[<prev] [next>] [day] [month] [year] [list]
Message-ID: <9c3f30d166d7d4e08afd81d462413dff1703776a.1766164851.git.pabeni@redhat.com>
Date: Fri, 19 Dec 2025 18:21:03 +0100
From: Paolo Abeni <pabeni@...hat.com>
To: netdev@...r.kernel.org
Subject: [PATCH net] trace: fix UBSAN warning in __remove_instance
From: Darrick J. Wong <djwong@...nel.org>
xfs/558 triggers the following UBSAN warning:
------------[ cut here ]------------
UBSAN: shift-out-of-bounds in /storage/home/djwong/cdev/work/linux-xfs/kernel/trace/trace.c:10510:10
shift exponent 32 is too large for 32-bit type 'int'
CPU: 1 UID: 0 PID: 888674 Comm: rmdir Not tainted 6.19.0-rc1-xfsx #rc1 PREEMPT(lazy) dbf607ef4c142c563f76d706e71af9731d7b9c90
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-4.module+el8.8.0+21164+ed375313 04/01/2014
Call Trace:
<TASK>
dump_stack_lvl+0x4a/0x70
ubsan_epilogue+0x5/0x2b
__ubsan_handle_shift_out_of_bounds.cold+0x5e/0x113
__remove_instance.part.0.constprop.0.cold+0x18/0x26f
instance_rmdir+0xf3/0x110
tracefs_syscall_rmdir+0x4d/0x90
vfs_rmdir+0x139/0x230
do_rmdir+0x143/0x230
__x64_sys_rmdir+0x1d/0x20
do_syscall_64+0x44/0x230
entry_SYSCALL_64_after_hwframe+0x4b/0x53
RIP: 0033:0x7f7ae8e51f17
Code: f0 ff ff 73 01 c3 48 8b 0d de 2e 0e 00 f7 d8 64 89 01 48 83 c8 ff c3 66 2e 0f 1f 84 00 00 00 00 00 66 90 b8 54 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 01 c3 48 8b 15 b1 2e 0e 00 f7 d8 64 89 02 b8
RSP: 002b:00007ffd90743f08 EFLAGS: 00000246 ORIG_RAX: 0000000000000054
RAX: ffffffffffffffda RBX: 00007ffd907440f8 RCX: 00007f7ae8e51f17
RDX: 00007f7ae8f3c5c0 RSI: 00007ffd90744a21 RDI: 00007ffd90744a21
RBP: 0000000000000002 R08: 0000000000000000 R09: 0000000000000000
R10: 00007f7ae8f35ac0 R11: 0000000000000246 R12: 00007ffd90744a21
R13: 0000000000000001 R14: 00007f7ae8f8b000 R15: 000055e5283e6a98
</TASK>
---[ end trace ]---
whilst tearing down an ftrace instance. TRACE_FLAGS_MAX_SIZE is now 64,
so the mask comparison expression must be typecast to a u64 value to
avoid an overflow. AFAICT, ZEROED_TRACE_FLAGS is already cast to ULL
so this is ok.
Fixes: bbec8e28cac592 ("tracing: Allow tracer to add more than 32 options")
Signed-off-by: "Darrick J. Wong" <djwong@...nel.org>
Signed-off-by: Paolo Abeni <pabeni@...hat.com>
---
Note: DO NOT MERGE on net nor net-next. Reshared on the netdev ML to fix
CI failures caused by the blamed commit above. Will rot on PW until the
net PR or someone can access the CI hosts, whatever come first.
---
kernel/trace/trace.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index e575956ef9b5..6f2148df14d9 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -10507,7 +10507,7 @@ static int __remove_instance(struct trace_array *tr)
/* Disable all the flags that were enabled coming in */
for (i = 0; i < TRACE_FLAGS_MAX_SIZE; i++) {
- if ((1 << i) & ZEROED_TRACE_FLAGS)
+ if ((1ULL << i) & ZEROED_TRACE_FLAGS)
set_tracer_flag(tr, 1ULL << i, 0);
}
--
2.52.0
Powered by blists - more mailing lists