| Age | Commit message (Collapse) | Author |
|---|
|
The dst parameter of aspeed_spi_user_transfer_tx() is an MMIO address
obtained from chip->ahb_base, but it was typed as void * instead of
void __iomem *. This caused a sparse warning report. Fix the
parameter type to void __iomem * and drop the now-unnecessary
cast at the call site.
Reported-by: kernel test robot <lkp@intel.com>
Closes: https://lore.kernel.org/oe-kbuild-all/202605180441.uD3toFRJ-lkp@intel.com/
Signed-off-by: Chin-Ting Kuo <chin-ting_kuo@aspeedtech.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Link: https://patch.msgid.link/20260522071621.102507-2-chin-ting_kuo@aspeedtech.com
Signed-off-by: Mark Brown <broonie@kernel.org>
|
|
For lshift and rshift, the shift operations are performed in a loop over
32-bit words. The loop calculates the shifted value and write it to dst,
and then immediately reads from src to calculate the carry for the next
iteration. Because src and dst could point to the same memory location,
the carry is incorrectly calculated using the newly modified dst value
instead of the original src value.
Adding a temporary local variable to cache the original value before
writing to dst and using it for the carry calculation solves the
problem. In addition, partial overlap is rejected from control plane for
all kind of operations including byteorder. This was tested with the
following bytecode:
table test_table ip flags 0 use 1 handle 1
ip test_table test_chain use 3 type filter hook input prio 0 policy accept packets 0 bytes 0 flags 1
ip test_table test_chain 2
[ immediate reg 1 0x44332211 0x88776655 ]
[ bitwise reg 1 = ( reg 1 << 0x08000000 ) ]
[ cmp eq reg 1 0x66443322 0x00887766 ]
[ counter pkts 0 bytes 0 ]
ip test_table test_chain 4 3
[ immediate reg 1 0x44332211 0x88776655 ]
[ bitwise reg 1 = ( reg 1 << 0x08000000 ) ]
[ cmp eq reg 1 0x55443322 0x00887766 ]
[ counter pkts 21794 bytes 1917798 ]
Fixes: 567d746b55bc ("netfilter: bitwise: add support for shifts.")
Acked-by: Jeremy Sowden <jeremy@azazel.net>
Signed-off-by: Fernando Fernandez Mancera <fmancera@suse.de>
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
Functional coverage of nft_fib6_eval()'s nexthop enumeration over
three route shapes:
1) single external nexthop (nhid)
2) external nexthop group (nhid -> group)
3) old-style multipath (nexthop ... nexthop ...)
Each scenario places one nexthop on the input device (veth0). For
(2) and (3) the matching nexthop is the second member, so the walk
has to traverse beyond the primary nh. Two nft counters on prerouting
verify the data path: one increments only when fib reports veth0 as
the oif, the other counts "missing" results and must stay at zero.
./nft_fib_nexthop.sh
PASS: single external nexthop (nhid -> veth0)
PASS: nexthop group (dummy0 + veth0)
PASS: old-style multipath (sibling on veth0)
Suggested-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
fib6_info has a union:
union {
struct list_head fib6_siblings;
struct list_head nh_list;
};
Old-style multipath (ip -6 route add ... nexthop ... nexthop ...) uses
fib6_siblings. External nexthop (ip -6 route add ... nhid N) uses
nh_list, linked into &nh->f6i_list.
nft_fib6_info_nh_uses_dev() blindly walks &rt->fib6_siblings, causing
an OOB read past the struct nexthop slab when rt->nh is set:
==================================================================
BUG: KASAN: slab-out-of-bounds in nft_fib6_eval+0x1362/0x16c0
Read of size 8 at addr ffff888103a099d0 by task ping/386
CPU: 2 UID: 0 PID: 386 Comm: ping Not tainted 7.1.0-rc3+ #251 PREEMPT
Call Trace:
<IRQ>
dump_stack_lvl+0x76/0xa0
print_report+0xd1/0x5f0
kasan_report+0xe7/0x130
__asan_report_load8_noabort+0x14/0x30
nft_fib6_eval+0x1362/0x16c0
nft_do_chain+0x279/0x18c0
nft_do_chain_ipv6+0x1a8/0x230
nf_hook_slow+0xad/0x200
ipv6_rcv+0x152/0x380
__netif_receive_skb_one_core+0x118/0x1c0
==================================================================
Branch by route shape: when rt->nh is set, walk via
nexthop_for_each_fib6_nh() (also covers nh groups, which the original
code missed); otherwise walk fib6_siblings, guarded by READ_ONCE() of
rt->fib6_nsiblings as required by commit 31d7d67ba127 ("ipv6: annotate
data-races around rt->fib6_nsiblings").
Fixes: 1c32b24c234b ("netfilter: nft_fib_ipv6: switch to fib6_lookup")
Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
nft_fib6_info_nh_uses_dev() runs from nft_fib6_eval() in softirq under
rcu_read_lock(). fib6_siblings is modified by writers that hold
tb6_lock but do not wait for RCU readers, so the sibling walk should
use list_for_each_entry_rcu(): it adds READ_ONCE() on the ->next
pointer and lets CONFIG_PROVE_RCU_LIST validate the locking.
No functional change for non-debug builds.
Fixes: 1c32b24c234b ("netfilter: nft_fib_ipv6: switch to fib6_lookup")
Signed-off-by: Jiayuan Chen <jiayuan.chen@linux.dev>
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
Luxiao Xu says:
The function compat_mtw_from_user() converts ebtables extensions from
32-bit user structures to kernel native structures. However, it lacks
proper validation of the user-supplied match_size/target_size.
When certain extensions are processed, the kernel-side translation
logic may perform memory accesses based on the extension's expected
size. If the user provides a size smaller than what the extension
requires, it results in an out-of-bounds read as reported by KASAN.
This fix introduces a check to ensure match_size is at least as large
as the extension's required compatsize. This covers matches, watchers,
and targets, while maintaining compatibility with standard targets.
AFAIU this is relevant for matches that need to go though
match->compat_from_user() call. Those that use plain memcpy with the
user-provided size are ok because the caller checks that size vs the
start of the next rule entry offset (which itself is checked vs. total
size copied from userspace).
The ->compat_from_user() callbacks assume they can read compatsize bytes,
so they need this extra check.
Based on an earlier patch from Luxiao Xu.
Fixes: 81e675c227ec ("netfilter: ebtables: add CONFIG_COMPAT support")
Reported-by: Yuan Tan <yuantan098@gmail.com>
Reported-by: Yifan Wu <yifanwucs@gmail.com>
Reported-by: Juefei Pu <tomapufckgml@gmail.com>
Reported-by: Xin Liu <bird@lzu.edu.cn>
Signed-off-by: Luxiao Xu <rakukuip@gmail.com>
Signed-off-by: Ren Wei <n05ec@lzu.edu.cn>
Reviewed-by: Fernando Fernandez Mancera <fmancera@suse.de>
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
Several parts of network stack rely on iph->ihl validation
done by network stack before PRE_ROUTING.
Disable this feature for user namespaces for now.
tcp option handling is likely safe even for LOCAL_IN, so this
this leaves tcp option mangling via nft_exthdr.c as-is.
I don't think these are the only means to alter packets, but these
appear to be relatively prominent.
This could be relaxed later. Example:
- allow userns for ingress hook.
- allow userns if base is transport header.
Also, we should revalidate or restrict generally:
- Don't allow linklayer writes to spill into network header
- restrict ipv4 and ipv6 to 'known safe' writes, e.g.
saddr/daddr/check/tos
Reported-by: Qi Tang <tpluszz77@gmail.com>
Reported-by: Tong Liu <lyutoon@gmail.com>
Tested-by: Qi Tang <tpluszz77@gmail.com>
Link: https://lore.kernel.org/netfilter-devel/20260515100411.3141-1-fw@strlen.de/
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
With PREEMPT_RCU we get splat:
BUG: using smp_processor_id() in preemptible [..]
caller is cpu_mt+0x53/0xd0 net/netfilter/xt_cpu.c:37
CPU: 1 .. Comm: syz.3.1377 #0 PREEMPT(full)
Call Trace:
<TASK>
dump_stack_lvl+0xe8/0x150 lib/dump_stack.c:120
check_preemption_disabled+0xd3/0xe0 lib/smp_processor_id.c:47
cpu_mt+0x53/0xd0 net/netfilter/xt_cpu.c:37
[..]
Just use raw version instead.
This is similar to 14d14a5d2957 ("netfilter: nft_meta: use raw_smp_processor_id()").
Fixes: 0ca743a55991 ("netfilter: nf_tables: add compatibility layer for x_tables")
Reported-by: syzbot+690d3e3ffa7335ac10eb@syzkaller.appspotmail.com
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
Quoting reporter:
A race between GRE keymap insertion and destruction can corrupt the
kernel list or use a freed object. `nf_ct_gre_keymap_add()` publishes a
new keymap pointer before the embedded `list_head` is linked, while
`nf_ct_gre_keymap_destroy()` can concurrently delete and free that
same object. An unprivileged user can reach this through the PPTP
conntrack helper by racing PPTP control messages or helper teardown,
leading to KASAN-detectable list corruption/UAF in kernel context.
## Root Cause Analysis
`exp_gre()` installs GRE expectations for a PPTP control flow and then
adds two GRE keymap entries [..]
The add path publishes `ct_pptp_info->keymap[dir]` before linking the
embedded list node [..]
Concurrent teardown deletes that partially initialized object.
Make add/destroy symmetric: install both, destroy both while under lock.
Furthermore, we should refuse to publish a new mapping in case ct is going
away, else we may leak the allocation.
The "retrans" detection is strange: existing mapping is checked for key
equality with the new mapping, then for "is on the list" via list walk.
But I can't see how an existing keymap entry can be NOT on list.
Change this to only check if we're asked to map same tuple again -- if so,
skip re-install, else signal failure.
Last, add a bug trap for the keymap list; it has to be empty when namespace
is going away.
Reported-by: Leo Lin <leo@depthfirst.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
synproxy_tstamp_adjust() rewrites the TCP timestamp option in place
and then patches the TCP checksum via inet_proto_csum_replace4() on
the caller-supplied tcphdr pointer. Both ipv4_synproxy_hook() and
ipv6_synproxy_hook() obtain that pointer with skb_header_pointer()
before calling in, so it may either alias skb->head directly or
point at the caller's on-stack _tcph buffer.
Between obtaining the pointer and using it, the function calls
skb_ensure_writable(skb, optend), which on a cloned or non-linear
skb invokes pskb_expand_head() and frees the old skb->head. After
that point the cached th is stale:
caller (ipv[46]_synproxy_hook)
th = skb_header_pointer(skb, ..., &_tcph)
synproxy_tstamp_adjust(skb, protoff, th, ...)
skb_ensure_writable(skb, optend)
pskb_expand_head() /* kfree(old skb->head) */
...
inet_proto_csum_replace4(&th->check, ...)
/* writes into freed head, or
into the caller's stack copy
leaving the on-wire checksum
stale */
The option bytes are written through skb->data and are fine; only
the checksum update goes through th and so lands in the wrong
place. The result is either a write into freed slab memory or a
packet leaving with a checksum that does not match its payload.
Fix by re-deriving th from skb->data + protoff immediately after
skb_ensure_writable() succeeds, so the subsequent checksum update
targets the linear, writable header.
Fixes: 48b1de4c110a ("netfilter: add SYNPROXY core/target")
Assisted-by: kres (claude-opus-4-7)
Signed-off-by: Chris Mason <clm@meta.com>
Reviewed-by: Fernando Fernandez Mancera <fmancera@suse.de>
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
direction check
An unintended behavior in the TCP conntrack state machine allows a
connection to be forced into the CLOSE state using an RST packet with an
invalid sequence number.
Specifically, after a SYN packet is observed, an RST with an invalid SEQ
can transition the conntrack entry to TCP_CONNTRACK_CLOSE, regardless of
whether the RST corresponds to the expected reply direction. The relevant
code path assumes the RST is a response to an outgoing SYN, but does not
validate packet direction or ensure that a matching SYN was actually sent
in the opposite direction.
As a result, a crafted packet sequence consisting of a SYN followed by an
invalid-sequence RST can prematurely terminate an active NAT entry. This
makes connection teardown easier than intended.
So, tighten the state transition logic to ensure that RST-triggered
CLOSE transitions only occur when the RST is a valid response to a
previously observed SYN in the correct direction.
Cc: stable@vger.kernel.org
Fixes: 9fb9cbb1082d ("[NETFILTER]: Add nf_conntrack subsystem.")
Signed-off-by: Hamza Mahfooz <hamzamahfooz@linux.microsoft.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
|
|
If a firmware node is allocated on the stack (for instance: temporary
software node whose life-time we control) or on the heap - but using a
non-zeroing allocation function - and initialized using fwnode_init(),
its secondary pointer will contain uninitalized memory which likely will
be neither NULL nor IS_ERR() and so may end up being dereferenced (for
example: in dev_to_swnode()). Set fwnode->secondary to NULL on
initialization.
Cc: stable <stable@kernel.org>
Fixes: 01bb86b380a3 ("driver core: Add fwnode_init()")
Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Reviewed-by: Rafael J. Wysocki (Intel) <rafael@kernel.org>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Link: https://patch.msgid.link/20260506115701.23035-1-bartosz.golaszewski@oss.qualcomm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
At this point the superblock is already read, so jump to the label that
releases the buffer_head for it.
Fixes: d893fc670546 ("minix: handle set_blocksize failures")
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://patch.msgid.link/20260518130330.529085-1-hch@lst.de
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
|
|
After a kexec/kdump reboot, the macb Ethernet controller fails to
receive any packets, causing DHCP to hang indefinitely and the network
interface to be unusable despite link being up.
The root cause is that RP1's level-triggered MSI-X interrupt sources
(such as macb on hwirq 6) may have their internal state machines stuck
in the "waiting for IACK" state. This happens because the previous
kernel crashed before sending the acknowledgment for a pending level
interrupt.
In this stuck state, RP1 will not generate new MSI-X writes even though
the interrupt source remains asserted. Since no new MSI-X is sent, the
GIC never sees a new edge, the chained IRQ handler is never invoked,
and the interrupt is permanently lost.
Fix this by sending MSIX_CFG_IACK in rp1_irq_activate(). This
unconditionally resets the MSI-X state machine back to idle when a
child device requests its interrupt. If the interrupt source is still
asserted, RP1 will immediately issue a new MSI-X with the freshly
configured msg_addr/msg_data, and normal interrupt delivery resumes.
Writing IACK when the state machine is already idle (i.e., on a normal
cold boot) is harmless — it has no effect.
Fixes: 49d63971f963 ("misc: rp1: RaspberryPi RP1 misc driver")
Cc: stable <stable@kernel.org>
Signed-off-by: Xiaolei Wang <xiaolei.wang@windriver.com>
Link: https://patch.msgid.link/20260518073405.2115003-1-xiaolei.wang@windriver.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
When request_irq() fails, the region allocated by request_region()
is not released. Fix this by adding an error handling path with
proper goto labels to release the region.
Fixes: e9dc69956d4d ("staging: gpib: Add Computer Boards GPIB driver")
Closes: https://lore.kernel.org/oe-kbuild-all/202605160620.ReBOadPX-lkp@intel.com/
Signed-off-by: Hongling Zeng <zenghongling@kylinos.cn>
Cc: stable <stable@kernel.org>
Link: https://patch.msgid.link/20260518022939.16881-1-zenghongling@kylinos.cn
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
The parport subsystem registers port devices before they are fully
initialised, resulting in a race condition where client drivers such
as lp can attach to ports that are not completely initialised or even
being torn down.
When the port and client drivers are built as modules and loaded
around the same time during boot, this occasionally results in a
crash. I was able to make this happen reliably in a VM with a
PC-style parallel port by patching parport_pc to fail probing:
> --- a/drivers/parport/parport_pc.c
> +++ b/drivers/parport/parport_pc.c
> @@ -2069,7 +2069,7 @@ static struct parport *__parport_pc_probe_port(unsigned long int base,
> if (!p)
> goto out3;
>
> - base_res = request_region(base, 3, p->name);
> + base_res = NULL;
> if (!base_res)
> goto out4;
>
and then running:
while true; do
modprobe lp & modprobe parport_pc
wait
rmmod lp parport_pc
done
for a few seconds.
In the long term I think port registration should be changed to put
the call to device_add() inside parport_announce_port(), but since the
latter currently cannot fail this will require changing all port
drivers.
For now, add a flag to indicate whether a port has been "announced"
and only try to attach client drivers to ports when the flag is set.
Fixes: 6fa45a226897 ("parport: add device-model to parport subsystem")
Closes: https://bugs.debian.org/1130365
Closes: https://lore.kernel.org/all/6ba903ad-9897-42bb-8c2d-337385cc3746@molgen.mpg.de/
Cc: stable <stable@kernel.org>
Signed-off-by: Ben Hutchings <benh@debian.org>
Acked-by: Sudip Mukherjee <sudipm.mukherjee@gmail.com>
Link: https://patch.msgid.link/afo6uBv68GDevbMD@decadent.org.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
uio_pci_sva allocates struct uio_pci_sva_dev with devm_kzalloc() in
probe(), but then calls kfree(udev) both on the probe() error path
(label out_free) and again in remove().
Because devm_kzalloc() allocations are devres-managed and are freed
automatically when the device is detached (including after a failing
probe() and during driver unbind), the explicit kfree() can lead to a
double free.
If probe() fails after devm_kzalloc(), the error path frees udev and
devres cleanup will free it again when the core unwinds the partially
bound device. On normal driver removal, remove() frees udev and devres
will free it again when the device is detached.
This issue was identified by a static analysis tool I developed and
confirmed by manual review. Fix by removing the manual kfree() calls
and dropping the now-unused label.
Fixes: 3397c3cd859a2 ("uio: Add SVA support for PCI devices via uio_pci_generic_sva.c")
Cc: stable <stable@kernel.org>
Signed-off-by: Guangshuo Li <lgs201920130244@gmail.com>
Link: https://patch.msgid.link/20260505150256.614071-1-lgs201920130244@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
When huge_pmd_unshare() is called to unshare a PMD table, the
tlb_unshare_pmd_ptdesc() function sets tlb->unshared_tables=true
but the aarch64 tlb_flush() only checked tlb->freed_tables to
determine whether to use TLBF_NONE (vae1is, invalidates walk
cache) or TLBF_NOWALKCACHE (vale1is, leaf-only).
This caused the stale PMD page table entry to remain in the walk cache
after unshare, potentially leading to incorrect page table walks.
Fix by including unshared_tables in the check, so that when
unsharing tables, TLBF_NONE is used and the walk cache is properly
invalidated.
Here is the detailed distinction between vae1is and vale1is:
| Instruction Combination | Actual Invalidation Scope |
| ------------------------ | --------------------------------------------------|
| `VAE1IS` + TTL=`0` | All entries at all levels (full invalidation) |
| `VAE1IS` + TTL=`2` (L2) | Non-leaf at Level 0/1 + leaf at Level 2 |
| `VALE1IS` + TTL=`0` | Leaf entries at all levels (non-leaf not cleared) |
| `VALE1IS` + TTL=`2` (L2) | Leaf entry at Level 2 only |
Signed-off-by: Zeng Heng <zengheng4@huawei.com>
Fixes: 8ce720d5bd91 ("mm/hugetlb: fix excessive IPI broadcasts when unsharing PMD tables using mmu_gather")
Cc: <stable@vger.kernel.org>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
|
|
Several places in init/do_mounts.c allocate temporary buffers for
filesystem names or options using __get_free_page() or alloc_page().
Usage of alloc_page() APIs is not required there and only creates
unnecessary noise with castings or conversion from struct page to void *.
kmalloc() is a better API for these uses and it also provides better
scalability and more debugging possibilities.
Replace use of __get_free_page() and alloc_page() with kmalloc().
While on it, add a check for -ENOMEM condition in mount_root_generic().
Link: https://lore.kernel.org/all/635405e4-9423-4a25-a6e7-e03c8ea0bcbe@redhat.com
Signed-off-by: Mike Rapoport (Microsoft) <rppt@kernel.org>
Link: https://patch.msgid.link/20260520-init-v1-1-aaf2ebac5ad9@kernel.org
Reviewed-by: David Disseldorp <ddiss@suse.de>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
|
|
and inode_switch_wbs()"
Baokun Li <libaokun@linux.alibaba.com> says:
When a container exits, a race between cgroup_writeback_umount() and
inode_switch_wbs() / cleanup_offline_cgwb() can trigger
"VFS: Busy inodes after unmount" followed by a use-after-free on
percpu counters.
There is a window between inode_prepare_wbs_switch() returning true
(having passed the SB_ACTIVE check and grabbed the inode) and the
subsequent wb_queue_isw() call. If cgroup_writeback_umount() observes
the global isw_nr_in_flight counter as non-zero but flush_workqueue()
finds nothing queued, it returns early -- leaving a held inode
reference that blocks evict_inodes() and a later iput() that hits
freed percpu counters.
Patch 1 closes the race by extending the RCU read-side critical
section to cover the window from inode_prepare_wbs_switch() through
wb_queue_isw(), and adding synchronize_rcu() in the umount path so
that all in-flight switchers complete queueing before
flush_workqueue() runs. rcu_barrier() is intentionally retained so
the same hunk applies cleanly to stable trees that still queue
switches via queue_rcu_work().
Patch 2 removes the now-dead rcu_barrier() that was left over from
the queue_rcu_work() era (replaced by plain queue_work() in commit
e1b849cfa6b6 "writeback: Avoid contention on wb->list_lock when
switching inodes"). This is mainline-only.
Patch 3 replaces the global synchronize_rcu()/flush_workqueue() pair
with a per-sb counter (s_isw_nr_in_flight) plus three small helpers
(cgroup_writeback_pin / cgroup_writeback_unpin /
cgroup_writeback_drain), eliminating the global serialization
penalty. This also reverts the RCU extension from patch 1 since the
per-sb counter makes it unnecessary.
Performance
-----------
Measured on a 16 vCPU QEMU guest, all kernels share the same .config.
Background load: 4 ext4 superblocks each running
while :; do
mkdir /sys/fs/cgroup/<tag>-tmp$N
( echo $BASHPID > <tag>-tmp$N/cgroup.procs
dd if=/dev/zero of=$mp/burner bs=4k count=256 conv=notrunc \
oflag=sync)
rmdir /sys/fs/cgroup/<tag>-tmp$N
done
This drives both inode_switch_wbs() (different cgroups writing the
same inode) and cleanup_offline_cgwb() (dying memcgs), keeping the
global isw_nr_in_flight non-zero throughout the run. Latencies are
wall-clock around umount(8) on a separate target sb; only the target
sb's umount is measured.
Four kernels are compared at each step of the series:
base pre-fix mainline
+race base + patch 1 (race fix, keeps rcu_barrier)
+rmbarrier +race + patch 2 (drop rcu_barrier)
+persb +rmbarrier + patch 3 (per-sb counter)
Target sb runs its own cgwb churn:
p50 p95 p99 max
base 99.7 ms 112.9 ms 112.9 ms 127.2 ms
+race 110.2 ms 153.8 ms 153.8 ms 160.4 ms
+rmbarrier 67.6 ms 88.3 ms 88.3 ms 96.8 ms
+persb 7.9 ms 10.0 ms 10.0 ms 10.1 ms
Idle target umount under cross-sb cgwb-switch pressure:
p50 p95 p99 max
base 92.0 ms 123.5 ms 136.5 ms 141.3 ms
+race 118.8 ms 154.6 ms 164.7 ms 165.3 ms
+rmbarrier 62.7 ms 95.4 ms 108.1 ms 108.6 ms
+persb 5.3 ms 6.9 ms 7.4 ms 7.4 ms
8 concurrent umounts of idle sbs under the same pressure:
p50 p95 p99 max
base 137.5 ms 166.9 ms 166.9 ms 171.3 ms
+race 162.2 ms 183.9 ms 183.9 ms 217.0 ms
+rmbarrier 61.3 ms 99.5 ms 99.5 ms 113.7 ms
+persb 8.1 ms 9.1 ms 9.1 ms 9.5 ms
A no-pressure baseline run (no background load) measures ~5 ms p50
across all four kernels, validating that the methodology has no
systematic bias.
In-kernel cgroup_writeback_umount() cumulative cost across the same
run (bpftrace, ~340 calls covering all four scenarios):
cgroup_writeback_umount() time
base 21240 ms total (~62 ms / call)
+race (rcu_barrier+sync) 24966 ms total (~73 ms / call)
+rmbarrier (synchronize_rcu) 12371 ms total (~36 ms / call)
+persb (per-sb counter) 1.37 ms total ( ~4 us / call)
Under +persb the wait_var_event() condition is true on entry
whenever the target sb has nothing in flight, so synchronize_rcu()
and flush_workqueue() are never called on this path.
Notes:
- Patch 1 adds ~10-27 ms p50 over base by introducing
synchronize_rcu(). This is the cost of closing the race
correctly and is paid by stable backports as well.
- Patch 2 ("drop rcu_barrier()") was expected to be a pure cleanup
on mainline, but actually removes a real wait: rcu_barrier()
drains call_rcu() callbacks from *all* subsystems, and the
cgroup teardown path keeps that pipeline busy under this
workload. Removing it cuts ~43-101 ms p50 on top of patch 1.
- Patch 3 (per-sb counter) replaces the global wait entirely; the
target sb no longer waits for activity on unrelated sbs,
recovering near-baseline latency in all three scenarios.
* patches from https://patch.msgid.link/20260521095016.2791354-1-libaokun@linux.alibaba.com:
writeback: use a per-sb counter to drain inode wb switches at umount
writeback: drop now-unnecessary rcu_barrier() in cgroup_writeback_umount()
writeback: fix race between cgroup_writeback_umount() and inode_switch_wbs()
Link: https://patch.msgid.link/20260521095016.2791354-1-libaokun@linux.alibaba.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Tracking in-flight inode wb switches with a single global counter
(isw_nr_in_flight) plus a synchronize_rcu() based wait in
cgroup_writeback_umount() forces every umount to take a global hit
whenever any other superblock on the system has wb switches in flight,
even if the superblock being unmounted has none of its own.
Replace the global synchronize_rcu()/flush_workqueue() pair with a
per-sb counter, s_isw_nr_in_flight, plus three small helpers:
- cgroup_writeback_pin(sb) - increment counter
- cgroup_writeback_unpin(sb) - decrement and wake drainer if last
- cgroup_writeback_drain(sb) - wait for counter to reach zero
The wiring is:
- inode_prepare_wbs_switch() pins before checking SB_ACTIVE and
grabbing the inode; failure paths unpin before returning. A
lockless SB_ACTIVE check at the top of the function lets us skip
the atomic_inc/smp_mb dance once SB_ACTIVE has been cleared (it
is monotonic and never set back).
- process_inode_switch_wbs() unpins after the matching iput().
- cgroup_writeback_umount() drains the per-sb counter via
wait_var_event().
The smp_mb() pair between inode_prepare_wbs_switch() and
cgroup_writeback_umount() keeps the SB_ACTIVE / counter ordering:
either the umounter sees a non-zero counter and waits, or the
switcher sees SB_ACTIVE cleared and aborts before grabbing the
inode.
The global isw_nr_in_flight is left in place, since it is still used
to throttle in-flight switches via WB_FRN_MAX_IN_FLIGHT.
The rcu_read_lock() extension in inode_switch_wbs() and
cleanup_offline_cgwb() that the race fix added is no longer needed
and is reverted; the synchronize_rcu() that the race fix added to
cgroup_writeback_umount() is dropped as well.
The following numbers were measured on a 16 vCPU QEMU guest with 4
background superblocks each churning "create memcg -> write 1 MiB ->
rmdir memcg" to keep the global isw_nr_in_flight non-zero. Latencies
are wall-clock around umount(8); only the target sb's umount is
measured.
Target sb runs its own cgwb churn:
p50 p95 p99 max
global synchronize_rcu() 67.6 ms 88.3 ms 88.3 ms 96.8 ms
per-sb counter (this) 7.9 ms 10.0 ms 10.0 ms 10.1 ms
Idle target umount latency under cross-sb cgwb-switch pressure:
p50 p95 p99 max
global synchronize_rcu() 62.7 ms 95.4 ms 108.1 ms 108.6 ms
per-sb counter (this) 5.3 ms 6.9 ms 7.4 ms 7.4 ms
no-pressure baseline 4.9 ms 5.9 ms 6.3 ms 6.7 ms
8 concurrent umounts of idle sbs under the same pressure:
p50 p95 max
global synchronize_rcu() 61.3 ms 99.5 ms 113.7 ms
per-sb counter (this) 8.1 ms 9.1 ms 9.5 ms
In-kernel cgroup_writeback_umount() time across the same run
(bpftrace, ~340 calls covering all scenarios):
global synchronize_rcu() 12371 ms total (~36 ms / call)
per-sb counter (this) 1.37 ms total ( ~4 us / call)
Suggested-by: Christian Brauner <brauner@kernel.org>
Link: https://lore.kernel.org/r/177910456953.488929.2169908940676707307.b4-review@b4
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Baokun Li <libaokun@linux.alibaba.com>
Link: https://patch.msgid.link/20260521095016.2791354-4-libaokun@linux.alibaba.com
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
|
|
Commit e1b849cfa6b6 ("writeback: Avoid contention on wb->list_lock when
switching inodes") replaced the queue_rcu_work() based scheduling of
inode wb switches with a plain queue_work(). Since then no switcher
goes through call_rcu(), so rcu_barrier() in cgroup_writeback_umount()
has no callbacks of its own to wait for. It still drains unrelated
call_rcu() callbacks from other subsystems on busy systems, which
incidentally slows umount down; drop it.
Fixes: e1b849cfa6b6 ("writeback: Avoid contention on wb->list_lock when switching inodes")
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Baokun Li <libaokun@linux.alibaba.com>
Link: https://patch.msgid.link/20260521095016.2791354-3-libaokun@linux.alibaba.com
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
|
|
When a container exits, the following BUG_ON() is occasionally triggered:
==================================================================
VFS: Busy inodes after unmount of sdb (ext4)
------------[ cut here ]------------
kernel BUG at fs/super.c:695!
CPU: 3 PID: 6 Comm: containerd-shim Tainted: G OE K 6.6 #1
pstate: 63400009 (nZCv daif +PAN -UAO +TCO +DIT -SSBS BTYPE=--)
pc : generic_shutdown_super+0xf0/0x100
lr : generic_shutdown_super+0xf0/0x100
Call trace:
generic_shutdown_super+0xf0/0x100
kill_block_super+0x20/0x48
ext4_kill_sb+0x28/0x60
deactivate_locked_super+0x54/0x130
deactivate_super+0x84/0xa0
cleanup_mnt+0xa4/0x140
__cleanup_mnt+0x18/0x28
task_work_run+0x78/0xe0
do_notify_resume+0x204/0x240
==================================================================
The root cause is a race between cgroup_writeback_umount() and
inode_switch_wbs()/cleanup_offline_cgwb(). There is a window between
inode_prepare_wbs_switch() returning true and the subsequent
wb_queue_isw() call. Following is the process that triggers the issue:
CPU A (umount) | CPU B (writeback)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
inode_switch_wbs/cleanup_offline_cgwb
atomic_inc(&isw_nr_in_flight)
inode_prepare_wbs_switch
-> passes SB_ACTIVE check
__iget(inode)
generic_shutdown_super
sb->s_flags &= ~SB_ACTIVE
cgroup_writeback_umount(sb)
smp_mb()
atomic_read(&isw_nr_in_flight)
rcu_barrier()
-> no pending RCU callbacks
flush_workqueue(isw_wq)
-> nothing queued, returns
evict_inodes(sb)
-> Inode skipped as isw still holds a ref.
sop->put_super(sb)
/* destroys percpu counters */
-> VFS: Busy inodes after unmount!
wb_queue_isw()
queue_work(isw_wq, ...)
/* later in work function */
inode_switch_wbs_work_fn
process_inode_switch_wbs
iput() -> evict
percpu_counter_dec() // UAF!
Fix this by extending the RCU read-side critical section in
inode_switch_wbs() and cleanup_offline_cgwb() to cover from
inode_prepare_wbs_switch() through wb_queue_isw(). Since there is
no sleep in this window, rcu_read_lock() can be used. Then add a
synchronize_rcu() in cgroup_writeback_umount() before the existing
rcu_barrier(), so that all in-flight switchers that have passed the
SB_ACTIVE check have completed queue_work() before flush_workqueue()
is called.
The existing rcu_barrier() is intentionally retained so this fix can
be backported unchanged to stable kernels (5.10.y, 6.6.y, ...) that
still queue switches via queue_rcu_work(). It is a no-op on current
mainline (since commit e1b849cfa6b6 ("writeback: Avoid contention on
wb->list_lock when switching inodes")) and is removed in a follow-up
patch.
Fixes: a1a0e23e4903 ("writeback: flush inode cgroup wb switches instead of pinning super_block")
Cc: stable@vger.kernel.org
Suggested-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/all/mxnjq2l6guusfchvauxr3v7c4bwjasybxlleqbbh4efloeqspz@iqylk76ohufz
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Baokun Li <libaokun@linux.alibaba.com>
Link: https://patch.msgid.link/20260521095016.2791354-2-libaokun@linux.alibaba.com
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
|
|
In 6c37bebd8c926, we switched to looping over the list and dropping each
individual node, ostensibly without the lock held in the loop body.
If the kernel were using Rust Edition 2024, the comment would be
accurate, and the lock would not be held across the drop. However, the
kernel is currently using 2021, so tail expression lifetime extension
results in the lock being held across the drop. Explicitly binding the
expression result to a variable makes the lockguard no longer part of a
tail expression, causing the lock to be dropped before entering the loop
body.
This was detected via `CONFIG_PROVE_LOCKING` identifying an invalid wait
context at the drop site.
Reported-by: David Stevens <stevensd@google.com>
Signed-off-by: Matthew Maurer <mmaurer@google.com>
Cc: stable <stable@kernel.org>
Fixes: 6c37bebd8c92 ("rust_binder: avoid mem::take on delivered_deaths")
Reviewed-by: Alice Ryhl <aliceryhl@google.com>
Acked-by: Carlos Llamas <cmllamas@google.com>
Link: https://patch.msgid.link/20260403-lockhold-v1-1-c332b56cd8ae@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
When an outdated transaction is removed from `oneway_todo` due to
`TF_UPDATE_TXN`, its `Allocation` is dropped. The current implementation
of `Allocation::drop` calls `pending_oneway_finished()`, assuming the
transaction was executed. This leads to premature execution of the next
queued one-way transaction.
Fix this by taking the `oneway_node` from the `Allocation` of the
outdated transaction before it is dropped. This prevents
`Allocation::drop` from signaling completion.
We do not call `take_oneway_node()` from `Transaction::cancel` because
it's actually correct to call `pending_oneway_finished()` on cancel if
the transaction did not come from `oneway_todo`. This ensures that if
`BINDER_THREAD_EXIT` is invoked and cancels a oneway transaction, then
the next transaction is taken from `oneway_todo`.
This bug does not lead to any issues in the kernel, but may lead to
Binder delivering transactions to userspace earlier than userspace
expected to receive them.
Cc: stable <stable@kernel.org>
Fixes: eafedbc7c050 ("rust_binder: add Rust Binder driver")
Assisted-by: Antigravity:gemini
Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Acked-by: Carlos Llamas <cmllamas@google.com>
Link: https://patch.msgid.link/20260414-tf-update-txn-fix-v1-1-d2b83303acc9@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Enable modular build since the driver now has a proper module_exit()
handler. There's nothing specific to DZ hardware to prevent driver
unloading and reloading from working.
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Link: https://patch.msgid.link/alpine.DEB.2.21.2605062331420.46195@angie.orcam.me.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Prevent a crash from happening as the first serial port is initialised:
Console: switching to mono frame buffer device 160x64
fb0: PMAG-AA frame buffer device at tc0
DECstation Z85C30 serial driver version 0.10
CPU 0 Unable to handle kernel paging request at virtual address 0000002c, epc == 803ab00c, ra == 803aafe0
Oops[#1]:
CPU: 0 PID: 1 Comm: swapper Not tainted 6.4.0-rc3-00031-g84a9582fd203-dirty #57
$ 0 : 00000000 10012c00 803aaeb0 00000000
$ 4 : 80e12f60 80e12f50 80e12f58 81000030
$ 8 : 00000000 805ff37c 00000000 33433538
$12 : 65732030 00000006 80c2915d 6c616972
$16 : 80e12f00 807b7630 00000000 00000000
$20 : 00000004 00000348 000001a0 807623b8
$24 : 00000018 00000000
$28 : 80c24000 80c25d60 8078b148 803aafe0
Hi : 00000000
Lo : 00000000
epc : 803ab00c serial_base_ctrl_add+0x78/0xf4
ra : 803aafe0 serial_base_ctrl_add+0x4c/0xf4
Status: 10012c03 KERNEL EXL IE
Cause : 00000008 (ExcCode 02)
BadVA : 0000002c
PrId : 00000440 (R4400SC)
Modules linked in:
Process swapper (pid: 1, threadinfo=(ptrval), task=(ptrval), tls=00000000)
Stack : 80760000 00000cc0 00400044 00400040 803aa02c 80d61ab8 00000000 807b7630
80760000 807623b8 807b7628 803aa644 80386998 00000000 80e17780 80220f68
80e17780 80d61ab8 80c17d80 80e17780 80e17780 8063c798 80e17780 80383fa0
00000010 80e17780 00000000 80386998 807a0000 00000000 00400040 8038f848
807623b8 80d61ab8 00000004 80e17780 00000000 803a68e4 80c25e2c 803bb884
...
Call Trace:
[<803ab00c>] serial_base_ctrl_add+0x78/0xf4
[<803aa644>] serial_core_register_port+0x174/0x69c
[<8077e9ac>] zs_init+0xc8/0xfc
[<800404d4>] do_one_initcall+0x40/0x2ac
[<8076cecc>] kernel_init_freeable+0x1e4/0x270
[<80605bec>] kernel_init+0x20/0x108
[<800431e8>] ret_from_kernel_thread+0x14/0x1c
Code: 2442aeb0 ae120024 ae0200d0 <8c67002c> 50e00001 8c670000 3c06806e 3c05806e afb30010
---[ end trace 0000000000000000 ]---
(report at the offending commit) -- where a pointer is dereferenced that
has been derived from a null pointer to the port's parent device.
Since no device is available with legacy probing and it's not anymore a
preferable way to discover devices anyway, switch the driver to using a
platform device and use it as the port's parent device. Update resource
handling accordingly and only request the actual span of addresses used
within the slot, which will have had its resource already requested by
generic platform device code.
Use platform_driver_probe() not just because SCC devices are fixed with
solder on board and not straightforward to remove, but foremost because
the associated TTY's major device number is the same as used by the dz
driver and the first driver to claim it will prevent the other one from
using it. Either one DZ device or some SCC devices will be present in a
given system but never both at a time, and therefore we want the major
device number to be claimed by the first driver to actually successfully
bind to its device and platform_driver_probe() is a way to fulfil that.
An unfortunate consequence of the switch to a platform device is we now
hand the console over from the bootconsole much later in the bootstrap.
The firmware console handler appears good enough though to work so late
and in particular with interrupts enabled.
Since there is one way only remaining to reach zs_reset() now, remove
the port initialisation marker as no longer needed and go through the
channel reset unconditionally.
Fixes: 84a9582fd203 ("serial: core: Start managing serial controllers to enable runtime PM")
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Cc: stable@vger.kernel.org # needs to use .remove_new for <= 6.10
Link: https://patch.msgid.link/alpine.DEB.2.21.2605062328480.46195@angie.orcam.me.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Prevent a crash from happening as the first serial port is initialised:
Console: switching to colour frame buffer device 160x64
tgafb: SFB+ detected, rev=0x02
fb0: Digital ZLX-E1 frame buffer device at 0x1e000000
DECstation DZ serial driver version 1.04
CPU 0 Unable to handle kernel paging request at virtual address 000000bc, epc == 8048b3a4, ra == 80470a78
Oops[#1]:
CPU: 0 UID: 0 PID: 1 Comm: swapper/0 Not tainted 6.19.0-dirty #35 NONE
$ 0 : 00000000 1000ac00 00000004 804707ac
$ 4 : 00000000 80e20850 80e20858 81000030
$ 8 : 00000000 8072c81c 00000008 fefefeff
$12 : 6c616972 00000006 80c5917f 69726420
$16 : 80e20800 00000000 808f8968 80e20800
$20 : 00000000 807f5a90 808b0094 808d3bc8
$24 : 00000018 80479030
$28 : 80c2e000 80c2fd70 00000069 80470a78
Hi : 00000004
Lo : 00000000
epc : 8048b3a4 __dev_fwnode+0x0/0xc
ra : 80470a78 serial_base_ctrl_add+0xa0/0x168
Status: 1000ac04 IEp
Cause : 30000008 (ExcCode 02)
BadVA : 000000bc
PrId : 00000220 (R3000)
Modules linked in:
Process swapper/0 (pid: 1, threadinfo=(ptrval), task=(ptrval), tls=00000000)
Stack : 00400044 00400040 8046f4cc 00000000 808a6148 808a0000 808f8968 8086983c
808e0000 8046fc84 1000ac01 00000028 80e20700 802ba3f8 80e20700 80d34a94
80c1b900 80e20700 80e20700 80e20700 80e20700 80444650 00000000 00000000
00000000 807f5a90 808b0094 80447080 00400040 808e0000 80d34a94 808a6148
80d34a94 00000004 80e20700 00000000 8076974c 80469810 80c2fe3c 1000ac01
...
Call Trace:
[<8048b3a4>] __dev_fwnode+0x0/0xc
[<80470a78>] serial_base_ctrl_add+0xa0/0x168
[<8046fc84>] serial_core_register_port+0x1c8/0x974
[<808c6af0>] dz_init+0x74/0xc8
[<800470e0>] do_one_initcall+0x44/0x2d4
[<808b111c>] kernel_init_freeable+0x258/0x308
[<8072e434>] kernel_init+0x20/0x114
[<80049cd0>] ret_from_kernel_thread+0x14/0x1c
Code: 27bd0018 03e00008 2402ffea <8c8200bc> 03e00008 00000000 27bdffc0 afbe0038 afb30024
---[ end trace 0000000000000000 ]---
-- where a pointer is dereferenced that has been derived from a null
pointer to the port's parent device.
Since no device is available with legacy probing and it's not anymore a
preferable way to discover devices anyway, switch the driver to using a
platform device and use it as the port's parent device. Update resource
handling accordingly and only request the actual span of addresses used
within the slot, which will have had its resource already requested by
generic platform device code.
Use platform_driver_probe() not just because the DZ device is fixed with
solder on board and not straightforward to remove, but foremost because
the associated TTY's major device number is the same as used by the zs
driver and the first driver to claim it will prevent the other one from
using it. Either one DZ device or some SCC devices will be present in a
given system but never both at a time, and therefore we want the major
device number to be claimed by the first driver to actually successfully
bind to its device and platform_driver_probe() is a way to fulfil that.
An unfortunate consequence of the switch to a platform device is we now
hand the console over from the bootconsole much later in the bootstrap.
The firmware console handler appears good enough though to work so late
and in particular with interrupts enabled.
Conversely only starting the console port so late lets the reset code
fully utilise our delay handlers, so switch from udelay() to fsleep()
for transmitter draining so as to avoid busy-waiting for an excessive
amount of time.
Fixes: 84a9582fd203 ("serial: core: Start managing serial controllers to enable runtime PM")
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Cc: stable@vger.kernel.org # needs to use .remove_new for <= 6.10
Link: https://patch.msgid.link/alpine.DEB.2.21.2605062326540.46195@angie.orcam.me.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Switch the driver to using the channel reset rather than hardware reset,
simplifying handling by removing an interference between channels that
causes the other channel to become uninitialised afterwards.
There is little difference between the two kinds of reset in terms of
register settings that result, and we initialise the whole register set
right away anyway. However this prevents a hang from happening should
the console output handler in the firmware try to access the other port
whose transmitter has been disabled and line parameters messed up.
For example this will happen if the keyboard port (port A) is chosen for
the system console, unusually but not insanely for a headless system, as
the port is wired to a standard DA-15 connector and an adapter can be
easily made. Or with the next change in place this would happen for the
regular console port (port B), since the keyboard port (port A) will be
initialised first.
Just remove the unnecessary complication then, a channel reset is good
enough. We still need the initialisation marker, now per channel rather
than per SCC, as for the console port zs_reset() will be called twice:
once early on via zs_serial_console_init() for the console setup only,
and then again via zs_config_port() as the port is associated with a TTY
device.
Fixes: 8b4a40809e53 ("zs: move to the serial subsystem")
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Cc: stable@vger.kernel.org # v2.6.23+
Link: https://patch.msgid.link/alpine.DEB.2.21.2605062323430.46195@angie.orcam.me.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Calling zs_reset() in the course of setting up the serial device causes
line parameters to be reset and the transmitter disabled. We've been
lucky in that no message is usually produced to the kernel log between
this call and the later call to uart_set_options() in the course of
console setup done by zs_serial_console_init(), or the system would hang
as the console output handler in the firmware tried to access a port the
transmitter of which has been disabled and line parameters messed up.
This will change with the next change to the driver, so fix zs_reset()
such that line parameters are set for 9600n8 console operation as with
the system firmware and the transmitter re-enabled after reset. This
also means zs_pm() serves no purpose anymore, so drop it.
Fixes: 8b4a40809e53 ("zs: move to the serial subsystem")
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Cc: stable@vger.kernel.org # v2.6.23+
Link: https://patch.msgid.link/alpine.DEB.2.21.2605062308040.46195@angie.orcam.me.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Calling dz_reset() in the course of setting up the serial device causes
line parameters to be reset and the transmitter disabled. We've been
lucky in that no message is usually produced to the kernel log between
this call and the later call to uart_set_options() in the course of
console setup done by dz_serial_console_init(), or the system would hang
as the console output handler in the firmware tried to access a port the
transmitter of which has been disabled and line parameters messed up.
This will change with the next change to the driver, so fix dz_reset()
such that line parameters are set for 9600n8 console operation as with
the system firmware and the transmitter re-enabled after reset. This
also means dz_pm() serves no purpose anymore, so drop it.
Fixes: e6ee512f5a77 ("dz.c: Resource management")
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Cc: stable@vger.kernel.org # v2.6.25+
Link: https://patch.msgid.link/alpine.DEB.2.21.2605062302010.46195@angie.orcam.me.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
In the DZ interface as implemented by the DC7085 gate array the serial
transmitters are double buffered, meaning that at the time a transmitter
is ready to accept the next character there is one in the transmit shift
register still being sent to the line. Issuing a master clear at this
time causes this character to be lost, so wait an extra amount of time
sufficient for the transmit shift register to drain at 9600bps, which is
the baud rate setting used by the firmware console.
Mind the specified 1.4us TRDY recovery time in the course and continue
using iob() as the completion barrier, since the platforms involved use
a write buffer that can delay and combine writes, and reorder them with
respect to reads regardless of the MMIO locations accessed and we still
lack a platform-independent handler for that.
When called from dz_serial_console_init() this is too early for fsleep()
to work and even before lpj has been calculated and therefore the delay
is actually not sufficient for the transmitter to drain and is merely a
placeholder now. This will be addressed in a follow-up change.
Fixes: e6ee512f5a77 ("dz.c: Resource management")
Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Cc: stable@vger.kernel.org # v2.6.25+
Link: https://patch.msgid.link/alpine.DEB.2.21.2605062259080.46195@angie.orcam.me.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
dw8250_handle_irq() calls serial8250_handle_irq_locked() with the port
lock held via guard(uart_port_lock_irqsave). The guard destructor is
plain uart_port_unlock_irqrestore(), so a SysRq character captured into
port->sysrq_ch by uart_prepare_sysrq_char() is dropped without ever
being dispatched to handle_sysrq().
This is the same regression pattern as in serial8250_handle_irq(),
introduced when 883c5a2bc934 ("serial: 8250_dw: Rework
dw8250_handle_irq() locking and IIR handling") moved the function to
the guard()-based locking scheme without using the sysrq-aware unlock
helper.
Switch to guard(uart_port_lock_check_sysrq_irqsave) so that captured
sysrq_ch is dispatched on scope exit, matching the fix in
serial8250_handle_irq().
Fixes: 883c5a2bc934 ("serial: 8250_dw: Rework dw8250_handle_irq() locking and IIR handling")
Cc: stable@vger.kernel.org
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Jacques Nilo <jnilo@free.fr>
Link: https://patch.msgid.link/ed56fcaf4af24e4ed011a7bce206e0182acb761c.1778675349.git.jnilo@free.fr
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
serial8250_handle_irq() captures a SysRq character into port->sysrq_ch
inside serial8250_handle_irq_locked() via uart_prepare_sysrq_char()
(reached from serial8250_read_char()). Dispatch of that captured
character to handle_sysrq() is expected to happen at port-unlock time,
through uart_unlock_and_check_sysrq[_irqrestore]().
After commit 8324a54f604d ("serial: 8250: Add
serial8250_handle_irq_locked()") the function was reduced to a wrapper
that takes the port lock via guard(uart_port_lock_irqsave) whose
destructor is plain uart_port_unlock_irqrestore(). The sysrq-aware
unlock helper is no longer called, so port->sysrq_ch is captured but
never dispatched: BREAK + SysRq key is consumed silently.
This was the very condition Johan Hovold's 853a9ae29e978 ("serial:
8250: fix handle_irq locking", 2021) introduced
uart_unlock_and_check_sysrq_irqrestore() to address.
Switch to the new guard(uart_port_lock_check_sysrq_irqsave), whose
destructor is the sysrq-aware unlock helper, restoring the pre-split
behaviour. Update the Context: comment on serial8250_handle_irq_locked()
so future HW-specific 8250 wrappers know to use the same guard or the
explicit sysrq-aware unlock.
Verified on RTL8196E with CONFIG_MAGIC_SYSRQ_SERIAL=y: BREAK + 'h' on
the console UART produces the SysRq help dump in dmesg and the brk
counter in /proc/tty/driver/serial increments correctly.
Fixes: 8324a54f604d ("serial: 8250: Add serial8250_handle_irq_locked()")
Cc: stable@vger.kernel.org
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Jacques Nilo <jnilo@free.fr>
Link: https://patch.msgid.link/52692ae6c3501f7940347cef364ad7fcacaab7e5.1778675349.git.jnilo@free.fr
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
uart_handle_break() and uart_prepare_sysrq_char() (in
include/linux/serial_core.h) capture a SysRq character into
port->sysrq_ch while the port lock is held and rely on the unlock
helper -- uart_unlock_and_check_sysrq_irqrestore() -- to dispatch the
captured character to handle_sysrq() on scope exit.
The existing guard(uart_port_lock_irqsave) cannot be used by IRQ
handlers that process RX, because its destructor calls plain
uart_port_unlock_irqrestore() and silently drops port->sysrq_ch.
Add a dedicated guard(uart_port_lock_check_sysrq_irqsave) variant
whose destructor is the sysrq-aware unlock helper. The lock side is
identical to uart_port_lock_irqsave -- only the unlock-time behaviour
differs. Callers that may capture SysRq characters must use
guard(uart_port_lock_check_sysrq_irqsave); the existing
guard(uart_port_lock_irqsave) keeps its current plain-unlock semantics
for the many callers that do not process RX.
The new macro is placed after the CONFIG_MAGIC_SYSRQ_SERIAL block so
both definitions of uart_unlock_and_check_sysrq_irqrestore() (sysrq
enabled and disabled) are visible at expansion time. When
CONFIG_MAGIC_SYSRQ_SERIAL=n the destructor degenerates to plain
uart_port_unlock_irqrestore(), so there is no overhead.
No functional change on its own; users are converted in the following
patches.
Cc: stable@vger.kernel.org
Signed-off-by: Jacques Nilo <jnilo@free.fr>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Link: https://patch.msgid.link/3849af4bc55d5d2a424fa850844e94d641b2f8a6.1778675349.git.jnilo@free.fr
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Sashiko identified a deadlock when the console flow is engaged [1].
When console flow control is enabled (UPF_CONS_FLOW),
s3c24xx_serial_stop_tx() calls s3c24xx_serial_rx_enable() and
s3c24xx_serial_start_tx() calls s3c24xx_serial_rx_disable().
The serial core framework invokes the .stop_tx() and .start_tx()
callbacks with the port->lock spinlock already held. Furthermore, all
internal driver paths that invoke stop_tx (such as the DMA TX
completion handler s3c24xx_serial_tx_dma_complete() or the PIO TX IRQ
handler s3c24xx_serial_tx_irq()) also acquire port->lock prior to
calling it. (Note that s3c24xx_serial_start_tx() is only invoked by the
serial core).
However, s3c24xx_serial_rx_enable() and s3c24xx_serial_rx_disable()
unconditionally attempt to acquire port->lock again using
uart_port_lock_irqsave(). Since spinlocks are not recursive, this
causes a deadlock on the same CPU when console flow control is engaged.
Remove the redundant lock acquisition from both rx helper functions.
Cc: stable <stable@kernel.org>
Fixes: b497549a035c ("[ARM] S3C24XX: Split serial driver into core and per-cpu drivers")
Reported-by: John Ogness <john.ogness@linutronix.de>
Closes: https://sashiko.dev/#/patchset/20260506121606.5805-1-john.ogness%40linutronix.de [1]
Signed-off-by: Tudor Ambarus <tudor.ambarus@linaro.org>
Link: https://patch.msgid.link/20260515-samsung-tty-flow-control-deadlock-v1-1-93255edbc9bc@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
altera_jtaguart_probe() maps the register window before registering the
UART port, but it ignores failures from uart_add_one_port(). If port
registration fails, probe still returns success and the mapping remains
live until a later remove path that is not part of probe failure cleanup.
Return the uart_add_one_port() error and unmap the register window on
that failure path.
This issue was identified during our ongoing static-analysis research while
reviewing kernel code.
Fixes: 5bcd601049c6 ("serial: Add driver for the Altera JTAG UART")
Cc: stable <stable@kernel.org>
Co-developed-by: Ijae Kim <ae878000@gmail.com>
Signed-off-by: Ijae Kim <ae878000@gmail.com>
Signed-off-by: Myeonghun Pak <mhun512@gmail.com>
Link: https://patch.msgid.link/20260512065837.79528-1-mhun512@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
The hot path in vc_process_ucs() asks two independent questions about the
same code point -- "is it double-width?" and "is it zero-width?" -- and
was answering each with its own bsearch over its own table. For anything
past the leading bounds check that meant two scans of the BMP width
tables back to back for what is logically a single lookup.
Replace both with one ucs_get_width(cp) returning 0, 1, or 2 in a single
bsearch, while keeping the total table footprint at the same 2384 B as
before.
To do so, merge the zero-width and double-width ranges per region into
one sorted-by-`first` table. BMP entries stay 4 bytes; per-entry width
is hosted in spare bits of the non-BMP table's `last` field. Non-BMP
code points use only 20 of 32 bits, so each u32 has 12 unused high bits.
Store first/last shifted left by 12 and use the low 12 bits of `last`
for metadata: bit 11 is this entry's own width flag, bits 0..7 host an
8-bit chunk of the BMP double-width bitmap. Because the metadata bits
sit strictly below the lowest cp-scale bit, the bsearch comparator
remains a plain u32 compare on shifted keys with no masking.
In vc_process_ucs() the overwhelmingly common single-width path now
collapses to a single predicted branch:
if (likely(w == 1))
return 1;
Note: scripts/checkpatch.pl complains about "Macros with complex values
should be enclosed in parentheses" for the BMP_*WIDTH and
RANGE_*WIDTH macros. They are deliberately defined to expand to a
comma-separated (first, last) pair so they can populate the two
adjacent fields of a struct initializer; wrapping them in
parentheses would turn that into a comma-expression and defeat
the whole construction. Please ignore.
Signed-off-by: Nicolas Pitre <nico@fluxnic.net>
Link: https://patch.msgid.link/20260515034857.2514225-1-nico@fluxnic.net
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
There are rare cases in which the host gets stuck in the ISR because it
is flooded with messages during the startup phase.
The reason for the soft lockup in the ISR is the missing FIFO error IRQ
(FIFOE) handling. Not handling it and reporting IRQ_HANDLED triggers
the IRQ immediately again.
Fix this by adding a check for the FIFOE status and clearing the FIFO
if no data is ready (DR).
This behavior was observed on an AM62L device which uses the OMAP 8250
driver. Fix it for all 8250 drivers, since the OMAP driver's special
IRQ setup handling may trigger this behavior more frequently, but it
is not ensured that other 8250 drivers aren't affected.
Signed-off-by: Marco Felsch <m.felsch@pengutronix.de>
Link: https://patch.msgid.link/20260519-v7-1-topic-serial-8250-v1-1-56b04293a246@pengutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
The plain text binding file was superseded by the YAML schema in
commit d50f974c4f7f ("dt-bindings: serial: Convert rs485 bindings
to json-schema"). The file now contains only a redirect notice.
Remove it, and update references in serial_core.c and
serial-rs485.rst to point to the YAML schema.
Signed-off-by: Akash Sukhavasi <akash.sukhavasi@gmail.com>
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://patch.msgid.link/20260521162137.6325-1-akash.sukhavasi@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Add tracepoint support to the Qualcomm GENI serial driver to provide
runtime visibility into driver behavior without requiring invasive debug
patches.
The trace events cover UART termios configuration, clock setup, modem
control state, interrupt status, and TX/RX data, making it easier to
diagnose communication issues in the field.
Reviewed-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
Signed-off-by: Praveen Talari <praveen.talari@oss.qualcomm.com>
Link: https://patch.msgid.link/20260518-add-tracepoints-for-qcom-geni-serial-v3-1-b4addb151376@oss.qualcomm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
While being less compact, using named initializers allows to more easily
see which members of the structs are assigned which value without having
to lookup the declaration of the struct. And it's also more robust
against changes to the struct definition.
The mentioned robustness is relevant for a planned change to struct
i2c_device_id that replaces .driver_data by an anonymous union.
While touching all these arrays, unify usage of whitespace in the list
terminator.
This patch doesn't modify the compiled arrays, only their representation
in source form benefits. The former was confirmed with x86 and arm64
builds.
Signed-off-by: Uwe Kleine-König (The Capable Hub) <u.kleine-koenig@baylibre.com>
Link: https://patch.msgid.link/20260518101456.632410-2-u.kleine-koenig@baylibre.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
The clock notifier and matching work_struct in dw8250_data were added
in 2020 for the Baikal-T1 SoC, whose multiple UART ports share a
single reference clock and need to be informed when another consumer
re-rates that clock.
Baikal SoC support has since been removed from the kernel (see e.g.
commit 5d6c477687ae ("clk: baikal-t1: Remove not-going-to-be-supported
code for Baikal SoC") and the matching removals across bus/, mtd/,
PCI/, hwmon/, memory/). No remaining in-tree user needs the
cross-device baudclk rate-change notification path: the only
configuration that wired up the notifier was Baikal-T1's shared
reference clock topology.
Drop the now-unused clock-notifier and its deferred-update worker:
- struct dw8250_data fields clk_notifier and clk_work,
- the clk_to_dw8250_data() and work_to_dw8250_data() helpers,
- the dw8250_clk_work_cb() and dw8250_clk_notifier_cb() callbacks,
- the INIT_WORK / notifier_call setup in dw8250_probe(),
- the clk_notifier_register() / queue_work() in dw8250_probe(),
- the matching clk_notifier_unregister() / flush_work() in
dw8250_remove(),
- the stale comment in dw8250_set_termios() about the worker
blocking,
- the linux/notifier.h and linux/workqueue.h includes that are
no longer used.
dw8250_set_termios() keeps calling clk_set_rate() directly, which is
all the remaining single-UART configurations require.
Signed-off-by: Stepan Ionichev <sozdayvek@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://patch.msgid.link/20260514143746.23671-3-sozdayvek@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
dw8250_probe() registers the 8250 port via serial8250_register_8250_port()
and then, if the device has a clock, registers a clock notifier. If
clk_notifier_register() fails, probe returns the error but leaves the
8250 port registered. The matching serial8250_unregister_port() lives
in dw8250_remove(), which is not called when probe fails, so the port
slot stays occupied until the device is rebound or the system is
rebooted. The devm-allocated driver data is freed while the port still
references it (via the saved private_data and serial_in/serial_out
callbacks), so any access to that port slot before a rebind is a
use-after-free hazard.
Unregister the port on the clk_notifier_register() error path.
Fixes: cc816969d7b5 ("serial: 8250_dw: Fix common clocks usage race condition")
Cc: stable@vger.kernel.org
Signed-off-by: Stepan Ionichev <sozdayvek@gmail.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Link: https://patch.msgid.link/20260514143746.23671-2-sozdayvek@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
This is based on code removed in commit 89d4f98ae90d ("ARM: remove zte
zx platform"). I did not bring back the zx29-uart .compatible as the
arm,primecell-periphid does the job.
Reviewed-by: Linus Walleij <linusw@kernel.org>
Signed-off-by: Stefan Dösinger <stefandoesinger@gmail.com>
Link: https://patch.msgid.link/20260514-zx29uart-v1-2-68470ecc3977@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
The kernel documentation specifies that the console option 'r' can
be used to enable hardware flow control for console writes. The 8250
driver does include code for hardware flow control on the console if
cons_flow is set, but there is no code path that actually sets this.
However, that is not the only issue. The problems are:
1. Specifying the console option 'r' does not lead to cons_flow being
set.
2. Even if cons_flow would be set, serial8250_register_8250_port()
clears it.
3. When the console option 'r' is specified, uart_set_options()
attempts to initialize the port for CRTSCTS. However, afterwards
it does not set the UPSTAT_CTS_ENABLE status bit and therefore on
boot, uart_cts_enabled() is always false. This policy bit is
important for console drivers as a criteria if they may poll CTS.
4. Even though uart_set_options() attempts to initialize the port
for CRTSCTS, the 8250 set_termios() callback does not enable the
RTS signal (TIOCM_RTS) and thus the hardware is not properly
initialized for CTS polling.
5. Even if modem control was properly setup for CTS polling
(TIOCM_RTS), uart_configure_port() clears TIOCM_RTS, thus
breaking CTS polling.
6. wait_for_xmitr() and serial8250_console_write() use cons_flow
to decide if CTS polling should occur. However, the condition
should also include a check that it is not in RS485 mode and
CRTSCTS is actually enabled in the hardware.
Address all these issues as conservatively as possible by gating them
behind checks focussed on the user specifying console hardware flow
control support and the hardware being configured for CTS polling
at the time of the write to the UART.
Since checking the UPSTAT_CTS_ENABLE status bit is a part of the new
condition gate, these changes also support runtime termios updates to
disable/enable CRTSCTS.
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Link: https://patch.msgid.link/20260511152706.151498-4-john.ogness@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
wait_for_xmitr() calls wait_for_lsr() to wait for the transmission
registers to be empty. wait_for_lsr() can timeout after a reasonable
amount of time.
When console flow control is active, wait_for_xmitr() additionally
polls CTS, waiting for the peer to signal that it is ready to receive
more data.
If hardware flow control is enabled (auto CTS) and the peer deasserts
CTS, wait_for_lsr() will timeout. If additionally console flow
control is active and while polling CTS the peer asserts CTS, the
console will assume it can immediately transmit, even though the
transmission registers may not be empty. This can lead to data loss.
Avoid this problem by performing an extra wait_for_lsr() upon CTS
assertion if wait_for_lsr() previously timed out.
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Link: https://patch.msgid.link/20260511152706.151498-3-john.ogness@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Since console flow control policy is no longer part of uart_port.flags,
explicitly set the policy for the port.
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Link: https://patch.msgid.link/20260511152706.151498-2-john.ogness@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
serial8250_register_8250_port() conditionally copies uart->port.dev
from up->port.dev only when up->port.dev is non-NULL:
if (up->port.dev) {
uart->port.dev = up->port.dev;
...
}
So if both the existing uart slot and up have a NULL ->dev,
uart->port.dev remains NULL. The very next ACPI companion check
then dereferences it unconditionally:
if (!has_acpi_companion(uart->port.dev)) {
has_acpi_companion() reads dev->fwnode without a NULL guard
(include/linux/acpi.h), so this NULL-derefs the kernel for the
remaining no-dev case rather than just skipping the
mctrl_gpio_init() initialisation as intended.
smatch flags the inconsistency:
drivers/tty/serial/8250/8250_core.c:767
serial8250_register_8250_port() error: 'uart->port.dev' could be
null (see line 719)
Guard the call with a NULL check so register continues to work
for callers that legitimately have no parent device (legacy
non-OF/non-ACPI registrations).
No functional change for callers that pass a non-NULL ->dev.
Signed-off-by: Stepan Ionichev <sozdayvek@gmail.com>
Link: https://patch.msgid.link/20260508181237.11146-1-sozdayvek@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Add basic support for the A9 based Amlogic BY401 board, which describes
the following components: CPU, GIC, IRQ, Timer and UART.
These are capable of booting up into the serial console.
Reviewed-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: Xianwei Zhao <xianwei.zhao@amlogic.com>
Link: https://patch.msgid.link/20260423-a9-baisc-dts-v4-2-c26b480a068c@amlogic.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
