Re: [PATCH V4] scsi_debugfs: fix crash in scsi_show_rq()
On Wed, Nov 15, 2017 at 08:04:49PM +0800, Ming Lei wrote: > On Wed, Nov 15, 2017 at 07:28:00PM +0900, James Bottomley wrote: > > On Wed, 2017-11-15 at 18:09 +0800, Ming Lei wrote: > > > On Tue, Nov 14, 2017 at 10:14:52AM -0800, James Bottomley wrote: > > > > > > > > On Tue, 2017-11-14 at 08:55 +0800, Ming Lei wrote: > > > > > > > > > > Hi James, > > > > > > > > > > On Mon, Nov 13, 2017 at 10:55:52AM -0800, James Bottomley wrote: > > > > > > > > > > > > > > > > > > On Sat, 2017-11-11 at 10:43 +0800, Ming Lei wrote: > > > > > > > > > > > > > > > > > > > > > So from CPU1's review, cmd->cmnd is in a remote NUMA node, > > > > > > > __scsi_format_command() is executed much slower than > > > > > > > mempool_free(). > > > > > > > So when mempool_free() returns, __scsi_format_command() may > > > > > > > not fetched the buffer in L1 cache yet, then use-after-free > > > > > > > is still triggered. > > > > > > > > > > > > > > That is why I say this use-after-free is inevitable no matter > > > > > > > 'setting SCpnt->cmnd to NULL before calling mempool_free()' > > > > > > > or not. > > > > > > > > > > > > The bottom line is that there are several creative ways around > > > > > > this but the proposed code is currently broken and simply > > > > > > putting a comment in saying so doesn't make it acceptable. > > > > > > > > > > As I explained above, I didn't see one really workable way. Or > > > > > please correct it if I am wrong. > > > > > > > > I simply can't believe it's beyond the wit of man to solve a use > > > > after free race. About 40% of kernel techniques are devoted to > > > > this. All I really care about is not losing the PI information we > > > > previously had. I agree with Bart that NULL cmnd is a good > > > > indicator, so it seems reasonable to use it. If you have another > > > > mechanism, feel free to propose it. > > > > > > Hi James, > > > > > > This patch is my proposal, no others thought of yet. > > > > > > We can fix the use-after-free easily via lock, rcu and ..., but some > > > cost has to pay. In this case, we can't wait too long in show_rq(), > > > otherwise we may lose important debug info, so I do not have better > > > way. > > > > > > IMO this use-after-free is actually no harm, I don't think we have to > > > fix it, but it should be better to not let utility warn on this case. > > > > Fine, so lose the snide comment and set the cmnd to NULL *before* > > calling free so we narrow the race window. > > Hi James, > > Given we can't avoid the use-after-free, how about not do that way so > we won't lose the precious debug info too early? Hi James, Are you fine with V4? As I explained, the use-after-free can't be avoided, we have to make scsi_show_rq() to survive that, so we don't need to touch code in free path. Also we won't lose debug info too early in this way, not like 'set the cmnd to NULL *before* calling free'. Thanks, Ming
Re: [PATCH V4] scsi_debugfs: fix crash in scsi_show_rq()
On Wed, 2017-11-15 at 18:09 +0800, Ming Lei wrote: > On Tue, Nov 14, 2017 at 10:14:52AM -0800, James Bottomley wrote: > > > > On Tue, 2017-11-14 at 08:55 +0800, Ming Lei wrote: > > > > > > Hi James, > > > > > > On Mon, Nov 13, 2017 at 10:55:52AM -0800, James Bottomley wrote: > > > > > > > > > > > > On Sat, 2017-11-11 at 10:43 +0800, Ming Lei wrote: > > > > > > > > > > > > > > > So from CPU1's review, cmd->cmnd is in a remote NUMA node, > > > > > __scsi_format_command() is executed much slower than > > > > > mempool_free(). > > > > > So when mempool_free() returns, __scsi_format_command() may > > > > > not fetched the buffer in L1 cache yet, then use-after-free > > > > > is still triggered. > > > > > > > > > > That is why I say this use-after-free is inevitable no matter > > > > > 'setting SCpnt->cmnd to NULL before calling mempool_free()' > > > > > or not. > > > > > > > > The bottom line is that there are several creative ways around > > > > this but the proposed code is currently broken and simply > > > > putting a comment in saying so doesn't make it acceptable. > > > > > > As I explained above, I didn't see one really workable way. Or > > > please correct it if I am wrong. > > > > I simply can't believe it's beyond the wit of man to solve a use > > after free race. About 40% of kernel techniques are devoted to > > this. All I really care about is not losing the PI information we > > previously had. I agree with Bart that NULL cmnd is a good > > indicator, so it seems reasonable to use it. If you have another > > mechanism, feel free to propose it. > > Hi James, > > This patch is my proposal, no others thought of yet. > > We can fix the use-after-free easily via lock, rcu and ..., but some > cost has to pay. In this case, we can't wait too long in show_rq(), > otherwise we may lose important debug info, so I do not have better > way. > > IMO this use-after-free is actually no harm, I don't think we have to > fix it, but it should be better to not let utility warn on this case. Fine, so lose the snide comment and set the cmnd to NULL *before* calling free so we narrow the race window. James
Re: [PATCH V4] scsi_debugfs: fix crash in scsi_show_rq()
On Tue, Nov 14, 2017 at 10:14:52AM -0800, James Bottomley wrote: > On Tue, 2017-11-14 at 08:55 +0800, Ming Lei wrote: > > Hi James, > > > > On Mon, Nov 13, 2017 at 10:55:52AM -0800, James Bottomley wrote: > > > > > > On Sat, 2017-11-11 at 10:43 +0800, Ming Lei wrote: > > > > > > > > So from CPU1's review, cmd->cmnd is in a remote NUMA node, > > > > __scsi_format_command() is executed much slower than > > > > mempool_free(). > > > > So when mempool_free() returns, __scsi_format_command() may not > > > > fetched the buffer in L1 cache yet, then use-after-free > > > > is still triggered. > > > > > > > > That is why I say this use-after-free is inevitable no matter > > > > 'setting SCpnt->cmnd to NULL before calling mempool_free()' or > > > > not. > > > > > > The bottom line is that there are several creative ways around this > > > but the proposed code is currently broken and simply putting a > > > comment in saying so doesn't make it acceptable. > > > > As I explained above, I didn't see one really workable way. Or please > > correct it if I am wrong. > > I simply can't believe it's beyond the wit of man to solve a use after > free race. About 40% of kernel techniques are devoted to this. All I > really care about is not losing the PI information we previously had. > I agree with Bart that NULL cmnd is a good indicator, so it seems > reasonable to use it. If you have another mechanism, feel free to > propose it. Hi James, This patch is my proposal, no others thought of yet. We can fix the use-after-free easily via lock, rcu and ..., but some cost has to pay. In this case, we can't wait too long in show_rq(), otherwise we may lose important debug info, so I do not have better way. IMO this use-after-free is actually no harm, I don't think we have to fix it, but it should be better to not let utility warn on this case. Thanks, Ming
Re: [PATCH V4] scsi_debugfs: fix crash in scsi_show_rq()
On Tue, 2017-11-14 at 08:55 +0800, Ming Lei wrote: > Hi James, > > On Mon, Nov 13, 2017 at 10:55:52AM -0800, James Bottomley wrote: > > > > On Sat, 2017-11-11 at 10:43 +0800, Ming Lei wrote: > > > > > > So from CPU1's review, cmd->cmnd is in a remote NUMA node, > > > __scsi_format_command() is executed much slower than > > > mempool_free(). > > > So when mempool_free() returns, __scsi_format_command() may not > > > fetched the buffer in L1 cache yet, then use-after-free > > > is still triggered. > > > > > > That is why I say this use-after-free is inevitable no matter > > > 'setting SCpnt->cmnd to NULL before calling mempool_free()' or > > > not. > > > > The bottom line is that there are several creative ways around this > > but the proposed code is currently broken and simply putting a > > comment in saying so doesn't make it acceptable. > > As I explained above, I didn't see one really workable way. Or please > correct it if I am wrong. I simply can't believe it's beyond the wit of man to solve a use after free race. About 40% of kernel techniques are devoted to this. All I really care about is not losing the PI information we previously had. I agree with Bart that NULL cmnd is a good indicator, so it seems reasonable to use it. If you have another mechanism, feel free to propose it. James
Re: [PATCH V4] scsi_debugfs: fix crash in scsi_show_rq()
Hi James, On Mon, Nov 13, 2017 at 10:55:52AM -0800, James Bottomley wrote: > On Sat, 2017-11-11 at 10:43 +0800, Ming Lei wrote: > > On Fri, Nov 10, 2017 at 08:51:58AM -0800, James Bottomley wrote: > > > > > > On Fri, 2017-11-10 at 17:01 +0800, Ming Lei wrote: > > > > > > > > cmd->cmnd can be allocated/freed dynamically in case of > > > > T10_PI_TYPE2_PROTECTION, so we should check it in scsi_show_rq() > > > > because this request may have been freed already here, and cmd- > > > > >cmnd has been set as null. > > > > > > > > We choose to accept read-after-free and dump request data as far > > > > as possible. > > > > > > > > This patch fixs the following kernel crash when dumping request > > > > via block's debugfs interface: > > > > > > > > [ 252.962045] BUG: unable to handle kernel NULL pointer > > > > dereference > > > > at (null) > > > > [ 252.963007] IP: scsi_format_opcode_name+0x1a/0x1c0 > > > > [ 252.963007] PGD 25e75a067 P4D 25e75a067 PUD 25e75b067 PMD 0 > > > > [ 252.963007] Oops: [#1] PREEMPT SMP > > > > [ 252.963007] Dumping ftrace buffer: > > > > [ 252.963007](ftrace buffer empty) > > > > [ 252.963007] Modules linked in: scsi_debug ebtable_filter > > > > ebtables > > > > ip6table_filter ip6_tables xt_CHECKSUM iptable_mangle > > > > ipt_MASQUERADE > > > > nf_nat_masquerade_ipv4 iptable_nat nf_conntrack_ipv4 > > > > nf_defrag_ipv4 > > > > nf_nat_ipv4 nf_nat nf_conntrack libcrc32c bridge stp llc > > > > iptable_filter fuse ip_tables sd_mod sg mptsas mptscsih mptbase > > > > crc32c_intel ahci libahci nvme serio_raw scsi_transport_sas > > > > libata > > > > lpc_ich nvme_core virtio_scsi binfmt_misc dm_mod iscsi_tcp > > > > libiscsi_tcp libiscsi scsi_transport_iscsi null_blk configs > > > > [ 252.963007] CPU: 1 PID: 1881 Comm: cat Not tainted 4.14.0- > > > > rc2.blk_mq_io_hang+ #516 > > > > [ 252.963007] Hardware name: QEMU Standard PC (Q35 + ICH9, > > > > 2009), > > > > BIOS 1.9.3-1.fc25 04/01/2014 > > > > [ 252.963007] task: 88025e6f6000 task.stack: > > > > c90001bd > > > > [ 252.963007] RIP: 0010:scsi_format_opcode_name+0x1a/0x1c0 > > > > [ 252.963007] RSP: 0018:c90001bd3c50 EFLAGS: 00010286 > > > > [ 252.963007] RAX: 4843 RBX: 0050 RCX: > > > > > > > > [ 252.963007] RDX: RSI: 0050 RDI: > > > > c90001bd3cd8 > > > > [ 252.963007] RBP: c90001bd3c88 R08: 1000 R09: > > > > > > > > [ 252.963007] R10: 880275134000 R11: 88027513406c R12: > > > > 0050 > > > > [ 252.963007] R13: c90001bd3cd8 R14: R15: > > > > > > > > [ 252.963007] FS: 7f4d11762700() > > > > GS:88027fc4() > > > > knlGS: > > > > [ 252.963007] CS: 0010 DS: ES: CR0: 80050033 > > > > [ 252.963007] CR2: CR3: 00025e789003 CR4: > > > > 003606e0 > > > > [ 252.963007] DR0: DR1: DR2: > > > > > > > > [ 252.963007] DR3: DR6: fffe0ff0 DR7: > > > > 0400 > > > > [ 252.963007] Call Trace: > > > > [ 252.963007] __scsi_format_command+0x27/0xc0 > > > > [ 252.963007] scsi_show_rq+0x5c/0xc0 > > > > [ 252.963007] ? seq_printf+0x4e/0x70 > > > > [ 252.963007] ? blk_flags_show+0x5b/0xf0 > > > > [ 252.963007] __blk_mq_debugfs_rq_show+0x116/0x130 > > > > [ 252.963007] blk_mq_debugfs_rq_show+0xe/0x10 > > > > [ 252.963007] seq_read+0xfe/0x3b0 > > > > [ 252.963007] ? __handle_mm_fault+0x631/0x1150 > > > > [ 252.963007] full_proxy_read+0x54/0x90 > > > > [ 252.963007] __vfs_read+0x37/0x160 > > > > [ 252.963007] ? security_file_permission+0x9b/0xc0 > > > > [ 252.963007] vfs_read+0x96/0x130 > > > > [ 252.963007] SyS_read+0x55/0xc0 > > > > [ 252.963007] entry_SYSCALL_64_fastpath+0x1a/0xa5 > > > > [ 252.963007] RIP: 0033:0x7f4d1127e9b0 > > > > [ 252.963007] RSP: 002b:7ffd27082568 EFLAGS: 0246 > > > > ORIG_RAX: > > > > > > > > [ 252.963007] RAX: ffda RBX: 7f4d1154bb20 RCX: > > > > 7f4d1127e9b0 > > > > [ 252.963007] RDX: 0002 RSI: 7f4d115a7000 RDI: > > > > 0003 > > > > [ 252.963007] RBP: 00021010 R08: R09: > > > > > > > > [ 252.963007] R10: 037b R11: 0246 R12: > > > > 00022000 > > > > [ 252.963007] R13: 7f4d1154bb78 R14: 1000 R15: > > > > 0002 > > > > [ 252.963007] Code: c6 e8 1b ca 24 00 eb 8c e8 74 2c ae ff 0f 1f > > > > 40 > > > > 00 0f 1f 44 00 00 55 48 89 e5 41 56 41 55 41 54 53 49 89 fd 49 89 > > > > f4 > > > > 48 83 ec 18 <44> 0f b6 32 48 c7 45 c8 00 00 00 00 65 48 8b 04 25 > > > > 28 > > > > 00 00 00 > > > > [ 252.963007] RIP: scsi_format_opcode_name+0x1a/0x1c0 RSP: > > > > c90001bd3c50 > > > > [ 252.963007] CR2: > > > > [ 252.963007] ---[ end
Re: [PATCH V4] scsi_debugfs: fix crash in scsi_show_rq()
On Sat, 2017-11-11 at 10:43 +0800, Ming Lei wrote: > On Fri, Nov 10, 2017 at 08:51:58AM -0800, James Bottomley wrote: > > > > On Fri, 2017-11-10 at 17:01 +0800, Ming Lei wrote: > > > > > > cmd->cmnd can be allocated/freed dynamically in case of > > > T10_PI_TYPE2_PROTECTION, so we should check it in scsi_show_rq() > > > because this request may have been freed already here, and cmd- > > > >cmnd has been set as null. > > > > > > We choose to accept read-after-free and dump request data as far > > > as possible. > > > > > > This patch fixs the following kernel crash when dumping request > > > via block's debugfs interface: > > > > > > [ 252.962045] BUG: unable to handle kernel NULL pointer > > > dereference > > > at (null) > > > [ 252.963007] IP: scsi_format_opcode_name+0x1a/0x1c0 > > > [ 252.963007] PGD 25e75a067 P4D 25e75a067 PUD 25e75b067 PMD 0 > > > [ 252.963007] Oops: [#1] PREEMPT SMP > > > [ 252.963007] Dumping ftrace buffer: > > > [ 252.963007](ftrace buffer empty) > > > [ 252.963007] Modules linked in: scsi_debug ebtable_filter > > > ebtables > > > ip6table_filter ip6_tables xt_CHECKSUM iptable_mangle > > > ipt_MASQUERADE > > > nf_nat_masquerade_ipv4 iptable_nat nf_conntrack_ipv4 > > > nf_defrag_ipv4 > > > nf_nat_ipv4 nf_nat nf_conntrack libcrc32c bridge stp llc > > > iptable_filter fuse ip_tables sd_mod sg mptsas mptscsih mptbase > > > crc32c_intel ahci libahci nvme serio_raw scsi_transport_sas > > > libata > > > lpc_ich nvme_core virtio_scsi binfmt_misc dm_mod iscsi_tcp > > > libiscsi_tcp libiscsi scsi_transport_iscsi null_blk configs > > > [ 252.963007] CPU: 1 PID: 1881 Comm: cat Not tainted 4.14.0- > > > rc2.blk_mq_io_hang+ #516 > > > [ 252.963007] Hardware name: QEMU Standard PC (Q35 + ICH9, > > > 2009), > > > BIOS 1.9.3-1.fc25 04/01/2014 > > > [ 252.963007] task: 88025e6f6000 task.stack: > > > c90001bd > > > [ 252.963007] RIP: 0010:scsi_format_opcode_name+0x1a/0x1c0 > > > [ 252.963007] RSP: 0018:c90001bd3c50 EFLAGS: 00010286 > > > [ 252.963007] RAX: 4843 RBX: 0050 RCX: > > > > > > [ 252.963007] RDX: RSI: 0050 RDI: > > > c90001bd3cd8 > > > [ 252.963007] RBP: c90001bd3c88 R08: 1000 R09: > > > > > > [ 252.963007] R10: 880275134000 R11: 88027513406c R12: > > > 0050 > > > [ 252.963007] R13: c90001bd3cd8 R14: R15: > > > > > > [ 252.963007] FS: 7f4d11762700() > > > GS:88027fc4() > > > knlGS: > > > [ 252.963007] CS: 0010 DS: ES: CR0: 80050033 > > > [ 252.963007] CR2: CR3: 00025e789003 CR4: > > > 003606e0 > > > [ 252.963007] DR0: DR1: DR2: > > > > > > [ 252.963007] DR3: DR6: fffe0ff0 DR7: > > > 0400 > > > [ 252.963007] Call Trace: > > > [ 252.963007] __scsi_format_command+0x27/0xc0 > > > [ 252.963007] scsi_show_rq+0x5c/0xc0 > > > [ 252.963007] ? seq_printf+0x4e/0x70 > > > [ 252.963007] ? blk_flags_show+0x5b/0xf0 > > > [ 252.963007] __blk_mq_debugfs_rq_show+0x116/0x130 > > > [ 252.963007] blk_mq_debugfs_rq_show+0xe/0x10 > > > [ 252.963007] seq_read+0xfe/0x3b0 > > > [ 252.963007] ? __handle_mm_fault+0x631/0x1150 > > > [ 252.963007] full_proxy_read+0x54/0x90 > > > [ 252.963007] __vfs_read+0x37/0x160 > > > [ 252.963007] ? security_file_permission+0x9b/0xc0 > > > [ 252.963007] vfs_read+0x96/0x130 > > > [ 252.963007] SyS_read+0x55/0xc0 > > > [ 252.963007] entry_SYSCALL_64_fastpath+0x1a/0xa5 > > > [ 252.963007] RIP: 0033:0x7f4d1127e9b0 > > > [ 252.963007] RSP: 002b:7ffd27082568 EFLAGS: 0246 > > > ORIG_RAX: > > > > > > [ 252.963007] RAX: ffda RBX: 7f4d1154bb20 RCX: > > > 7f4d1127e9b0 > > > [ 252.963007] RDX: 0002 RSI: 7f4d115a7000 RDI: > > > 0003 > > > [ 252.963007] RBP: 00021010 R08: R09: > > > > > > [ 252.963007] R10: 037b R11: 0246 R12: > > > 00022000 > > > [ 252.963007] R13: 7f4d1154bb78 R14: 1000 R15: > > > 0002 > > > [ 252.963007] Code: c6 e8 1b ca 24 00 eb 8c e8 74 2c ae ff 0f 1f > > > 40 > > > 00 0f 1f 44 00 00 55 48 89 e5 41 56 41 55 41 54 53 49 89 fd 49 89 > > > f4 > > > 48 83 ec 18 <44> 0f b6 32 48 c7 45 c8 00 00 00 00 65 48 8b 04 25 > > > 28 > > > 00 00 00 > > > [ 252.963007] RIP: scsi_format_opcode_name+0x1a/0x1c0 RSP: > > > c90001bd3c50 > > > [ 252.963007] CR2: > > > [ 252.963007] ---[ end trace 83c5bddfbaa6573c ]--- > > > [ 252.963007] Kernel panic - not syncing: Fatal exception > > > [ 252.963007] Dumping ftrace buffer: > > > [ 252.963007](ftrace buffer empty) > > > [ 252.963007] Kernel Offset: disabled > > > [ 252.963007] ---[ end Kernel panic - not syncing:
Re: [PATCH V4] scsi_debugfs: fix crash in scsi_show_rq()
On Fri, Nov 10, 2017 at 08:51:58AM -0800, James Bottomley wrote: > On Fri, 2017-11-10 at 17:01 +0800, Ming Lei wrote: > > cmd->cmnd can be allocated/freed dynamically in case of > > T10_PI_TYPE2_PROTECTION, > > so we should check it in scsi_show_rq() because this request may have > > been freed already here, and cmd->cmnd has been set as null. > > > > We choose to accept read-after-free and dump request data as far as > > possible. > > > > This patch fixs the following kernel crash when dumping request via > > block's > > debugfs interface: > > > > [ 252.962045] BUG: unable to handle kernel NULL pointer dereference > > at (null) > > [ 252.963007] IP: scsi_format_opcode_name+0x1a/0x1c0 > > [ 252.963007] PGD 25e75a067 P4D 25e75a067 PUD 25e75b067 PMD 0 > > [ 252.963007] Oops: [#1] PREEMPT SMP > > [ 252.963007] Dumping ftrace buffer: > > [ 252.963007](ftrace buffer empty) > > [ 252.963007] Modules linked in: scsi_debug ebtable_filter ebtables > > ip6table_filter ip6_tables xt_CHECKSUM iptable_mangle ipt_MASQUERADE > > nf_nat_masquerade_ipv4 iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 > > nf_nat_ipv4 nf_nat nf_conntrack libcrc32c bridge stp llc > > iptable_filter fuse ip_tables sd_mod sg mptsas mptscsih mptbase > > crc32c_intel ahci libahci nvme serio_raw scsi_transport_sas libata > > lpc_ich nvme_core virtio_scsi binfmt_misc dm_mod iscsi_tcp > > libiscsi_tcp libiscsi scsi_transport_iscsi null_blk configs > > [ 252.963007] CPU: 1 PID: 1881 Comm: cat Not tainted 4.14.0- > > rc2.blk_mq_io_hang+ #516 > > [ 252.963007] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), > > BIOS 1.9.3-1.fc25 04/01/2014 > > [ 252.963007] task: 88025e6f6000 task.stack: c90001bd > > [ 252.963007] RIP: 0010:scsi_format_opcode_name+0x1a/0x1c0 > > [ 252.963007] RSP: 0018:c90001bd3c50 EFLAGS: 00010286 > > [ 252.963007] RAX: 4843 RBX: 0050 RCX: > > > > [ 252.963007] RDX: RSI: 0050 RDI: > > c90001bd3cd8 > > [ 252.963007] RBP: c90001bd3c88 R08: 1000 R09: > > > > [ 252.963007] R10: 880275134000 R11: 88027513406c R12: > > 0050 > > [ 252.963007] R13: c90001bd3cd8 R14: R15: > > > > [ 252.963007] FS: 7f4d11762700() GS:88027fc4() > > knlGS: > > [ 252.963007] CS: 0010 DS: ES: CR0: 80050033 > > [ 252.963007] CR2: CR3: 00025e789003 CR4: > > 003606e0 > > [ 252.963007] DR0: DR1: DR2: > > > > [ 252.963007] DR3: DR6: fffe0ff0 DR7: > > 0400 > > [ 252.963007] Call Trace: > > [ 252.963007] __scsi_format_command+0x27/0xc0 > > [ 252.963007] scsi_show_rq+0x5c/0xc0 > > [ 252.963007] ? seq_printf+0x4e/0x70 > > [ 252.963007] ? blk_flags_show+0x5b/0xf0 > > [ 252.963007] __blk_mq_debugfs_rq_show+0x116/0x130 > > [ 252.963007] blk_mq_debugfs_rq_show+0xe/0x10 > > [ 252.963007] seq_read+0xfe/0x3b0 > > [ 252.963007] ? __handle_mm_fault+0x631/0x1150 > > [ 252.963007] full_proxy_read+0x54/0x90 > > [ 252.963007] __vfs_read+0x37/0x160 > > [ 252.963007] ? security_file_permission+0x9b/0xc0 > > [ 252.963007] vfs_read+0x96/0x130 > > [ 252.963007] SyS_read+0x55/0xc0 > > [ 252.963007] entry_SYSCALL_64_fastpath+0x1a/0xa5 > > [ 252.963007] RIP: 0033:0x7f4d1127e9b0 > > [ 252.963007] RSP: 002b:7ffd27082568 EFLAGS: 0246 ORIG_RAX: > > > > [ 252.963007] RAX: ffda RBX: 7f4d1154bb20 RCX: > > 7f4d1127e9b0 > > [ 252.963007] RDX: 0002 RSI: 7f4d115a7000 RDI: > > 0003 > > [ 252.963007] RBP: 00021010 R08: R09: > > > > [ 252.963007] R10: 037b R11: 0246 R12: > > 00022000 > > [ 252.963007] R13: 7f4d1154bb78 R14: 1000 R15: > > 0002 > > [ 252.963007] Code: c6 e8 1b ca 24 00 eb 8c e8 74 2c ae ff 0f 1f 40 > > 00 0f 1f 44 00 00 55 48 89 e5 41 56 41 55 41 54 53 49 89 fd 49 89 f4 > > 48 83 ec 18 <44> 0f b6 32 48 c7 45 c8 00 00 00 00 65 48 8b 04 25 28 > > 00 00 00 > > [ 252.963007] RIP: scsi_format_opcode_name+0x1a/0x1c0 RSP: > > c90001bd3c50 > > [ 252.963007] CR2: > > [ 252.963007] ---[ end trace 83c5bddfbaa6573c ]--- > > [ 252.963007] Kernel panic - not syncing: Fatal exception > > [ 252.963007] Dumping ftrace buffer: > > [ 252.963007](ftrace buffer empty) > > [ 252.963007] Kernel Offset: disabled > > [ 252.963007] ---[ end Kernel panic - not syncing: Fatal exception > > > > Fixes: 0eebd005dd07(scsi: Implement blk_mq_ops.show_rq()) > > Cc: Bart Van Assche> > Cc: Omar Sandoval > > Cc: Martin K. Petersen > > Cc: James Bottomley > > Cc: Hannes Reinecke
Re: [PATCH V4] scsi_debugfs: fix crash in scsi_show_rq()
On Fri, 2017-11-10 at 17:01 +0800, Ming Lei wrote: > cmd->cmnd can be allocated/freed dynamically in case of > T10_PI_TYPE2_PROTECTION, > so we should check it in scsi_show_rq() because this request may have > been freed already here, and cmd->cmnd has been set as null. > > We choose to accept read-after-free and dump request data as far as > possible. > > This patch fixs the following kernel crash when dumping request via > block's > debugfs interface: > > [ 252.962045] BUG: unable to handle kernel NULL pointer dereference > at (null) > [ 252.963007] IP: scsi_format_opcode_name+0x1a/0x1c0 > [ 252.963007] PGD 25e75a067 P4D 25e75a067 PUD 25e75b067 PMD 0 > [ 252.963007] Oops: [#1] PREEMPT SMP > [ 252.963007] Dumping ftrace buffer: > [ 252.963007](ftrace buffer empty) > [ 252.963007] Modules linked in: scsi_debug ebtable_filter ebtables > ip6table_filter ip6_tables xt_CHECKSUM iptable_mangle ipt_MASQUERADE > nf_nat_masquerade_ipv4 iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 > nf_nat_ipv4 nf_nat nf_conntrack libcrc32c bridge stp llc > iptable_filter fuse ip_tables sd_mod sg mptsas mptscsih mptbase > crc32c_intel ahci libahci nvme serio_raw scsi_transport_sas libata > lpc_ich nvme_core virtio_scsi binfmt_misc dm_mod iscsi_tcp > libiscsi_tcp libiscsi scsi_transport_iscsi null_blk configs > [ 252.963007] CPU: 1 PID: 1881 Comm: cat Not tainted 4.14.0- > rc2.blk_mq_io_hang+ #516 > [ 252.963007] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), > BIOS 1.9.3-1.fc25 04/01/2014 > [ 252.963007] task: 88025e6f6000 task.stack: c90001bd > [ 252.963007] RIP: 0010:scsi_format_opcode_name+0x1a/0x1c0 > [ 252.963007] RSP: 0018:c90001bd3c50 EFLAGS: 00010286 > [ 252.963007] RAX: 4843 RBX: 0050 RCX: > > [ 252.963007] RDX: RSI: 0050 RDI: > c90001bd3cd8 > [ 252.963007] RBP: c90001bd3c88 R08: 1000 R09: > > [ 252.963007] R10: 880275134000 R11: 88027513406c R12: > 0050 > [ 252.963007] R13: c90001bd3cd8 R14: R15: > > [ 252.963007] FS: 7f4d11762700() GS:88027fc4() > knlGS: > [ 252.963007] CS: 0010 DS: ES: CR0: 80050033 > [ 252.963007] CR2: CR3: 00025e789003 CR4: > 003606e0 > [ 252.963007] DR0: DR1: DR2: > > [ 252.963007] DR3: DR6: fffe0ff0 DR7: > 0400 > [ 252.963007] Call Trace: > [ 252.963007] __scsi_format_command+0x27/0xc0 > [ 252.963007] scsi_show_rq+0x5c/0xc0 > [ 252.963007] ? seq_printf+0x4e/0x70 > [ 252.963007] ? blk_flags_show+0x5b/0xf0 > [ 252.963007] __blk_mq_debugfs_rq_show+0x116/0x130 > [ 252.963007] blk_mq_debugfs_rq_show+0xe/0x10 > [ 252.963007] seq_read+0xfe/0x3b0 > [ 252.963007] ? __handle_mm_fault+0x631/0x1150 > [ 252.963007] full_proxy_read+0x54/0x90 > [ 252.963007] __vfs_read+0x37/0x160 > [ 252.963007] ? security_file_permission+0x9b/0xc0 > [ 252.963007] vfs_read+0x96/0x130 > [ 252.963007] SyS_read+0x55/0xc0 > [ 252.963007] entry_SYSCALL_64_fastpath+0x1a/0xa5 > [ 252.963007] RIP: 0033:0x7f4d1127e9b0 > [ 252.963007] RSP: 002b:7ffd27082568 EFLAGS: 0246 ORIG_RAX: > > [ 252.963007] RAX: ffda RBX: 7f4d1154bb20 RCX: > 7f4d1127e9b0 > [ 252.963007] RDX: 0002 RSI: 7f4d115a7000 RDI: > 0003 > [ 252.963007] RBP: 00021010 R08: R09: > > [ 252.963007] R10: 037b R11: 0246 R12: > 00022000 > [ 252.963007] R13: 7f4d1154bb78 R14: 1000 R15: > 0002 > [ 252.963007] Code: c6 e8 1b ca 24 00 eb 8c e8 74 2c ae ff 0f 1f 40 > 00 0f 1f 44 00 00 55 48 89 e5 41 56 41 55 41 54 53 49 89 fd 49 89 f4 > 48 83 ec 18 <44> 0f b6 32 48 c7 45 c8 00 00 00 00 65 48 8b 04 25 28 > 00 00 00 > [ 252.963007] RIP: scsi_format_opcode_name+0x1a/0x1c0 RSP: > c90001bd3c50 > [ 252.963007] CR2: > [ 252.963007] ---[ end trace 83c5bddfbaa6573c ]--- > [ 252.963007] Kernel panic - not syncing: Fatal exception > [ 252.963007] Dumping ftrace buffer: > [ 252.963007](ftrace buffer empty) > [ 252.963007] Kernel Offset: disabled > [ 252.963007] ---[ end Kernel panic - not syncing: Fatal exception > > Fixes: 0eebd005dd07(scsi: Implement blk_mq_ops.show_rq()) > Cc: Bart Van Assche> Cc: Omar Sandoval > Cc: Martin K. Petersen > Cc: James Bottomley > Cc: Hannes Reinecke > Signed-off-by: Ming Lei > --- > V2: > - fix typo > V3: > - prefer to dump data and accept read-after-free > - add some comment > V4: > - read the two variable into local variable first, for avoiding > free between