I/O to a disk via read/write is not limited by the number of segments allowed
by the host adapter; the kernel can split requests if needed, and the limit
imposed by the host adapter can be very low (256k or so) to avoid that SG_IO
returns EINVAL if memory is heavily fragmented.

Since this value is only interesting for SG_IO-based I/O, do not include
it in the max_transfer and only take it into account when patching the
block limits VPD page in the scsi-generic device.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 block/file-posix.c     | 3 +--
 hw/scsi/scsi-generic.c | 6 ++++--
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/block/file-posix.c b/block/file-posix.c
index 58db526cc2..e3241a0dd3 100644
--- a/block/file-posix.c
+++ b/block/file-posix.c
@@ -1239,8 +1239,7 @@ static void raw_refresh_limits(BlockDriverState *bs, Error **errp)
 
         ret = sg_get_max_segments(s->fd);
         if (ret > 0) {
-            bs->bl.max_transfer = MIN(bs->bl.max_transfer,
-                                      ret * qemu_real_host_page_size);
+            bs->bl.max_iov = ret;
         }
     }
 
diff --git a/hw/scsi/scsi-generic.c b/hw/scsi/scsi-generic.c
index 98c30c5d5c..82e1e2ee79 100644
--- a/hw/scsi/scsi-generic.c
+++ b/hw/scsi/scsi-generic.c
@@ -179,10 +179,12 @@ static void scsi_handle_inquiry_reply(SCSIGenericReq *r, SCSIDevice *s)
         (r->req.cmd.buf[1] & 0x01)) {
         page = r->req.cmd.buf[2];
         if (page == 0xb0) {
-            uint32_t max_transfer =
-                blk_get_max_transfer(s->conf.blk) / s->blocksize;
+            uint32_t max_transfer = blk_get_max_transfer(s->conf.blk);
+            uint32_t max_iov = blk_get_max_iov(s->conf.blk);
 
             assert(max_transfer);
+            max_transfer = MIN_NON_ZERO(max_transfer, max_iov * qemu_real_host_page_size)
+                / s->blocksize;
             stl_be_p(&r->buf[8], max_transfer);
             /* Also take care of the opt xfer len. */
             stl_be_p(&r->buf[12],
-- 
2.31.1

On 03.06.21 15:37, Paolo Bonzini wrote:
> I/O to a disk via read/write is not limited by the number of segments allowed
> by the host adapter; the kernel can split requests if needed, and the limit
> imposed by the host adapter can be very low (256k or so) to avoid that SG_IO
> returns EINVAL if memory is heavily fragmented.
>
> Since this value is only interesting for SG_IO-based I/O, do not include
> it in the max_transfer and only take it into account when patching the
> block limits VPD page in the scsi-generic device.
>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>   block/file-posix.c     | 3 +--
>   hw/scsi/scsi-generic.c | 6 ++++--
>   2 files changed, 5 insertions(+), 4 deletions(-)

Reviewed-by: Max Reitz <mreitz@redhat.com>

On 03.06.21 15:37, Paolo Bonzini wrote:
> I/O to a disk via read/write is not limited by the number of segments allowed
> by the host adapter; the kernel can split requests if needed, and the limit
> imposed by the host adapter can be very low (256k or so) to avoid that SG_IO
> returns EINVAL if memory is heavily fragmented.
>
> Since this value is only interesting for SG_IO-based I/O, do not include
> it in the max_transfer and only take it into account when patching the
> block limits VPD page in the scsi-generic device.
>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>   block/file-posix.c     | 3 +--
>   hw/scsi/scsi-generic.c | 6 ++++--
>   2 files changed, 5 insertions(+), 4 deletions(-)
>
> diff --git a/block/file-posix.c b/block/file-posix.c
> index 58db526cc2..e3241a0dd3 100644
> --- a/block/file-posix.c
> +++ b/block/file-posix.c
> @@ -1239,8 +1239,7 @@ static void raw_refresh_limits(BlockDriverState *bs, Error **errp)
>   
>           ret = sg_get_max_segments(s->fd);
>           if (ret > 0) {
> -            bs->bl.max_transfer = MIN(bs->bl.max_transfer,
> -                                      ret * qemu_real_host_page_size);
> +            bs->bl.max_iov = ret;
>           }
>       }
>   
> diff --git a/hw/scsi/scsi-generic.c b/hw/scsi/scsi-generic.c
> index 98c30c5d5c..82e1e2ee79 100644
> --- a/hw/scsi/scsi-generic.c
> +++ b/hw/scsi/scsi-generic.c
> @@ -179,10 +179,12 @@ static void scsi_handle_inquiry_reply(SCSIGenericReq *r, SCSIDevice *s)
>           (r->req.cmd.buf[1] & 0x01)) {
>           page = r->req.cmd.buf[2];
>           if (page == 0xb0) {
> -            uint32_t max_transfer =
> -                blk_get_max_transfer(s->conf.blk) / s->blocksize;
> +            uint32_t max_transfer = blk_get_max_transfer(s->conf.blk);
> +            uint32_t max_iov = blk_get_max_iov(s->conf.blk);
>   
>               assert(max_transfer);
> +            max_transfer = MIN_NON_ZERO(max_transfer, max_iov * qemu_real_host_page_size)
> +                / s->blocksize;

Now that I ran checkpatch for patch 3, I saw that it complains about 
this line being longer than 80 characters. I think it could be split so 
it doesn’t exceed that limit. It looks a bit like you intentionally 
exceeded the warning limit, but didn’t exceed the error limit (of 90). 
Is that so?

Max

Yes, I try to avoid the 90 character limit (though sometimes that's hard
too, hello QOM) but I don't think much of the lower one.

Paolo

Il mar 15 giu 2021, 18:20 Max Reitz <mreitz@redhat.com> ha scritto:

> On 03.06.21 15:37, Paolo Bonzini wrote:
> > I/O to a disk via read/write is not limited by the number of segments
> allowed
> > by the host adapter; the kernel can split requests if needed, and the
> limit
> > imposed by the host adapter can be very low (256k or so) to avoid that
> SG_IO
> > returns EINVAL if memory is heavily fragmented.
> >
> > Since this value is only interesting for SG_IO-based I/O, do not include
> > it in the max_transfer and only take it into account when patching the
> > block limits VPD page in the scsi-generic device.
> >
> > Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> > ---
> >   block/file-posix.c     | 3 +--
> >   hw/scsi/scsi-generic.c | 6 ++++--
> >   2 files changed, 5 insertions(+), 4 deletions(-)
> >
> > diff --git a/block/file-posix.c b/block/file-posix.c
> > index 58db526cc2..e3241a0dd3 100644
> > --- a/block/file-posix.c
> > +++ b/block/file-posix.c
> > @@ -1239,8 +1239,7 @@ static void raw_refresh_limits(BlockDriverState
> *bs, Error **errp)
> >
> >           ret = sg_get_max_segments(s->fd);
> >           if (ret > 0) {
> > -            bs->bl.max_transfer = MIN(bs->bl.max_transfer,
> > -                                      ret * qemu_real_host_page_size);
> > +            bs->bl.max_iov = ret;
> >           }
> >       }
> >
> > diff --git a/hw/scsi/scsi-generic.c b/hw/scsi/scsi-generic.c
> > index 98c30c5d5c..82e1e2ee79 100644
> > --- a/hw/scsi/scsi-generic.c
> > +++ b/hw/scsi/scsi-generic.c
> > @@ -179,10 +179,12 @@ static void
> scsi_handle_inquiry_reply(SCSIGenericReq *r, SCSIDevice *s)
> >           (r->req.cmd.buf[1] & 0x01)) {
> >           page = r->req.cmd.buf[2];
> >           if (page == 0xb0) {
> > -            uint32_t max_transfer =
> > -                blk_get_max_transfer(s->conf.blk) / s->blocksize;
> > +            uint32_t max_transfer = blk_get_max_transfer(s->conf.blk);
> > +            uint32_t max_iov = blk_get_max_iov(s->conf.blk);
> >
> >               assert(max_transfer);
> > +            max_transfer = MIN_NON_ZERO(max_transfer, max_iov *
> qemu_real_host_page_size)
> > +                / s->blocksize;
>
> Now that I ran checkpatch for patch 3, I saw that it complains about
> this line being longer than 80 characters. I think it could be split so
> it doesn’t exceed that limit. It looks a bit like you intentionally
> exceeded the warning limit, but didn’t exceed the error limit (of 90).
> Is that so?
>
> Max
>
>