> diff --git a/fs/io_uring.c b/fs/io_uring.c > index 86a2bd721900..485832deb7ea 100644
Jens Thanks! I tested that and it seems to resolve the GPF. I now get a Bad file descriptor error in fio (which is I think what we'd expect). If you turn that patch into an official kernel patch feel free to add Tested-by: Stephen Bates <[email protected]> batesste@io_uring-vm1:~/io_uring-test$ sudo fio fio/io_uring.fio io_uring: (g=0): rw=rw, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=io_uring, iodepth=1 fio-3.14-7-g7184a Starting 1 process fio: io_u error on file /dev/nullb0: Bad file descriptor: read offset=0, buflen=4096 Cheers Stephen On 2019-06-21, 2:13 PM, "Jens Axboe" <[email protected]> wrote: On 6/21/19 6:40 AM, Stephen Bates wrote: > Hi > > I hit the following General Protection Fault when testing io_uring via the io_uring engine in fio. This was on a VM running 5.2-rc5 and the latest version of fio. The issue occurs for both null_blk and fake NVMe drives. I have not tested bare metal or real NVMe SSDs. The fio script used is given below. > > [io_uring] > time_based=1 > runtime=60 > filename=/dev/nvme2n1 (note /dev/nullb0 also fails) > ioengine=io_uring > bs=4k > rw=readwrite > direct=1 > fixedbufs=1 > sqthread_poll=1 > sqthread_poll_cpu=0 > > [ 964.540374] general protection fault: 0000 [#1] SMP PTI > [ 964.542041] CPU: 0 PID: 872 Comm: io_uring-sq Not tainted 5.2.0-rc5-cpacket-io-uring #1 > [ 964.545589] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014 > [ 964.549761] RIP: 0010:fput_many+0x7/0x90 > [ 964.551522] Code: 01 48 85 ff 74 17 55 48 89 e5 53 48 8b 1f e8 a0 f9 ff ff 48 85 db 48 89 df 75 f0 5b 5d f3 c3 0f 1f 40 00 0f 1f 44 00 00 89 f6 <f0> 48 29 77 38 74 01 c3 55 48 89 e5 53 48 89 fb 65 48 \ > 8b 3c 25 c0 > [ 964.559031] RSP: 0018:ffffadeb817ebc50 EFLAGS: 00010246 > [ 964.561112] RAX: 0000000000000004 RBX: ffff8f46ad477480 RCX: 0000000000001805 > [ 964.563911] RDX: 0000000000000000 RSI: 0000000000000001 RDI: f18b51b9a39552b5 > [ 964.566580] RBP: ffffadeb817ebc58 R08: ffff8f46b7a318c0 R09: 000000000000015d > [ 964.569109] R10: ffffadeb817ebce8 R11: 0000000000000020 R12: ffff8f46ad4cd000 > [ 964.571623] R13: 00000000fffffff7 R14: ffffadeb817ebe30 R15: 0000000000000004 > [ 964.574153] FS: 0000000000000000(0000) GS:ffff8f46b7a00000(0000) knlGS:0000000000000000 > [ 964.577020] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 > [ 964.578917] CR2: 000055828f0bbbf0 CR3: 0000000232176004 CR4: 00000000003606f0 > [ 964.581221] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 > [ 964.583511] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 > [ 964.585808] Call Trace: > [ 964.586626] ? fput+0x13/0x20 > [ 964.587613] io_free_req+0x20/0x40 > [ 964.588733] io_put_req+0x1b/0x20 > [ 964.589795] io_submit_sqe+0x40a/0x680 > [ 964.590919] ? __switch_to_asm+0x34/0x70 > [ 964.592090] ? __switch_to_asm+0x40/0x70 > [ 964.593270] io_submit_sqes+0xb9/0x160 > [ 964.594392] ? io_submit_sqes+0xb9/0x160 > [ 964.595564] ? __switch_to_asm+0x40/0x70 > [ 964.596737] ? __switch_to_asm+0x34/0x70 > [ 964.597918] ? __schedule+0x3f2/0x6a0 > [ 964.599015] ? __switch_to_asm+0x34/0x70 > [ 964.600444] io_sq_thread+0x1af/0x470 > [ 964.601568] ? __switch_to_asm+0x34/0x70 > [ 964.602655] ? wait_woken+0x80/0x80 > [ 964.603625] ? __switch_to+0x85/0x410 > [ 964.604638] ? __switch_to_asm+0x40/0x70 > [ 964.605726] ? __switch_to_asm+0x34/0x70 > [ 964.606811] ? __schedule+0x3f2/0x6a0 > [ 964.607827] kthread+0x105/0x140 > [ 964.608725] ? io_submit_sqes+0x160/0x160 > [ 964.609836] ? kthread+0x105/0x140 > [ 964.610780] ? io_submit_sqes+0x160/0x160 > [ 964.611887] ? kthread_destroy_worker+0x50/0x50 > [ 964.613158] ret_from_fork+0x35/0x40 > [ 964.614148] Modules linked in: crct10dif_pclmul crc32_pclmul ghash_clmulni_intel aesni_intel aes_x86_64 crypto_simd cryptd glue_helper joydev input_leds serio_raw mac_hid sch_fq_codel sunrpc null_blk \ > ip_tables x_tables autofs4 8139too psmouse 8139cp floppy mii i2c_piix4 pata_acpi > [ 964.620856] ---[ end trace bdbba818b310272c ]--- Try this patch. Technically, it's not valid to use sqthread without fixed files registered through io_uring_register(), and this case looks to me like we're just not initializing ->file before we end up failing the request due to a violation of that requirement. Not tested, on vacation... diff --git a/fs/io_uring.c b/fs/io_uring.c index 86a2bd721900..485832deb7ea 100644 --- a/fs/io_uring.c +++ b/fs/io_uring.c @@ -579,6 +579,7 @@ static struct io_kiocb *io_get_req(struct io_ring_ctx *ctx, state->cur_req++; } + req->file = NULL; req->ctx = ctx; req->flags = 0; /* one is dropped after submission, the other at completion */ @@ -1801,10 +1802,8 @@ static int io_req_set_file(struct io_ring_ctx *ctx, const struct sqe_submit *s, req->sequence = ctx->cached_sq_head - 1; } - if (!io_op_needs_file(s->sqe)) { - req->file = NULL; + if (!io_op_needs_file(s->sqe)) return 0; - } if (flags & IOSQE_FIXED_FILE) { if (unlikely(!ctx->user_files || -- Jens Axboe
