scsi/mpt3sas: fix a KASAN report
Subsystems (2)
PATCH 1/1 scsi/mpt3sas: fix a KASAN report
From: Peng Hao <[email protected]>
==================================================================
[ 103.297144] BUG: KASAN: slab-out-of-bounds in _ctl_do_mpt_command+0xdd9/0x2f30 [mpt3sas] [ 103.306623] Read of size 8 at addr ffff888116747d48 by task sas3ircu/3658
[ 103.314781] CPU: 69 PID: 3658 Comm: sas3ircu Not tainted 6.10.0-rc2+ #18 [ 103.314787] Hardware name: Inspur SA5212M5/SA5212M5, BIOS 4.3.17 07/09/2021 [ 103.314790] Call Trace: [ 103.336743] <TASK> [ 103.336745] dump_stack_lvl+0x8d/0x140 [ 103.342826] ? __pfx_dump_stack_lvl+0x10/0x10 [ 103.342829] ? _printk+0xb0/0xf0 [ 103.342833] ? __kasan_check_write+0x18/0x20 [ 103.342841] print_report+0x199/0x570 [ 103.342844] ? _ctl_do_mpt_command+0xdd9/0x2f30 [mpt3sas] [ 103.342875] ? kasan_complete_mode_report_info+0x45/0x240 [ 103.342879] ? _ctl_do_mpt_command+0xdd9/0x2f30 [mpt3sas] [ 103.342910] kasan_report+0xd8/0x110 [ 103.342913] ? _ctl_do_mpt_command+0xdd9/0x2f30 [mpt3sas] [ 103.342944] kasan_check_range+0x262/0x2f0 [ 103.342981] _ctl_do_mpt_command+0xdd9/0x2f30 [mpt3sas] [ 103.343068] _ctl_ioctl_main+0x732/0x930 [mpt3sas] [ 103.343136] _ctl_ioctl+0x1a/0x20 [mpt3sas] [ 103.455340] __se_sys_ioctl+0xd0/0x120 [ 103.455345] __x64_sys_ioctl+0x7f/0x90 [ 103.455349] x64_sys_call+0x1586/0x2d00 [ 103.467041] do_syscall_64+0x60/0x110 [ 103.467044] ? clear_bhb_loop+0x45/0xa0 [ 103.474781] entry_SYSCALL_64_after_hwframe+0x76/0x7e [ 103.479952] RIP: 0033:0x442829 [ 103.483127] Code: ff 85 c0 48 8b 6c 24 10 48 0f 49 d3 48 8b 5c 24 08 48 83 c4 18 48 89 d0 c3 90 90 90 90 90 90 90 90 48 c7 c0 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 0f 83 4b 31 00 00 c3 90 90 90 90 90 90 90 90 90 [ 103.502096] RSP: 002b:00007fff2bbd7e68 EFLAGS: 00000246 ORIG_RAX: 0000000000000010 [ 103.502101] RAX: ffffffffffffffda RBX: 000000001ae96c80 RCX: 0000000000442829 [ 103.502104] RDX: 000000001ae971e0 RSI: 00000000c0484c14 RDI: 0000000000000006 [ 103.502107] RBP: 00007fff2bbd7eb0 R08: 00000000c0484c14 R09: 0000001202000000 [ 103.531597] R10: 000000001ae97224 R11: 0000000000000246 R12: 00007fff2bbd8d30 [ 103.531600] R13: 0000000000000001 R14: 0000000000000000 R15: 0000000000000000 [ 103.531604] </TASK>
[ 103.531608] Allocated by task 9: [ 103.553368] kasan_save_track+0x30/0x70 [ 103.553373] kasan_save_alloc_info+0x3c/0x50 [ 103.553378] __kasan_kmalloc+0xa2/0xc0 [ 103.553381] __kmalloc_noprof+0x1bd/0x3f0 [ 103.553387] mpt3sas_base_attach+0x1176/0x20f0 [mpt3sas] [ 103.553418] _scsih_probe+0x131a/0x23f0 [mpt3sas] [ 103.553449] local_pci_probe+0xf8/0x160 [ 103.553454] work_for_cpu_fn+0x5b/0x90 [ 103.553458] process_scheduled_works+0x72d/0xdc0 [ 103.553462] worker_thread+0x85b/0xc40 [ 103.553466] kthread+0x2ad/0x340 [ 103.553471] ret_from_fork+0x41/0x70 [ 103.553475] ret_from_fork_asm+0x1a/0x30
[ 103.555079] The buggy address belongs to the object at ffff888116747d00
which belongs to the cache kmalloc-96 of size 96
[ 103.567583] The buggy address is located 72 bytes inside of
allocated 76-byte region [ffff888116747d00, ffff888116747d4c)[ 103.589088] Memory state around the buggy address: [ 103.593999] ffff888116747c00: 00 00 00 00 00 00 00 00 00 04 fc fc fc fc fc fc [ 103.601375] ffff888116747c80: 00 00 00 00 00 00 00 00 00 04 fc fc fc fc fc fc [ 103.608750] >ffff888116747d00: 00 00 00 00 00 00 00 00 00 04 fc fc fc fc fc fc [ 103.616125] ^ [ 103.621822] ffff888116747d80: 00 00 00 00 00 00 00 00 00 00 00 00 fc fc fc fc [ 103.629192] ffff888116747e00: 00 00 00 00 00 00 00 00 00 fc fc fc fc fc fc fc [ 103.636560] ==================================================================
Although it appears to be a KASAN report, it is actually a concurrency issue.
Signed-off-by Peng Hao <[email protected]>
--- drivers/scsi/mpt3sas/mpt3sas_base.c | 3 ++- drivers/scsi/mpt3sas/mpt3sas_ctl.c | 2 +- 2 files changed, 3 insertions(+), 2 deletions(-)
drivers/scsi/mpt3sas/mpt3sas_base.c +2 -1
@@ -8577,7 +8577,8 @@ mpt3sas_base_attach(struct MPT3SAS_ADAPTER *ioc)/* ctl module internal command bits */ ioc->ctl_cmds.reply =kzalloc (ioc->reply_sz,GFP_KERNEL ); - ioc->ctl_cmds.sense =kzalloc (SCSI_SENSE_BUFFERSIZE ,GFP_KERNEL ); +void *sense =kzalloc (SCSI_SENSE_BUFFERSIZE ,GFP_KERNEL ); +WRITE_ONCE (ioc->ctl_cmds.sense, sense); ioc->ctl_cmds.status =MPT3_CMD_NOT_USED ;mutex_init (&ioc->ctl_cmds.mutex);
drivers/scsi/mpt3sas/mpt3sas_ctl.c +1 -1
@@ -1096,7 +1096,7 @@ _ctl_do_mpt_command(struct MPT3SAS_ADAPTER *ioc, struct mpt3_ioctl_command karg, MPI2_FUNCTION_NVME_ENCAPSULATED) ? NVME_ERROR_RESPONSE_SIZE :SCSI_SENSE_BUFFERSIZE ; sz =min_t (u32, karg.max_sense_bytes, sz_arg); -if (copy_to_user (karg.sense_data_ptr, ioc->ctl_cmds.sense, +if (copy_to_user (karg.sense_data_ptr,READ_ONCE (ioc->ctl_cmds.sense), sz)) {pr_err ("failure at %s:%d/%s()!\n" , __FILE__, __LINE__, __func__);
Signature · 1 line
2.27.0
Quoted 1 line · On 7/9/24 3:55 AM, [email protected] wrote:
Although it appears to be a KASAN report, it is actually a concurrency issue.
Please explain the concurrency issue. Please also explain why the introduction of WRITE_ONCE() / READ_ONCE() solves the concurrency issue rather than papering over it. Note: I expect the latter rather than the former.
Thanks,
Bart.