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

[Fam Zheng <famz@redhat.com>]

On Tue, 04/03 17:41, Daniel Henrique Barboza wrote:
> 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:

Oops, thanks. Apparently I failed to test my patch, sorry. :(

Fam