Re: [Qemu-devel] [PATCH] scsi-block: Handle error from host devices

[Daniel Henrique Barboza]

Hi Fam,

I've tried this patch and found issues when booting a VM using SCSI 
passthrough. This is the backtrace from gdb from the segfault that 
happens in the middle of kernel boot:

Thread 1 "qemu-system-ppc" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7ffff7ff63a0 (LWP 16830)]
0x0000000000000000 in ?? ()
(gdb) bt
#0  0x0000000000000000 in ?? ()
#1  0x00000001007b8da8 in scsi_block_sgio_cb (opaque=0x10212e580, ret=0) 
at /home/danielhb/qemu/hw/scsi/scsi-disk.c:2772
#2  0x0000000100993f68 in blk_aio_complete (acb=0x101909520) at 
/home/danielhb/qemu/block/block-backend.c:1331
#3  0x0000000100994ccc in blk_aio_ioctl_entry (opaque=0x101909520) at 
/home/danielhb/qemu/block/block-backend.c:1542
#4  0x0000000100ac0954 in coroutine_trampoline (i0=28666944, i1=1) at 
/home/danielhb/qemu/util/coroutine-ucontext.c:116
#5  0x00007ffff789574c in makecontext () at 
../sysdeps/unix/sysv/linux/powerpc/powerpc64/makecontext.S:136
#6  0x0000000000000000 in ?? ()

The segfault is happening at this line:

static void scsi_block_sgio_cb(void *opaque, int ret)
{
    SCSIBlockReq *req = opaque;

    if (!ret &&
        (req->io_header.status ||
         req->io_header.host_status ||
         req->io_header.driver_status)) {
        ret = -EIO;
    }
    req->cb(req->cb_opaque, ret); <-----------------
}


This is happening because inside scsi_block_do_sgio you're not setting 
req->cb, just req->cb_opaque. Setting req->cb made the VM boot again:


--- a/hw/scsi/scsi-disk.c
+++ b/hw/scsi/scsi-disk.c
@@ -2851,6 +2851,7 @@ static BlockAIOCB *scsi_block_do_sgio(SCSIBlockReq 
*req,
     io_header->usr_ptr = r;
     io_header->flags |= SG_FLAG_DIRECT_IO;

+    req->cb = cb;
     req->cb_opaque = opaque;
     aiocb = blk_aio_ioctl(s->qdev.conf.blk, SG_IO, io_header,
                           scsi_block_sgio_cb, req);




Thanks,


Daniel

On 04/02/2018 12:23 AM, Fam Zheng wrote:
> The callback of blk_aio_ioctl is not sensible to SCSI errors, so
> werror=stop doesn't work if ioctl returns 0 but the scsi status is
> error.
>
> Peek at the sg_io_hdr_t fields and amend ret to fix that.
>
> Signed-off-by: Fam Zheng <address@hidden>
> ---
>   hw/scsi/scsi-disk.c | 20 +++++++++++++++++++-
>   1 file changed, 19 insertions(+), 1 deletion(-)
>
> diff --git a/hw/scsi/scsi-disk.c b/hw/scsi/scsi-disk.c
> index f5ab767ab5..2c43830586 100644
> --- a/hw/scsi/scsi-disk.c
> +++ b/hw/scsi/scsi-disk.c
> @@ -2651,10 +2651,26 @@ typedef struct SCSIBlockReq {
>       /* Selected bytes of the original CDB, copied into our own CDB.  */
>       uint8_t cmd, cdb1, group_number;
>
> +    BlockCompletionFunc *cb;
> +    void *cb_opaque;
> +
>       /* CDB passed to SG_IO.  */
>       uint8_t cdb[16];
>   } SCSIBlockReq;
>
> +static void scsi_block_sgio_cb(void *opaque, int ret)
> +{
> +    SCSIBlockReq *req = opaque;
> +
> +    if (!ret &&
> +        (req->io_header.status ||
> +         req->io_header.host_status ||
> +         req->io_header.driver_status)) {
> +        ret = -EIO;
> +    }
> +    req->cb(req->cb_opaque, ret);
> +}
> +
>   static BlockAIOCB *scsi_block_do_sgio(SCSIBlockReq *req,
>                                         int64_t offset, QEMUIOVector *iov,
>                                         int direction,
> @@ -2734,7 +2750,9 @@ static BlockAIOCB *scsi_block_do_sgio(SCSIBlockReq *req,
>       io_header->usr_ptr = r;
>       io_header->flags |= SG_FLAG_DIRECT_IO;
>
> -    aiocb = blk_aio_ioctl(s->qdev.conf.blk, SG_IO, io_header, cb, opaque);
> +    req->cb_opaque = opaque;
> +    aiocb = blk_aio_ioctl(s->qdev.conf.blk, SG_IO, io_header,
> +                          scsi_block_sgio_cb, req);
>       assert(aiocb != NULL);
>       return aiocb;
>   }