[PATCH v4 2/7] scsi-generic: pass max_segments via max_iov field in BlockLimits
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 --- 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 536998a1d6..670c577bfe 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
Re: [PATCH v4 2/7] scsi-generic: pass max_segments via max_iov field in BlockLimits
On Tue, Jun 08, 2021 at 03:16:29PM +0200, 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. to avoid SG_IO returning EINVAL > > 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 > --- > block/file-posix.c | 3 +-- > hw/scsi/scsi-generic.c | 6 -- > 2 files changed, 5 insertions(+), 4 deletions(-) > Reviewed-by: Eric Blake -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [PATCH v4 2/7] scsi-generic: pass max_segments via max_iov field in BlockLimits
On Tue, 2021-06-08 at 15:16 +0200, 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 > --- > 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 536998a1d6..670c577bfe 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; Actually I think that both max transfer size and max segement count, are only relevant for SCSI passthrough since kernel I think emualates both for regular I/O, so I think that we shoudn't expose them to qemu at all. In my version of the patches I removed both bl.max_transfer and bl.max_iov setup from the file-posix driver and replaced it with bs->bl.max_ioctl_transfer (you call it max_hw_transfer) In my version the bl.max_ioctl_transfer is a merged limit of the max transfer size and the max iovec number. https://www.mail-archive.com/qemu-devel@nongnu.org/msg768264.html Best regards, Maxim Levitsky > } > } > > 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],