 0xffffffffffffffff] var_off=(0, 0)
r0 after reg_bounds_sync: u64=[0x1, 0] var_off=(0, 0)
Improving the range refinement for JSET to cover all cases is tricky. We
also don't expect many users to rely on JSET given LLVM doesn't generate
those instructions. So instead of improving the range refinement for
JSETs, Eduard suggested we forget the ranges whenever we're narrowing
tnums after a JSET. This patch implements that approach.
Reported-by: syzbot+c711ce17dd78e5d4fdcf@syzkaller.appspotmail.com
Suggested-by: Eduard Zingerman <eddyz87@gmail.com>
Acked-by: Yonghong Song <yonghong.song@linux.dev>
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Signed-off-by: Paul Chaignon <paul.chaignon@gmail.com>
Link: https://lore.kernel.org/r/9d4fd6432a095d281f815770608fdcd16028ce0b.1752171365.git.paul.chaignon@gmail.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
[ shung-hsi.yu: no detection or kernel warning for invariant violation before
6.8, but the same umin=1,umax=0 state can occur when jset is preceed by r0 < 1.
Changes were made to adapt to older range refinement logic before commit
67420501e8 ("bpf: generalize reg_set_min_max() to handle non-const register
comparisons"). ]
Signed-off-by: Shung-Hsi Yu <shung-hsi.yu@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 426ca15c7f ]
This patch enhances GSO segment handling by properly checking
the SKB_GSO_DODGY flag for frag_list GSO packets, addressing
low throughput issues observed when a station accesses IPv4
servers via hotspots with an IPv6-only upstream interface.
Specifically, it fixes a bug in GSO segmentation when forwarding
GRO packets containing a frag_list. The function skb_segment_list
cannot correctly process GRO skbs that have been converted by XLAT,
since XLAT only translates the header of the head skb. Consequently,
skbs in the frag_list may remain untranslated, resulting in protocol
inconsistencies and reduced throughput.
To address this, the patch explicitly sets the SKB_GSO_DODGY flag
for GSO packets in XLAT's IPv4/IPv6 protocol translation helpers
(bpf_skb_proto_4_to_6 and bpf_skb_proto_6_to_4). This marks GSO
packets as potentially modified after protocol translation. As a
result, GSO segmentation will avoid using skb_segment_list and
instead falls back to skb_segment for packets with the SKB_GSO_DODGY
flag. This ensures that only safe and fully translated frag_list
packets are processed by skb_segment_list, resolving protocol
inconsistencies and improving throughput when forwarding GRO packets
converted by XLAT.
Signed-off-by: Jibin Zhang <jibin.zhang@mediatek.com>
Fixes: 9fd1ff5d2a ("udp: Support UDP fraglist GRO/GSO.")
Cc: stable@vger.kernel.org
Link: https://patch.msgid.link/20260126152114.1211-1-jibin.zhang@mediatek.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Li hongliang <1468888505@139.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 17bd3bd82f ]
Detect tcp gso fraglist skbs with corrupted geometry (see below) and
pass these to skb_segment instead of skb_segment_list, as the first
can segment them correctly.
Valid SKB_GSO_FRAGLIST skbs
- consist of two or more segments
- the head_skb holds the protocol headers plus first gso_size
- one or more frag_list skbs hold exactly one segment
- all but the last must be gso_size
Optional datapath hooks such as NAT and BPF (bpf_skb_pull_data) can
modify these skbs, breaking these invariants.
In extreme cases they pull all data into skb linear. For TCP, this
causes a NULL ptr deref in __tcpv4_gso_segment_list_csum at
tcp_hdr(seg->next).
Detect invalid geometry due to pull, by checking head_skb size.
Don't just drop, as this may blackhole a destination. Convert to be
able to pass to regular skb_segment.
Approach and description based on a patch by Willem de Bruijn.
Link: https://lore.kernel.org/netdev/20240428142913.18666-1-shiming.cheng@mediatek.com/
Link: https://lore.kernel.org/netdev/20240922150450.3873767-1-willemdebruijn.kernel@gmail.com/
Fixes: bee88cd5bd ("net: add support for segmenting TCP fraglist GSO packets")
Cc: stable@vger.kernel.org
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20240926085315.51524-1-nbd@nbd.name
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Li hongliang <1468888505@139.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit a143545855 ]
Add proper locking in mmp_pdma_residue() to prevent use-after-free when
accessing descriptor list and descriptor contents.
The race occurs when multiple threads call tx_status() while the tasklet
on another CPU is freeing completed descriptors:
CPU 0 CPU 1
----- -----
mmp_pdma_tx_status()
mmp_pdma_residue()
-> NO LOCK held
list_for_each_entry(sw, ..)
DMA interrupt
dma_do_tasklet()
-> spin_lock(&desc_lock)
list_move(sw->node, ...)
spin_unlock(&desc_lock)
| dma_pool_free(sw) <- FREED!
-> access sw->desc <- UAF!
This issue can be reproduced when running dmatest on the same channel with
multiple threads (threads_per_chan > 1).
Fix by protecting the chain_running list iteration and descriptor access
with the chan->desc_lock spinlock.
Signed-off-by: Juan Li <lijuan@linux.spacemit.com>
Signed-off-by: Guodong Xu <guodong@riscstar.com>
Link: https://patch.msgid.link/20251216-mmp-pdma-race-v1-1-976a224bb622@riscstar.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
[ Minor context conflict resolved. ]
Signed-off-by: Wenshan Lan <jetlan9@163.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 25fd7ee7bf ]
The syscall number is a user-controlled value used to index into the
syscall table. Use array_index_nospec() to clamp this value after the
bounds check to prevent speculative out-of-bounds access and subsequent
data leakage via cache side channels.
Signed-off-by: Lukas Gerlach <lukas.gerlach@cispa.de>
Link: https://patch.msgid.link/20251218191332.35849-3-lukas.gerlach@cispa.de
Signed-off-by: Paul Walmsley <pjw@kernel.org>
[ Added linux/nospec.h for array_index_nospec() to make sure compile without error ]
Signed-off-by: Leon Chen <leonchen.oss@139.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 4e159150a9 ]
[BUG]
There is an internal report that over 1000 processes are
waiting at the io_schedule_timeout() of balance_dirty_pages(), causing
a system hang and trigger a kernel coredump.
The kernel is v6.4 kernel based, but the root problem still applies to
any upstream kernel before v6.18.
[CAUSE]
>From Jan Kara for his wisdom on the dirty page balance behavior first.
This cgroup dirty limit was what was actually playing the role here
because the cgroup had only a small amount of memory and so the dirty
limit for it was something like 16MB.
Dirty throttling is responsible for enforcing that nobody can dirty
(significantly) more dirty memory than there's dirty limit. Thus when
a task is dirtying pages it periodically enters into balance_dirty_pages()
and we let it sleep there to slow down the dirtying.
When the system is over dirty limit already (either globally or within
a cgroup of the running task), we will not let the task exit from
balance_dirty_pages() until the number of dirty pages drops below the
limit.
So in this particular case, as I already mentioned, there was a cgroup
with relatively small amount of memory and as a result with dirty limit
set at 16MB. A task from that cgroup has dirtied about 28MB worth of
pages in btrfs btree inode and these were practically the only dirty
pages in that cgroup.
So that means the only way to reduce the dirty pages of that cgroup is
to writeback the dirty pages of btrfs btree inode, and only after that
those processes can exit balance_dirty_pages().
Now back to the btrfs part, btree_writepages() is responsible for
writing back dirty btree inode pages.
The problem here is, there is a btrfs internal threshold that if the
btree inode's dirty bytes are below the 32M threshold, it will not
do any writeback.
This behavior is to batch as much metadata as possible so we won't write
back those tree blocks and then later re-COW them again for another
modification.
This internal 32MiB is higher than the existing dirty page size (28MiB),
meaning no writeback will happen, causing a deadlock between btrfs and
cgroup:
- Btrfs doesn't want to write back btree inode until more dirty pages
- Cgroup/MM doesn't want more dirty pages for btrfs btree inode
Thus any process touching that btree inode is put into sleep until
the number of dirty pages is reduced.
Thanks Jan Kara a lot for the analysis of the root cause.
[ENHANCEMENT]
Since kernel commit b55102826d ("btrfs: set AS_KERNEL_FILE on the
btree_inode"), btrfs btree inode pages will only be charged to the root
cgroup which should have a much larger limit than btrfs' 32MiB
threshold.
So it should not affect newer kernels.
But for all current LTS kernels, they are all affected by this problem,
and backporting the whole AS_KERNEL_FILE may not be a good idea.
Even for newer kernels I still think it's a good idea to get
rid of the internal threshold at btree_writepages(), since for most cases
cgroup/MM has a better view of full system memory usage than btrfs' fixed
threshold.
For internal callers using btrfs_btree_balance_dirty() since that
function is already doing internal threshold check, we don't need to
bother them.
But for external callers of btree_writepages(), just respect their
requests and write back whatever they want, ignoring the internal
btrfs threshold to avoid such deadlock on btree inode dirty page
balancing.
CC: stable@vger.kernel.org
CC: Jan Kara <jack@suse.cz>
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[ The context change is due to the commit 41044b41ad
("btrfs: add helper to get fs_info from struct inode pointer")
in v6.9 and the commit c66f2afc71
("btrfs: remove pointless writepages callback wrapper")
in v6.10 which are irrelevant to the logic of this patch. ]
Signed-off-by: Rahul Sharma <black.hawk@163.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 5d5fe8bcd3 ]
Fix the following:
BUG: KCSAN: data-race in rxrpc_peer_keepalive_worker / rxrpc_send_data_packet
which is reporting an issue with the reads and writes to ->last_tx_at in:
conn->peer->last_tx_at = ktime_get_seconds();
and:
keepalive_at = peer->last_tx_at + RXRPC_KEEPALIVE_TIME;
The lockless accesses to these to values aren't actually a problem as the
read only needs an approximate time of last transmission for the purposes
of deciding whether or not the transmission of a keepalive packet is
warranted yet.
Also, as ->last_tx_at is a 64-bit value, tearing can occur on a 32-bit
arch.
Fix both of these by switching to an unsigned int for ->last_tx_at and only
storing the LSW of the time64_t. It can then be reconstructed at need
provided no more than 68 years has elapsed since the last transmission.
Fixes: ace45bec6d ("rxrpc: Fix firewall route keepalive")
Reported-by: syzbot+6182afad5045e6703b3d@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/r/695e7cfb.050a0220.1c677c.036b.GAE@google.com/
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Marc Dionne <marc.dionne@auristor.com>
cc: Simon Horman <horms@kernel.org>
cc: linux-afs@lists.infradead.org
cc: stable@kernel.org
Link: https://patch.msgid.link/1107124.1768903985@warthog.procyon.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ The context change is due to the commit f3a123b254
("rxrpc: Allow the app to store private data on peer structs"),
the commit 372d12d191
("rxrpc: Add a reason indicator to the tx_data tracepoint"),
the commit 153f90a066
("rxrpc: Use ktimes for call timeout tracking and set the timer lazily")
and the commit 9d1d2b5934
("rxrpc: rxgk: Implement the yfs-rxgk security class (GSSAPI))
which are irrelevant to the logic of this patch. ]
Signed-off-by: Rahul Sharma <black.hawk@163.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit d9f3e47d3f ]
There are two problems with the recursive correction:
1. It may cause denial-of-service. In fec_read_bufs, there is a loop that
has 253 iterations. For each iteration, we may call verity_hash_for_block
recursively. There is a limit of 4 nested recursions - that means that
there may be at most 253^4 (4 billion) iterations. Red Hat QE team
actually created an image that pushes dm-verity to this limit - and this
image just makes the udev-worker process get stuck in the 'D' state.
2. It doesn't work. In fec_read_bufs we store data into the variable
"fio->bufs", but fio bufs is shared between recursive invocations, if
"verity_hash_for_block" invoked correction recursively, it would
overwrite partially filled fio->bufs.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reported-by: Guangwu Zhang <guazhang@redhat.com>
Reviewed-by: Sami Tolvanen <samitolvanen@google.com>
Reviewed-by: Eric Biggers <ebiggers@kernel.org>
[ The context change is due to the commit bdf253d580
("dm-verity: remove support for asynchronous hashes")
in v6.18 and the commit 9356fcfe0a
("dm verity: set DM_TARGET_SINGLETON feature flag") in v6.9
which are irrelevant to the logic of this patch. ]
Signed-off-by: Rahul Sharma <black.hawk@163.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 2c28769a51 ]
If rxrpc_recvmsg() fails because MSG_DONTWAIT was specified but the call at
the front of the recvmsg queue already has its mutex locked, it requeues
the call - whether or not the call is already queued. The call may be on
the queue because MSG_PEEK was also passed and so the call was not dequeued
or because the I/O thread requeued it.
The unconditional requeue may then corrupt the recvmsg queue, leading to
things like UAFs or refcount underruns.
Fix this by only requeuing the call if it isn't already on the queue - and
moving it to the front if it is already queued. If we don't queue it, we
have to put the ref we obtained by dequeuing it.
Also, MSG_PEEK doesn't dequeue the call so shouldn't call
rxrpc_notify_socket() for the call if we didn't use up all the data on the
queue, so fix that also.
Fixes: 540b1c48c3 ("rxrpc: Fix deadlock between call creation and sendmsg/recvmsg")
Reported-by: Faith <faith@zellic.io>
Reported-by: Pumpkin Chang <pumpkin@devco.re>
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Marc Dionne <marc.dionne@auristor.com>
cc: Nir Ohfeld <niro@wiz.io>
cc: Willy Tarreau <w@1wt.eu>
cc: Simon Horman <horms@kernel.org>
cc: linux-afs@lists.infradead.org
cc: stable@kernel.org
Link: https://patch.msgid.link/95163.1768428203@warthog.procyon.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[Use spin_unlock instead of spin_unlock_irq to maintain context consistency.]
Signed-off-by: Robert Garcia <rob_garcia@163.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 4865c768b5 ]
For filesystems with more than 2^32 blocks inodes using indirect block
based format cannot use blocks beyond the 32-bit limit.
ext4_mb_scan_groups_linear() takes care to not select these unsupported
groups for such inodes however other functions selecting groups for
allocation don't. So far this is harmless because the other selection
functions are used only with mb_optimize_scan and this is currently
disabled for inodes with indirect blocks however in the following patch
we want to enable mb_optimize_scan regardless of inode format.
Reviewed-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Zhang Yi <yi.zhang@huawei.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Acked-by: Pedro Falcato <pfalcato@suse.de>
Cc: stable@kernel.org
Link: https://patch.msgid.link/20260114182836.14120-3-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
[ Drop a few hunks not needed in older trees ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit f0aa6a37a3 ]
Recalculate features when XDP is detached.
Before:
# ip li set dev eth0 xdp obj xdp_dummy.bpf.o sec xdp
# ip li set dev eth0 xdp off
# ethtool -k eth0 | grep gro
rx-gro-hw: off [requested on]
After:
# ip li set dev eth0 xdp obj xdp_dummy.bpf.o sec xdp
# ip li set dev eth0 xdp off
# ethtool -k eth0 | grep gro
rx-gro-hw: on
The fact that HW-GRO doesn't get re-enabled automatically is just
a minor annoyance. The real issue is that the features will randomly
come back during another reconfiguration which just happens to invoke
netdev_update_features(). The driver doesn't handle reconfiguring
two things at a time very robustly.
Starting with commit 98ba1d931f ("bnxt_en: Fix RSS logic in
__bnxt_reserve_rings()") we only reconfigure the RSS hash table
if the "effective" number of Rx rings has changed. If HW-GRO is
enabled "effective" number of rings is 2x what user sees.
So if we are in the bad state, with HW-GRO re-enablement "pending"
after XDP off, and we lower the rings by / 2 - the HW-GRO rings
doing 2x and the ethtool -L doing / 2 may cancel each other out,
and the:
if (old_rx_rings != bp->hw_resc.resv_rx_rings &&
condition in __bnxt_reserve_rings() will be false.
The RSS map won't get updated, and we'll crash with:
BUG: kernel NULL pointer dereference, address: 0000000000000168
RIP: 0010:__bnxt_hwrm_vnic_set_rss+0x13a/0x1a0
bnxt_hwrm_vnic_rss_cfg_p5+0x47/0x180
__bnxt_setup_vnic_p5+0x58/0x110
bnxt_init_nic+0xb72/0xf50
__bnxt_open_nic+0x40d/0xab0
bnxt_open_nic+0x2b/0x60
ethtool_set_channels+0x18c/0x1d0
As we try to access a freed ring.
The issue is present since XDP support was added, really, but
prior to commit 98ba1d931f ("bnxt_en: Fix RSS logic in
__bnxt_reserve_rings()") it wasn't causing major issues.
Fixes: 1054aee823 ("bnxt_en: Use NETIF_F_GRO_HW.")
Fixes: 98ba1d931f ("bnxt_en: Fix RSS logic in __bnxt_reserve_rings()")
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Link: https://patch.msgid.link/20250109043057.2888953-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ The context change is due to the commit 1f6e77cb9b
("bnxt_en: Add bnxt_l2_filter hash table.") in v6.8 and the commit
8336a974f3 ("bnxt_en: Save user configured filters in a lookup list")
in v6.9 which are irrelevant to the logic of this patch. ]
Signed-off-by: Rahul Sharma <black.hawk@163.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 11bb2ffb67 ]
Commit '9329933699b3 ("soc: qcom: pmic_glink: Make client-lock
non-sleeping")' moved the pmic_glink client list under a spinlock, as it
is accessed by the rpmsg/glink callback, which in turn is invoked from
IRQ context.
This means that ucsi_unregister() is now called from atomic context,
which isn't feasible as it's expecting a sleepable context. An effort is
under way to get GLINK to invoke its callbacks in a sleepable context,
but until then lets schedule the unregistration.
A side effect of this is that ucsi_unregister() can now happen
after the remote processor, and thereby the communication link with it, is
gone. pmic_glink_send() is amended with a check to avoid the resulting NULL
pointer dereference.
This does however result in the user being informed about this error by
the following entry in the kernel log:
ucsi_glink.pmic_glink_ucsi pmic_glink.ucsi.0: failed to send UCSI write request: -5
Fixes: 9329933699 ("soc: qcom: pmic_glink: Make client-lock non-sleeping")
Cc: stable@vger.kernel.org
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Tested-by: Amit Pundir <amit.pundir@linaro.org>
Reviewed-by: Johan Hovold <johan+linaro@kernel.org>
Tested-by: Johan Hovold <johan+linaro@kernel.org>
Signed-off-by: Bjorn Andersson <quic_bjorande@quicinc.com>
Link: https://lore.kernel.org/r/20240820-pmic-glink-v6-11-races-v3-2-eec53c750a04@quicinc.com
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
[ The context change is due to the commit 584e8df589
("usb: typec: ucsi: extract common code for command handling")
in v6.11 which is irrelevant to the logic of this patch. ]
Signed-off-by: Rahul Sharma <black.hawk@163.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 857bf90562 ]
Ben Coddington reports seeing a hang in the following stack trace:
0 [ffffd0b50e1774e0] __schedule at ffffffff9ca05415
1 [ffffd0b50e177548] schedule at ffffffff9ca05717
2 [ffffd0b50e177558] bit_wait at ffffffff9ca061e1
3 [ffffd0b50e177568] __wait_on_bit at ffffffff9ca05cfb
4 [ffffd0b50e1775c8] out_of_line_wait_on_bit at ffffffff9ca05ea5
5 [ffffd0b50e177618] pnfs_roc at ffffffffc154207b [nfsv4]
6 [ffffd0b50e1776b8] _nfs4_proc_delegreturn at ffffffffc1506586 [nfsv4]
7 [ffffd0b50e177788] nfs4_proc_delegreturn at ffffffffc1507480 [nfsv4]
8 [ffffd0b50e1777f8] nfs_do_return_delegation at ffffffffc1523e41 [nfsv4]
9 [ffffd0b50e177838] nfs_inode_set_delegation at ffffffffc1524a75 [nfsv4]
10 [ffffd0b50e177888] nfs4_process_delegation at ffffffffc14f41dd [nfsv4]
11 [ffffd0b50e1778a0] _nfs4_opendata_to_nfs4_state at ffffffffc1503edf [nfsv4]
12 [ffffd0b50e1778c0] _nfs4_open_and_get_state at ffffffffc1504e56 [nfsv4]
13 [ffffd0b50e177978] _nfs4_do_open at ffffffffc15051b8 [nfsv4]
14 [ffffd0b50e1779f8] nfs4_do_open at ffffffffc150559c [nfsv4]
15 [ffffd0b50e177a80] nfs4_atomic_open at ffffffffc15057fb [nfsv4]
16 [ffffd0b50e177ad0] nfs4_file_open at ffffffffc15219be [nfsv4]
17 [ffffd0b50e177b78] do_dentry_open at ffffffff9c09e6ea
18 [ffffd0b50e177ba8] vfs_open at ffffffff9c0a082e
19 [ffffd0b50e177bd0] dentry_open at ffffffff9c0a0935
The issue is that the delegreturn is being asked to wait for a layout
return that cannot complete because a state recovery was initiated. The
state recovery cannot complete until the open() finishes processing the
delegations it was given.
The solution is to propagate the existing flags that indicate a
non-blocking call to the function pnfs_roc(), so that it knows not to
wait in this situation.
Reported-by: Benjamin Coddington <bcodding@hammerspace.com>
Fixes: 29ade5db12 ("pNFS: Wait on outstanding layoutreturns to complete in pnfs_roc()")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
[ Minor conflict resolved. ]
Signed-off-by: Li hongliang <1468888505@139.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit cce0be6eb4 ]
Wang Zhaolong reports a deadlock involving NFSv4.1 state recovery
waiting on kthreadd, which is attempting to reclaim memory by calling
nfs_release_folio(). The latter cannot make progress due to state
recovery being needed.
It seems that the only safe thing to do here is to kick off a writeback
of the folio, without waiting for completion, or else kicking off an
asynchronous commit.
Reported-by: Wang Zhaolong <wangzhaolong@huaweicloud.com>
Fixes: 96780ca55e ("NFS: fix up nfs_release_folio() to try to release the page")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
[ Minor conflict resolved. ]
Signed-off-by: Li hongliang <1468888505@139.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit fada32ed6d ]
nfs_folio_length is unsafe to use without having the folio locked and a
check for a NULL ->f_mapping that protects against truncations and can
lead to kernel crashes. E.g. when running xfstests generic/065 with
all nfs trace points enabled.
Follow the model of the XFS trace points and pass in an explіcit offset
and length. This has the additional benefit that these values can
be more accurate as some of the users touch partial folio ranges.
Fixes: eb5654b3b8 ("NFS: Enable tracing of nfs_invalidate_folio() and nfs_launder_folio()")
Reported-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
[ Minor conflict resolved. ]
Signed-off-by: Li hongliang <1468888505@139.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit f157dd6613 ]
When evicting an inode the first thing we do is to setup tracing for it,
which implies fetching the root's id. But in btrfs_evict_inode() the
root might be NULL, as implied in the next check that we do in
btrfs_evict_inode().
Hence, we either should set the ->root_objectid to 0 in case the root is
NULL, or we move tracing setup after checking that the root is not
NULL. Setting the rootid to 0 at least gives us the possibility to trace
this call even in the case when the root is NULL, so that's the solution
taken here.
Fixes: 1abe9b8a13 ("Btrfs: add initial tracepoint support for btrfs")
Reported-by: syzbot+d991fea1b4b23b1f6bf8@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=d991fea1b4b23b1f6bf8
Signed-off-by: Miquel Sabaté Solà <mssola@mssola.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[ Adjust context ]
Signed-off-by: Bin Lan <lanbincn@139.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 0e9df1c905 ]
During linear map pgtable creation, each pgtable is fixmapped /
fixunmapped twice; once during allocation to zero the memory, and a
again during population to write the entries. This means each table has
2 TLB invalidations issued against it. Let's fix this so that each table
is only fixmapped/fixunmapped once, halving the number of TLBIs, and
improving performance.
Achieve this by separating allocation and initialization (zeroing) of
the page. The allocated page is now fixmapped directly by the walker and
initialized, before being populated and finally fixunmapped.
This approach keeps the change small, but has the side effect that late
allocations (using __get_free_page()) must also go through the generic
memory clearing routine. So let's tell __get_free_page() not to zero the
memory to avoid duplication.
Additionally this approach means that fixmap/fixunmap is still used for
late pgtable modifications. That's not technically needed since the
memory is all mapped in the linear map by that point. That's left as a
possible future optimization if found to be needed.
Execution time of map_mem(), which creates the kernel linear map page
tables, was measured on different machines with different RAM configs:
| Apple M2 VM | Ampere Altra| Ampere Altra| Ampere Altra
| VM, 16G | VM, 64G | VM, 256G | Metal, 512G
---------------|-------------|-------------|-------------|-------------
| ms (%) | ms (%) | ms (%) | ms (%)
---------------|-------------|-------------|-------------|-------------
before | 11 (0%) | 161 (0%) | 656 (0%) | 1654 (0%)
after | 10 (-11%) | 104 (-35%) | 438 (-33%) | 1223 (-26%)
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
Suggested-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Itaru Kitayama <itaru.kitayama@fujitsu.com>
Tested-by: Eric Chanudet <echanude@redhat.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20240412131908.433043-4-ryan.roberts@arm.com
Signed-off-by: Will Deacon <will@kernel.org>
[ Ryan: Trivial backport ]
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 1fcb7cea8a ]
After removing uneccessary TLBIs, the next bottleneck when creating the
page tables for the linear map is DSB and ISB, which were previously
issued per-pte in __set_pte(). Since we are writing multiple ptes in a
given pte table, we can elide these barriers and insert them once we
have finished writing to the table.
Execution time of map_mem(), which creates the kernel linear map page
tables, was measured on different machines with different RAM configs:
| Apple M2 VM | Ampere Altra| Ampere Altra| Ampere Altra
| VM, 16G | VM, 64G | VM, 256G | Metal, 512G
---------------|-------------|-------------|-------------|-------------
| ms (%) | ms (%) | ms (%) | ms (%)
---------------|-------------|-------------|-------------|-------------
before | 78 (0%) | 435 (0%) | 1723 (0%) | 3779 (0%)
after | 11 (-86%) | 161 (-63%) | 656 (-62%) | 1654 (-56%)
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
Tested-by: Itaru Kitayama <itaru.kitayama@fujitsu.com>
Tested-by: Eric Chanudet <echanude@redhat.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20240412131908.433043-3-ryan.roberts@arm.com
Signed-off-by: Will Deacon <will@kernel.org>
[ Ryan: Trivial backport ]
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 5c63db59c5 ]
A large part of the kernel boot time is creating the kernel linear map
page tables. When rodata=full, all memory is mapped by pte. And when
there is lots of physical ram, there are lots of pte tables to populate.
The primary cost associated with this is mapping and unmapping the pte
table memory in the fixmap; at unmap time, the TLB entry must be
invalidated and this is expensive.
Previously, each pmd and pte table was fixmapped/fixunmapped for each
cont(pte|pmd) block of mappings (16 entries with 4K granule). This means
we ended up issuing 32 TLBIs per (pmd|pte) table during the population
phase.
Let's fix that, and fixmap/fixunmap each page once per population, for a
saving of 31 TLBIs per (pmd|pte) table. This gives a significant boot
speedup.
Execution time of map_mem(), which creates the kernel linear map page
tables, was measured on different machines with different RAM configs:
| Apple M2 VM | Ampere Altra| Ampere Altra| Ampere Altra
| VM, 16G | VM, 64G | VM, 256G | Metal, 512G
---------------|-------------|-------------|-------------|-------------
| ms (%) | ms (%) | ms (%) | ms (%)
---------------|-------------|-------------|-------------|-------------
before | 168 (0%) | 2198 (0%) | 8644 (0%) | 17447 (0%)
after | 78 (-53%) | 435 (-80%) | 1723 (-80%) | 3779 (-78%)
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
Tested-by: Itaru Kitayama <itaru.kitayama@fujitsu.com>
Tested-by: Eric Chanudet <echanude@redhat.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20240412131908.433043-2-ryan.roberts@arm.com
Signed-off-by: Will Deacon <will@kernel.org>
[ Ryan: Trivial backport ]
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit e1aa5ef892 upstream.
Current clk_csr_i setting of Loongson STMMAC (including LS7A1000/2000
and LS2K1000/2000/3000) are copy & paste from other drivers. In fact,
Loongson STMMAC use 125MHz clocks and need 62 freq division to within
2.5MHz, meeting most PHY MDC requirement. So fix by setting clk_csr_i
to 100-150MHz, otherwise some PHYs may link fail.
Cc: stable@vger.kernel.org
Fixes: 30bba69d7d ("stmmac: pci: Add dwmac support for Loongson")
Signed-off-by: Hongliang Wang <wanghongliang@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 968f19c5b1 upstream.
[BUG]
It is a long known bug that VM image on btrfs can lead to data csum
mismatch, if the qemu is using direct-io for the image (this is commonly
known as cache mode 'none').
[CAUSE]
Inside the VM, if the fs is EXT4 or XFS, or even NTFS from Windows, the
fs is allowed to dirty/modify the folio even if the folio is under
writeback (as long as the address space doesn't have AS_STABLE_WRITES
flag inherited from the block device).
This is a valid optimization to improve the concurrency, and since these
filesystems have no extra checksum on data, the content change is not a
problem at all.
But the final write into the image file is handled by btrfs, which needs
the content not to be modified during writeback, or the checksum will
not match the data (checksum is calculated before submitting the bio).
So EXT4/XFS/NTRFS assume they can modify the folio under writeback, but
btrfs requires no modification, this leads to the false csum mismatch.
This is only a controlled example, there are even cases where
multi-thread programs can submit a direct IO write, then another thread
modifies the direct IO buffer for whatever reason.
For such cases, btrfs has no sane way to detect such cases and leads to
false data csum mismatch.
[FIX]
I have considered the following ideas to solve the problem:
- Make direct IO to always skip data checksum
This not only requires a new incompatible flag, as it breaks the
current per-inode NODATASUM flag.
But also requires extra handling for no csum found cases.
And this also reduces our checksum protection.
- Let hardware handle all the checksum
AKA, just nodatasum mount option.
That requires trust for hardware (which is not that trustful in a lot
of cases), and it's not generic at all.
- Always fallback to buffered write if the inode requires checksum
This was suggested by Christoph, and is the solution utilized by this
patch.
The cost is obvious, the extra buffer copying into page cache, thus it
reduces the performance.
But at least it's still user configurable, if the end user still wants
the zero-copy performance, just set NODATASUM flag for the inode
(which is a common practice for VM images on btrfs).
Since we cannot trust user space programs to keep the buffer
consistent during direct IO, we have no choice but always falling back
to buffered IO. At least by this, we avoid the more deadly false data
checksum mismatch error.
Cc: stable@vger.kernel.org # 6.6
[ Conflicts caused by code extracted into direct-io.c ]
Suggested-by: Christoph Hellwig <hch@infradead.org>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@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>
[ Upstream commit 94a8cea54c ]
fstests test generic/388 occasionally reproduces a warning in
ext4_put_super() associated with the dirty clusters count:
WARNING: CPU: 7 PID: 76064 at fs/ext4/super.c:1324 ext4_put_super+0x48c/0x590 [ext4]
Tracing the failure shows that the warning fires due to an
s_dirtyclusters_counter value of -1. IOW, this appears to be a
spurious decrement as opposed to some sort of leak. Further tracing
of the dirty cluster count deltas and an LLM scan of the resulting
output identified the cause as a double decrement in the error path
between ext4_mb_mark_diskspace_used() and the caller
ext4_mb_new_blocks().
First, note that generic/388 is a shutdown vs. fsstress test and so
produces a random set of operations and shutdown injections. In the
problematic case, the shutdown triggers an error return from the
ext4_handle_dirty_metadata() call(s) made from
ext4_mb_mark_context(). The changed value is non-zero at this point,
so ext4_mb_mark_diskspace_used() does not exit after the error
bubbles up from ext4_mb_mark_context(). Instead, the former
decrements both cluster counters and returns the error up to
ext4_mb_new_blocks(). The latter falls into the !ar->len out path
which decrements the dirty clusters counter a second time, creating
the inconsistency.
To avoid this problem and simplify ownership of the cluster
reservation in this codepath, lift the counter reduction to a single
place in the caller. This makes it more clear that
ext4_mb_new_blocks() is responsible for acquiring cluster
reservation (via ext4_claim_free_clusters()) in the !delalloc case
as well as releasing it, regardless of whether it ends up consumed
or returned due to failure.
Fixes: 0087d9fb3f ("ext4: Fix s_dirty_blocks_counter if block allocation failed with nodelalloc")
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Baokun Li <libaokun1@huawei.com>
Link: https://patch.msgid.link/20260113171905.118284-1-bfoster@redhat.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
[ Drop mballoc-test changes ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit d625a2b08c ]
It reports a bug from device w/ zufs:
F2FS-fs (dm-64): Inconsistent segment (173822) type [1, 0] in SSA and SIT
F2FS-fs (dm-64): Stopped filesystem due to reason: 4
Thread A Thread B
- f2fs_expand_inode_data
- f2fs_allocate_pinning_section
- f2fs_gc_range
- do_garbage_collect w/ segno #x
- writepage
- f2fs_allocate_data_block
- new_curseg
- allocate segno #x
The root cause is: fallocate on pinning file may race w/ block allocation
as above, result in do_garbage_collect() from fallocate() may migrate
segment which is just allocated by a log, the log will update segment type
in its in-memory structure, however GC will get segment type from on-disk
SSA block, once segment type changes by log, we can detect such
inconsistency, then shutdown filesystem.
In this case, on-disk SSA shows type of segno #173822 is 1 (SUM_TYPE_NODE),
however segno #173822 was just allocated as data type segment, so in-memory
SIT shows type of segno #173822 is 0 (SUM_TYPE_DATA).
Change as below to fix this issue:
- check whether current section is empty before gc
- add sanity checks on do_garbage_collect() to avoid any race case, result
in migrating segment used by log.
- btw, it fixes misc issue in printed logs: "SSA and SIT" -> "SIT and SSA".
Fixes: 9703d69d9d ("f2fs: support file pinning for zoned devices")
Cc: Daeho Jeong <daehojeong@google.com>
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
[ Use IS_CURSEC instead of is_cursec according to
commit c1cfc87e49 ("f2fs: introduce is_cur{seg,sec}()"). ]
Signed-off-by: Robert Garcia <rob_garcia@163.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 186ac39b8a ]
In xfs_qm_dqflush(), when a dquot flush fails due to corruption
(the out_abort error path), the original code removed the dquot log
item from the AIL before calling xfs_force_shutdown(). This ordering
introduces a subtle race condition that can lead to data loss after
a crash.
The AIL tracks the oldest dirty metadata in the journal. The position
of the tail item in the AIL determines the log tail LSN, which is the
oldest LSN that must be preserved for crash recovery. When an item is
removed from the AIL, the log tail can advance past the LSN of that item.
The race window is as follows: if the dquot item happens to be at
the tail of the log, removing it from the AIL allows the log tail
to advance. If a concurrent log write is sampling the tail LSN at
the same time and subsequently writes a complete checkpoint (i.e.,
one containing a commit record) to disk before the shutdown takes
effect, the journal will no longer protect the dquot's last
modification. On the next mount, log recovery will not replay the
dquot changes, even though they were never written back to disk,
resulting in silent data loss.
Fix this by calling xfs_force_shutdown() before xfs_trans_ail_delete()
in the out_abort path. Once the log is shut down, no new log writes
can complete with an updated tail LSN, making it safe to remove the
dquot item from the AIL.
Cc: stable@vger.kernel.org
Fixes: b707fffda6 ("xfs: abort consistently on dquot flush failure")
Signed-off-by: Long Li <leo.lilong@huawei.com>
Reviewed-by: Carlos Maiolino <cmaiolino@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
[ adapted error path to preserve existing out_unlock label between xfs_trans_ail_delete and xfs_dqfunlock ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 362c490980 ]
xfs_bmap_update_diff_items() sorts bmap intents by inode number using
a subtraction of two xfs_ino_t (uint64_t) values, with the result
truncated to int. This is incorrect when two inode numbers differ by
more than INT_MAX (2^31 - 1), which is entirely possible on large XFS
filesystems.
Fix this by replacing the subtraction with cmp_int().
Cc: <stable@vger.kernel.org> # v4.9
Fixes: 9f3afb57d5 ("xfs: implement deferred bmbt map/unmap operations")
Signed-off-by: Long Li <leo.lilong@huawei.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
[ replaced `bi_entry()` macro with `container_of()` and inlined `cmp_int()` as a manual three-way comparison expression ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 340cea84f6 ]
Today whenever we deal with a file, in addition to holding
a reference on the dentry, we also get a reference on the
superblock. This happens in two cases:
1. when a new cinode is allocated
2. when an oplock break is being processed
The reasoning for holding the superblock ref was to make sure
that when umount happens, if there are users of inodes and
dentries, it does not try to clean them up and wait for the
last ref to superblock to be dropped by last of such users.
But the side effect of doing that is that umount silently drops
a ref on the superblock and we could have deferred closes and
lease breaks still holding these refs.
Ideally, we should ensure that all of these users of inodes and
dentries are cleaned up at the time of umount, which is what this
code is doing.
This code change allows these code paths to use a ref on the
dentry (and hence the inode). That way, umount is
ensured to clean up SMB client resources when it's the last
ref on the superblock (For ex: when same objects are shared).
The code change also moves the call to close all the files in
deferred close list to the umount code path. It also waits for
oplock_break workers to be flushed before calling
kill_anon_super (which eventually frees up those objects).
Fixes: 24261fc23d ("cifs: delay super block destruction until all cifsFileInfo objects are gone")
Fixes: 705c79101c ("smb: client: fix use-after-free in cifs_oplock_break")
Cc: <stable@vger.kernel.org>
Signed-off-by: Shyam Prasad N <sprasad@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
[ adapted kmalloc_obj() macro to kmalloc(sizeof()) ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[ Upstream commit 881a0263d5 ]
Quanyang observed that when using an NFS rootfs on an AMD ZynqMp board,
the rootfs may take an extended time to recover after a suspend.
Upon investigation, it was determined that the issue originates from a
problem in the macb driver.
According to the Zynq UltraScale TRM [1], when transmit is disabled,
the transmit buffer queue pointer resets to point to the address
specified by the transmit buffer queue base address register.
In the current implementation, the code merely resets `queue->tx_head`
and `queue->tx_tail` to '0'. This approach presents several issues:
- Packets already queued in the tx ring are silently lost,
leading to memory leaks since the associated skbs cannot be released.
- Concurrent write access to `queue->tx_head` and `queue->tx_tail` may
occur from `macb_tx_poll()` or `macb_start_xmit()` when these values
are reset to '0'.
- The transmission may become stuck on a packet that has already been sent
out, with its 'TX_USED' bit set, but has not yet been processed. However,
due to the manipulation of 'queue->tx_head' and 'queue->tx_tail',
`macb_tx_poll()` incorrectly assumes there are no packets to handle
because `queue->tx_head == queue->tx_tail`. This issue is only resolved
when a new packet is placed at this position. This is the root cause of
the prolonged recovery time observed for the NFS root filesystem.
To resolve this issue, shuffle the tx ring and tx skb array so that
the first unsent packet is positioned at the start of the tx ring.
Additionally, ensure that updates to `queue->tx_head` and
`queue->tx_tail` are properly protected with the appropriate lock.
[1] https://docs.amd.com/v/u/en-US/ug1085-zynq-ultrascale-trm
Fixes: bf9cf80cab ("net: macb: Fix tx/rx malfunction after phy link down and up")
Reported-by: Quanyang Wang <quanyang.wang@windriver.com>
Signed-off-by: Kevin Hao <haokexin@gmail.com>
Cc: stable@vger.kernel.org
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20260307-zynqmp-v2-1-6ef98a70e1d0@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ #include context ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Chao Yu
Zqiang
Antheas Kapenekakis
Oleg Nesterov
Jeongjun Park
Lijo Lazar
Eric Dumazet
Timur Kristóf
Vladimir Oltean
Paul Chaignon
Jibin Zhang
Felix Fietkau
Steven Rostedt
Guodong Xu
Lukas Gerlach
Qu Wenruo
Christoph Hellwig
David Howells
Tom Lendacky
Borislav Petkov (AMD)
Mikulas Patocka
Jan Kara
Jakub Kicinski
Bjorn Andersson
Trond Myklebust
Miquel Sabaté Solà
Ryan Roberts
Huacai Chen
Brian Foster
Eric Biggers
Long Li
Thorsten Blum
Shyam Prasad N
Kevin Hao