Re: [PATCH] scsi-generic: replace logical block count of response of READ CAPACITY

["Philippe Mathieu-Daudé" <philmd@redhat.com>]

Hi Claudio,

On 12/21/21 10:48, Claudio Fontana wrote:
> Hi Paolo, Hannes,
> 
> any thoughts on the following issue?
> 
> Introduction:
> 
> When using SAN storage for providing block devices to guests, configured as SCSI-passthrough devices, increasing the space available in the VM is a use case.
> 
> To do it, it is currently necessary to:
> 
> 1) expand storage on the actual SAN,
> 2) run a "virsh blockresize" or equivalent command to make QEMU aware of the new size, and finally
> 3) do a "rescan-scsi-bus.sh" or equivalent operation in the guest to make the running guest aware of the increased disk size.
> 
> The problem:
> 
> As of now, the administrator needs to make sure that step 3 won't be done before step 2 has been executed, or the resulting state will be inconsistent.
> In practice this creates organizational issues to try to sync up host/storage admins and guest OS admin, and is therefore error prone (due to these human factors).
> 
> The proposal:
> 
> The patch I replied to here from Ma Lin tries to avoid the inconsistent state, by having "rescan-scsi-bus.sh" still report the old size in the guest until QEMU itself is aware of the new disk size.
> 
> The patch:
> 
> Before the patch, the SCSI READ_CAPACITY command in the guest os directly receives the unmodified response from the storage backend.
> 
> After the patch, QEMU intercepts the READ_CAPACITY response and replaces the maximum LBA with the information which is saved in QEMU.
> 
> This means: after resizing the storage on the SAN backend, the host administrator must explicitly notify about CAPACITY HAS CHANGED by issuing a block resize command through QMP or libvirt,
> even for SCSI passthrough disks.
> 
> Any ideas on this patch or on possible alternatives?

I am not an SCSI expert, but Lin's description and yours sound
coherent to me.

One minor nitpick below.

> On 11/20/21 11:15 AM, Lin Ma wrote:
>> While using SCSI passthrough, Following scenario makes qemu doesn't
>> realized the capacity change of remote scsi target:
>> 1. online resize the scsi target.
>> 2. issue 'rescan-scsi-bus.sh -s ...' in host.
>> 3. issue 'rescan-scsi-bus.sh -s ...' in vm.
>>
>> In above scenario I used to experienced errors while accessing the
>> additional disk space in vm. I think the reasonable operations should
>> be:
>> 1. online resize the scsi target.
>> 2. issue 'rescan-scsi-bus.sh -s ...' in host.
>> 3. issue 'block_resize' via qmp to notify qemu.
>> 4. issue 'rescan-scsi-bus.sh -s ...' in vm.
>>
>> The errors disappear once I notify qemu by block_resize via qmp.
>>
>> So this patch replaces the number of logical blocks of READ CAPACITY
>> response from scsi target by qemu's bs->total_sectors. If the user in
>> vm wants to access the additional disk space, The administrator of
>> host must notify qemu once resizeing the scsi target.
>>
>> Bonus is that domblkinfo of libvirt can reflect the consistent capacity
>> information between host and vm in case of missing block_resize in qemu.
>> E.g:
>> ...
>>     <disk type='block' device='lun'>
>>       <driver name='qemu' type='raw'/>
>>       <source dev='/dev/sdc' index='1'/>
>>       <backingStore/>
>>       <target dev='sda' bus='scsi'/>
>>       <alias name='scsi0-0-0-0'/>
>>       <address type='drive' controller='0' bus='0' target='0' unit='0'/>
>>     </disk>
>> ...
>>
>> Before:
>> 1. online resize the scsi target.
>> 2. host:~  # rescan-scsi-bus.sh -s /dev/sdc
>> 3. guest:~ # rescan-scsi-bus.sh -s /dev/sda
>> 4  host:~  # virsh domblkinfo --domain $DOMAIN --human --device sda
>> Capacity:       4.000 GiB
>> Allocation:     0.000 B
>> Physical:       8.000 GiB
>>
>> 5. guest:~ # lsblk /dev/sda
>> NAME   MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
>> sda      8:0    0   8G  0 disk
>> └─sda1   8:1    0   2G  0 part
>>
>> After:
>> 1. online resize the scsi target.
>> 2. host:~  # rescan-scsi-bus.sh -s /dev/sdc
>> 3. guest:~ # rescan-scsi-bus.sh -s /dev/sda
>> 4  host:~  # virsh domblkinfo --domain $DOMAIN --human --device sda
>> Capacity:       4.000 GiB
>> Allocation:     0.000 B
>> Physical:       8.000 GiB
>>
>> 5. guest:~ # lsblk /dev/sda
>> NAME   MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
>> sda      8:0    0   4G  0 disk
>> └─sda1   8:1    0   2G  0 part
>>
>> Signed-off-by: Lin Ma <lma@suse.com>
>> ---
>>  hw/scsi/scsi-generic.c | 10 ++++++++--
>>  1 file changed, 8 insertions(+), 2 deletions(-)
>>
>> diff --git a/hw/scsi/scsi-generic.c b/hw/scsi/scsi-generic.c
>> index 0306ccc7b1..343b51c2c0 100644
>> --- a/hw/scsi/scsi-generic.c
>> +++ b/hw/scsi/scsi-generic.c
>> @@ -315,11 +315,17 @@ static void scsi_read_complete(void * opaque, int ret)
>>      if (r->req.cmd.buf[0] == READ_CAPACITY_10 &&
>>          (ldl_be_p(&r->buf[0]) != 0xffffffffU || s->max_lba == 0)) {
>>          s->blocksize = ldl_be_p(&r->buf[4]);
>> -        s->max_lba = ldl_be_p(&r->buf[0]) & 0xffffffffULL;
>> +        BlockBackend *blk = s->conf.blk;
>> +        BlockDriverState *bs = blk_bs(blk);
>> +        s->max_lba = bs->total_sectors - 1;

I'd add a refresh_max_lba() helper:

 static void refresh_max_lba(SCSIDevice *s)
 {
     BlockBackend *blk = s->conf.blk;
     BlockDriverState *bs = blk_bs(blk);
     uint64_t max_lba = bs->total_sectors - 1;

     if (max_lba != s->max_lba) {
         trace_scsi_generic_max_lba_refreshed(s->max_lba, max_lba);
         s->max_lba = max_lba;
     }
 }

Otherwise, to the best of my knowledge:
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>

>> +        stl_be_p(&r->buf[0], s->max_lba);
>>      } else if (r->req.cmd.buf[0] == SERVICE_ACTION_IN_16 &&
>>                 (r->req.cmd.buf[1] & 31) == SAI_READ_CAPACITY_16) {
>>          s->blocksize = ldl_be_p(&r->buf[8]);
>> -        s->max_lba = ldq_be_p(&r->buf[0]);
>> +        BlockBackend *blk = s->conf.blk;
>> +        BlockDriverState *bs = blk_bs(blk);
>> +        s->max_lba = bs->total_sectors - 1;
>> +        stq_be_p(&r->buf[0], s->max_lba);
>>      }
>>      blk_set_guest_block_size(s->conf.blk, s->blocksize);
>>  
>>
> 
>