From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:36419) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1aOAOY-0002yn-NP for qemu-devel@nongnu.org; Tue, 26 Jan 2016 15:39:23 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1aOAOX-0003Oy-LD for qemu-devel@nongnu.org; Tue, 26 Jan 2016 15:39:22 -0500 References: <1453225191-11871-1-git-send-email-jsnow@redhat.com> From: John Snow Message-ID: <56A7D972.3050003@redhat.com> Date: Tue, 26 Jan 2016 15:39:14 -0500 MIME-Version: 1.0 In-Reply-To: <1453225191-11871-1-git-send-email-jsnow@redhat.com> Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] [PATCH v2 0/6] ide: fix atapi software reset List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-block@nongnu.org Cc: kwolf@redhat.com, pbonzini@redhat.com, qemu-devel@nongnu.org, stefanha@redhat.com Ping (I know I could just send a PR, but in this case I'd like someone to look over it. I've gone over pretty terrible detail to convince myself this won't break anything IDE-wise, so if it looks sane block-wise, I'd be satisfied with an ACK.) On 01/19/2016 12:39 PM, John Snow wrote: > The ATAPI software reset function is implemented somewhat lackadaisically. > > Firstly, it is valid only for ATAPI drives - not HDs. If a HD should > receive this command while BSY, it should be ignored like any other > command instead of aborted. A non-BSY HD is free to abort the command > in the usual fashion to indicate it doesn't understand or doesn't support > that command. > > Second, for drives that should accept a software reset, they should not > "forget" about all pending AIO during the reset. Since a software reset > resets the DRQ and BSY flags, it is possible to 'stack' multiple > concurrent reads using DMA and alternately chaining software reset and > DMA reads. We mustn't reset BSY/DRQ until we are confident that we > have canceled existing AIO. > > Third, the existing software reset routine does not perform a very > rigorous reset. > > This series corrects this by: > > (1) Correcting ide_exec_cmd to correctly ignore, not abort, software > reset commands for ide-hd devices that are busy executing a command. > > (2) Improving the software reset routine to cancel buffered DMA, then > fall back to synchronously waiting for any pending DMA to finish > before returning, insuring that the reset completes sanely. > > (3) Use existing reset routines to comprehensively reset the device. > > Reported-by: Kevin Wolf > Signed-off-by: John Snow > > ________________________________________________________________________________ > > For convenience, this branch is available at: > https://github.com/jnsnow/qemu.git branch ide-reset-fix > https://github.com/jnsnow/qemu/tree/ide-reset-fix > > This version is tagged ide-reset-fix-v2: > https://github.com/jnsnow/qemu/releases/tag/ide-reset-fix-v2 > > John Snow (6): > ide: Prohibit RESET on IDE drives > ide: code motion > ide: move buffered DMA cancel to core > ide: replace blk_drain_all by blk_drain > ide: Add silent DRQ cancellation > ide: fix device_reset to not ignore pending AIO > > hw/ide/core.c | 215 ++++++++++++++++++++++++++++++++++++------------------ > hw/ide/internal.h | 1 + > hw/ide/pci.c | 36 +-------- > 3 files changed, 144 insertions(+), 108 deletions(-) >