Comment 1 for bug 2039575

------- Comment From <email address hidden> 2023-10-17 10:23 EDT-------
--- MATTHEW R. commented ---

I encountered this issue recently while verifying a KVM feature. It's worth noting that if a kernel has UBSAN enabled (https://docs.kernel.org/dev-tools/ubsan.html) then rather than silently updating an incorrect stats counter you will also get a UBSAN array-index-out-of-bounds warning every time this occurs. In my case, I bumped into this because I was using an Ubuntu kernel which came with UBSAN enabled. Example of the warning:

[ 26.335369] ================================================================================
[ 26.335381] UBSAN: array-index-out-of-bounds in /build/linux-O6Qi7m/linux-5.15.0/net/smc/af_smc.c:2402:3
[ 26.335385] index -1 is out of range for type 'u64 [9]'
[ 26.335388] CPU: 0 PID: 274 Comm: iperf3 Tainted: G E 5.15.0-79-generic #86-Ubuntu
[ 26.335391] Hardware name: IBM 8561 T01 772 (KVM/Linux)
[ 26.335393] Call Trace:
[ 26.335397] [<00000000cd92e63a>] dump_stack_lvl+0x62/0x80
[ 26.335404] [<00000000cd92e36c>] ubsan_epilogue+0x1c/0x48
[ 26.335406] [<00000000cd52d3c4>] __ubsan_handle_out_of_bounds+0x94/0xa0
[ 26.335411] [<000003ff8033f9da>] smc_sendmsg+0x2aa/0x2d0 [smc]
[ 26.335425] [<00000000cd6a79a4>] sock_sendmsg+0x64/0x80
[ 26.335431] [<00000000cd6a7a32>] sock_write_iter+0x72/0xa0
[ 26.335433] [<00000000cd1d4000>] new_sync_write+0x100/0x190
[ 26.335438] [<00000000cd1d4bb8>] vfs_write+0x1e8/0x280
[ 26.335440] [<00000000cd1d7014>] ksys_write+0xb4/0x100
[ 26.335442] [<00000000cd932c7c>] __do_syscall+0x1bc/0x1f0
[ 26.335446] [<00000000cd940148>] system_call+0x78/0xa0

This makes the issue much more visible.

Worse, if you have panic_on_warn enabled (like I did) then this warning will subsequently trigger a kernel panic.