From b353314d439095d948cdf257c07f25f640850032 Mon Sep 17 00:00:00 2001 Message-Id: In-Reply-To: References: From: Fam Zheng Date: Fri, 24 Apr 2015 08:44:47 -0500 Subject: [CHANGE 27/29] scsi: fix cancellation when I/O was completed but DMA was not. To: rhvirt-patches@redhat.com, jen@redhat.com RH-Author: Fam Zheng Message-id: <1429865088-13298-28-git-send-email-famz@redhat.com> Patchwork-id: 64928 O-Subject: [RHEL-6.7 qemu-kvm PATCH v7 27/28] scsi: fix cancellation when I/O was completed but DMA was not. Bugzilla: 1069519 RH-Acked-by: Paolo Bonzini RH-Acked-by: Stefan Hajnoczi RH-Acked-by: Max Reitz From: Paolo Bonzini Commit d577646 (scsi: Introduce scsi_req_cancel_complete, 2014-09-25) was supposed to have no semantic change, but it missed a case. When r->aiocb has already been NULLed, but DMA was not complete and the SCSI layer was waiting for scsi_req_continue, after the patch the SCSI layer will not call the .cancel callback of SCSIBusInfo. Fixes: d5776465ee9a55815792efa34d79de240f4ffd99 Cc: qemu-stable@nongnu.org Reported-by: Dr. David Alan Gilbert Tested-by: Dr. David Alan Gilbert Reviewed-by: Fam Zheng Signed-off-by: Paolo Bonzini (cherry picked from commit 488eef2f1d16c97cf7f9ebf644ecafa1ea1e9acc) Signed-off-by: Fam Zheng Signed-off-by: Jeff E. Nelson Conflicts: hw/scsi-bus.c Context conflict. Upstream the block device interface is now changed to "blk_aio_cancel", whereas we still use bdrv_aio_cancel in downstream. --- hw/scsi-bus.c | 2 ++ 1 file changed, 2 insertions(+) Signed-off-by: Jeff E. Nelson --- hw/scsi-bus.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/hw/scsi-bus.c b/hw/scsi-bus.c index 762d71f..b53764c 100644 --- a/hw/scsi-bus.c +++ b/hw/scsi-bus.c @@ -1521,6 +1521,8 @@ void scsi_req_cancel(SCSIRequest *req) req->io_canceled = true; if (req->aiocb) { bdrv_aio_cancel(req->aiocb); + } else { + scsi_req_cancel_complete(req); } } -- 2.1.0