linux-kernel - Re: Re: [PATCH v3 1/1] nvmet-tcp: Fix NULL pointer dereference during release

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <db63eaac-08e5-cb78-9aef-14b3c9955e8b@bytedance.com>
Date:   Fri, 2 Sep 2022 10:37:08 +0800
From:   zhenwei pi <pizhenwei@...edance.com>
To:     Sagi Grimberg <sagi@...mberg.me>, hch@....de
Cc:     kch@...dia.com, linux-nvme@...ts.infradead.org,
        linux-kernel@...r.kernel.org
Subject: Re: Re: [PATCH v3 1/1] nvmet-tcp: Fix NULL pointer dereference during
 release

On 9/1/22 18:52, Sagi Grimberg wrote:
> 
> 
> On 9/1/22 09:14, zhenwei pi wrote:
>> nvmet-tcp frees CMD buffers in nvmet_tcp_uninit_data_in_cmds(),
>> and waits the inflight IO requests in nvmet_sq_destroy(). During wait
>> the inflight IO requests, the callback nvmet_tcp_queue_response()
>> is called from backend after IO complete, this leads a typical
>> Use-After-Free issue like this:
>>
>>   BUG: kernel NULL pointer dereference, address: 0000000000000008
>>   #PF: supervisor read access in kernel mode
>>   #PF: error_code(0x0000) - not-present page
>>   PGD 0 P4D 0
>>   Oops: 0000 [#1] PREEMPT SMP NOPTI
>>   CPU: 14 PID: 207 Comm: kworker/14:1H Kdump: loaded Tainted: 
>> G            E      6.0.0-rc2.bm.1-amd64 #12
>>   Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 
>> rel-1.16.0-0-gd239552ce722-prebuilt.qemu.org 04/01/2014
>>   Workqueue: nvmet_tcp_wq nvmet_tcp_io_work [nvmet_tcp]
>>   RIP: 0010:shash_ahash_digest+0x2b/0x110
>>   Code: 1f 44 00 00 41 57 41 56 41 55 41 54 55 48 89 fd 53 48 89 f3 48 
>> 83 ec 08 44 8b 67 30 45 85 e4 74 1c 48 8b 57 38 b8 00 10 00 00 <44> 8b 
>> 7a 08 44 29 f8 39 42 0c 0f 46 42 0c 41 39 c4 76 43 48 8b 03
>>   RSP: 0018:ffffc900006e3dd8 EFLAGS: 00010206
>>   RAX: 0000000000001000 RBX: ffff888104ac1650 RCX: 0000000000000000
>>   RDX: 0000000000000000 RSI: ffff888104ac1650 RDI: ffff888104ac1600
>>   RBP: ffff888104ac1600 R08: ffff8881073980c8 R09: ffff8881057798b8
>>   R10: 8080808080808080 R11: 0000000000000000 R12: 0000000000001000
>>   R13: 0000000000000000 R14: ffff88810601a1cc R15: ffff888107398000
>>   FS:  0000000000000000(0000) GS:ffff88823fd80000(0000) 
>> knlGS:0000000000000000
>>   CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>>   CR2: 0000000000000008 CR3: 000000010a8e8000 CR4: 0000000000350ee0
>>   Call Trace:
>>    <TASK>
>>    nvmet_tcp_io_work+0xa1c/0xb1c [nvmet_tcp]
>>    ? __switch_to+0x106/0x420
>>    process_one_work+0x1ae/0x380
>>    ? process_one_work+0x380/0x380
>>    worker_thread+0x30/0x360
>>    ? process_one_work+0x380/0x380
>>    kthread+0xe6/0x110
>>    ? kthread_complete_and_exit+0x20/0x20
>>    ret_from_fork+0x1f/0x30
>>
>> Suggested by Sagi, separate nvmet_tcp_uninit_data_in_cmds() into two
>> steps:
>>    uninit data in cmds                  <- new step 1
>>    nvmet_sq_destroy();
>>    cancel_work_sync(&queue->io_work);
>>    free CMD buffers                     <- new step 2
>>
>> Signed-off-by: zhenwei pi <pizhenwei@...edance.com>
>> ---
>>   drivers/nvme/target/tcp.c | 18 ++++++++++++++++--
>>   1 file changed, 16 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
>> index a3694a32f6d5..e18eae5ef90b 100644
>> --- a/drivers/nvme/target/tcp.c
>> +++ b/drivers/nvme/target/tcp.c
>> @@ -1423,14 +1423,27 @@ static void 
>> nvmet_tcp_uninit_data_in_cmds(struct nvmet_tcp_queue *queue)
>>       for (i = 0; i < queue->nr_cmds; i++, cmd++) {
>>           if (nvmet_tcp_need_data_in(cmd))
>>               nvmet_req_uninit(&cmd->req);
>> +    }
>> +
>> +    if (!queue->nr_cmds && nvmet_tcp_need_data_in(&queue->connect)) {
>> +        /* failed in connect */
>> +        nvmet_req_uninit(&queue->connect.req);
>> +    }
>> +}
>> +static void nvmet_tcp_free_cmd_data_in_buffers(struct nvmet_tcp_queue 
>> *queue)
>> +{
>> +    struct nvmet_tcp_cmd *cmd = queue->cmds;
>> +    int i;
>> +
>> +    for (i = 0; i < queue->nr_cmds; i++, cmd++) {
> 
>          if (nvmet_tcp_need_data_in(cmd))
> 
>>           nvmet_tcp_unmap_pdu_iovec(cmd);
> 
> This function should no longer exist in the code.
> 

Hi, Sagi

Sorry, I didn't notice this patch is on the way...
https://lore.kernel.org/linux-nvme/Yw%2F0cZFCQQpx3hhZ@ZenIV/T/#t

I'll send the v4 after this patch apply. Thanks!

>>           nvmet_tcp_free_cmd_buffers(cmd);
>>       }
>>       if (!queue->nr_cmds && nvmet_tcp_need_data_in(&queue->connect)) {
>> -        /* failed in connect */
>> -        nvmet_tcp_finish_cmd(&queue->connect);
>> +        nvmet_tcp_unmap_pdu_iovec(&queue->connect);
> 
> This function should no longer exist in the code.
> 
>> +        nvmet_tcp_free_cmd_buffers(&queue->connect);
>>       }
>>   }
>> @@ -1452,6 +1465,7 @@ static void nvmet_tcp_release_queue_work(struct 
>> work_struct *w)
>>       nvmet_tcp_uninit_data_in_cmds(queue);
>>       nvmet_sq_destroy(&queue->nvme_sq);
>>       cancel_work_sync(&queue->io_work);
>> +    nvmet_tcp_free_cmd_data_in_buffers(queue);
>>       sock_release(queue->sock);
>>       nvmet_tcp_free_cmds(queue);
>>       if (queue->hdr_digest || queue->data_digest)

-- 
zhenwei pi