Re: INFO: task syz-executor can't die for more than 143 seconds. (2)

[Bob Liu]

On 10/24/19 6:08 PM, Tetsuo Handa wrote:
> On 2019/10/23 16:56, syzbot wrote:
>> Hello,
>>
>> syzbot found the following crash on:
>>
>> HEAD commit:    c4b9850b Add linux-next specific files for 20191018
>> git tree:       linux-next
>> console output: https://urldefense.proofpoint.com/v2/url?u=https-3A__syzkaller.appspot.com_x_log.txt-3Fx-3D177b3ab0e00000&d=DwICaQ&c=RoP1YumCXCgaWHvlZYR8PZh8Bv7qIrMUB65eapI_JnE&r=1ktT0U2YS_I8Zz2o-MS1YcCAzWZ6hFGtyTgvVMGM7gI&m=wOlNeKk9puri9Fvxn8bGDrlWHd-4GPMeJ9rb2CVqXaE&s=PZfBliKlYjm16VnyPzu-3i0SgqlbByIB0iI8jVhcGuk&e= 
>> kernel config:  https://urldefense.proofpoint.com/v2/url?u=https-3A__syzkaller.appspot.com_x_.config-3Fx-3Dc940ef12efcd1ec&d=DwICaQ&c=RoP1YumCXCgaWHvlZYR8PZh8Bv7qIrMUB65eapI_JnE&r=1ktT0U2YS_I8Zz2o-MS1YcCAzWZ6hFGtyTgvVMGM7gI&m=wOlNeKk9puri9Fvxn8bGDrlWHd-4GPMeJ9rb2CVqXaE&s=z8tV220wKFTQIJH1tSYLUl8ecAnll94C_mFVcHkuTlc&e= 
>> dashboard link: https://urldefense.proofpoint.com/v2/url?u=https-3A__syzkaller.appspot.com_bug-3Fextid-3Db48daca8639150bc5e73&d=DwICaQ&c=RoP1YumCXCgaWHvlZYR8PZh8Bv7qIrMUB65eapI_JnE&r=1ktT0U2YS_I8Zz2o-MS1YcCAzWZ6hFGtyTgvVMGM7gI&m=wOlNeKk9puri9Fvxn8bGDrlWHd-4GPMeJ9rb2CVqXaE&s=VZ2ZdnAqEd_AkXJujidN3EgwscGpUAdsZjuObKjXN-U&e= 
>> compiler:       gcc (GCC) 9.0.0 20181231 (experimental)
>> syz repro:      https://urldefense.proofpoint.com/v2/url?u=https-3A__syzkaller.appspot.com_x_repro.syz-3Fx-3D1356b8ff600000&d=DwICaQ&c=RoP1YumCXCgaWHvlZYR8PZh8Bv7qIrMUB65eapI_JnE&r=1ktT0U2YS_I8Zz2o-MS1YcCAzWZ6hFGtyTgvVMGM7gI&m=wOlNeKk9puri9Fvxn8bGDrlWHd-4GPMeJ9rb2CVqXaE&s=Q_svUYj2OBYmIJXnResNzOWVUCyjRpxnpun2Cu15S9M&e= 
>> C reproducer:   https://urldefense.proofpoint.com/v2/url?u=https-3A__syzkaller.appspot.com_x_repro.c-3Fx-3D14f48687600000&d=DwICaQ&c=RoP1YumCXCgaWHvlZYR8PZh8Bv7qIrMUB65eapI_JnE&r=1ktT0U2YS_I8Zz2o-MS1YcCAzWZ6hFGtyTgvVMGM7gI&m=wOlNeKk9puri9Fvxn8bGDrlWHd-4GPMeJ9rb2CVqXaE&s=FGZNxR7w-rU29MhJxJtno-c_wUXCJHgPC5V1YNp7h58&e= 
> 
> The reproducer is trying to allocate 64TB of disk space on /dev/nullb0 using fallocate()
> but __blkdev_issue_zero_pages() cannot bail out upon SIGKILL (and therefore cannot
> terminate for minutes). Can we make it killable? 

Yeah, I think we can consider add fatal_signal_pending(current) checking in the while() loop..

> I don't know what action is needed
> for undoing this loop...
> 
>         while (nr_sects != 0) {
>                 bio = blk_next_bio(bio, __blkdev_sectors_to_bio_pages(nr_sects),
>                                    gfp_mask);
>                 bio->bi_iter.bi_sector = sector;
>                 bio_set_dev(bio, bdev);
>                 bio_set_op_attrs(bio, REQ_OP_WRITE, 0);
> 
>                 while (nr_sects != 0) {
>                         sz = min((sector_t) PAGE_SIZE, nr_sects << 9);
>                         bi_size = bio_add_page(bio, ZERO_PAGE(0), sz, 0);
>                         nr_sects -= bi_size >> 9;
>                         sector += bi_size >> 9;
>                         if (bi_size < sz)
>                                 break;
>                 }
>                 cond_resched();
>         }
>