| Message ID | 1312472437-26231-1-git-send-email-michael.jones@matrix-vision.de (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers |
Return-path: <linux-media-owner@vger.kernel.org> Envelope-to: mchehab@infradead.org Delivery-date: Thu, 04 Aug 2011 15:41:17 +0000 Received: from casper.infradead.org [85.118.1.10] by localhost.localdomain with IMAP (fetchmail-6.3.17) for <mchehab@localhost> (single-drop); Thu, 04 Aug 2011 12:42:02 -0300 (BRT) Received: from vger.kernel.org ([209.132.180.67]) by casper.infradead.org with esmtp (Exim 4.76 #1 (Red Hat Linux)) id 1Qp02u-00004s-SL; Thu, 04 Aug 2011 15:41:17 +0000 Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754579Ab1HDPlO (ORCPT <rfc822; kmpark@infradead.org> + 1 other); Thu, 4 Aug 2011 11:41:14 -0400 Received: from mail1.matrix-vision.com ([78.47.19.71]:40504 "EHLO mail1.matrix-vision.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751766Ab1HDPlO (ORCPT <rfc822; linux-media@vger.kernel.org>); Thu, 4 Aug 2011 11:41:14 -0400 Received: from mail1.matrix-vision.com (localhost [127.0.0.1]) by mail1.matrix-vision.com (Postfix) with ESMTP id DFEAF722AC; Thu, 4 Aug 2011 17:41:12 +0200 (CEST) Received: from erinome (g2.matrix-vision.com [80.152.136.245]) by mail1.matrix-vision.com (Postfix) with ESMTPA id B592F721B4; Thu, 4 Aug 2011 17:41:12 +0200 (CEST) Received: from erinome (localhost [127.0.0.1]) by erinome (Postfix) with ESMTP id E40796F8A; Thu, 4 Aug 2011 17:41:11 +0200 (CEST) Received: by erinome (Postfix, from userid 108) id D75226F9C; Thu, 4 Aug 2011 17:41:11 +0200 (CEST) Received: from ap437-joe.intern.matrix-vision.de (host65-46.intern.matrix-vision.de [192.168.65.46]) by erinome (Postfix) with ESMTPA id BAF546F8A; Thu, 4 Aug 2011 17:41:11 +0200 (CEST) From: Michael Jones <michael.jones@matrix-vision.de> To: linux-media@vger.kernel.org Cc: Laurent Pinchart <laurent.pinchart@ideasonboard.com>, Mauro Carvalho Chehab <mchehab@infradead.org> Subject: [PATCH] [media] omap3isp: queue: fail QBUF if buffer is too small Date: Thu, 4 Aug 2011 17:40:37 +0200 Message-Id: <1312472437-26231-1-git-send-email-michael.jones@matrix-vision.de> X-Mailer: git-send-email 1.7.6 X-MV-Disclaimer: true (erinome) X-AV-Checked: ClamAV using ClamSMTP (erinome) X-AV-Checked: ClamAV using ClamSMTP (mail1) Sender: linux-media-owner@vger.kernel.org Precedence: bulk List-ID: <linux-media.vger.kernel.org> X-Mailing-List: linux-media@vger.kernel.org |
Commit Message
Michael Jones
Aug. 4, 2011, 3:40 p.m. UTC
Add buffer length to sanity checks for QBUF.
Signed-off-by: Michael Jones <michael.jones@matrix-vision.de>
---
drivers/media/video/omap3isp/ispqueue.c | 3 +++
1 files changed, 3 insertions(+), 0 deletions(-)
Comments
Hi Michael, Thanks for the patch. On Thursday 04 August 2011 17:40:37 Michael Jones wrote: > Add buffer length to sanity checks for QBUF. > > Signed-off-by: Michael Jones <michael.jones@matrix-vision.de> > --- > drivers/media/video/omap3isp/ispqueue.c | 3 +++ > 1 files changed, 3 insertions(+), 0 deletions(-) > > diff --git a/drivers/media/video/omap3isp/ispqueue.c > b/drivers/media/video/omap3isp/ispqueue.c index 9c31714..4f6876f 100644 > --- a/drivers/media/video/omap3isp/ispqueue.c > +++ b/drivers/media/video/omap3isp/ispqueue.c > @@ -867,6 +867,9 @@ int omap3isp_video_queue_qbuf(struct isp_video_queue > *queue, if (buf->state != ISP_BUF_STATE_IDLE) > goto done; > > + if (vbuf->length < buf->vbuf.length) > + goto done; > + The vbuf->length value passed from userspace isn't used by the driver, so I'm not sure if verifying it is really useful. We verify the memory itself instead, to make sure that enough pages can be accessed. The application can always lie about the length, so we can't relying on it anyway. > if (vbuf->memory == V4L2_MEMORY_USERPTR && > vbuf->m.userptr != buf->vbuf.m.userptr) { > isp_video_buffer_cleanup(buf);
Hi Laurent, On 08/05/2011 10:59 AM, Laurent Pinchart wrote: > > Hi Michael, > > Thanks for the patch. > > On Thursday 04 August 2011 17:40:37 Michael Jones wrote: >> Add buffer length to sanity checks for QBUF. >> >> Signed-off-by: Michael Jones <michael.jones@matrix-vision.de> >> --- >> drivers/media/video/omap3isp/ispqueue.c | 3 +++ >> 1 files changed, 3 insertions(+), 0 deletions(-) >> >> diff --git a/drivers/media/video/omap3isp/ispqueue.c >> b/drivers/media/video/omap3isp/ispqueue.c index 9c31714..4f6876f 100644 >> --- a/drivers/media/video/omap3isp/ispqueue.c >> +++ b/drivers/media/video/omap3isp/ispqueue.c >> @@ -867,6 +867,9 @@ int omap3isp_video_queue_qbuf(struct isp_video_queue >> *queue, if (buf->state != ISP_BUF_STATE_IDLE) >> goto done; >> >> + if (vbuf->length < buf->vbuf.length) >> + goto done; >> + > > The vbuf->length value passed from userspace isn't used by the driver, so I'm > not sure if verifying it is really useful. We verify the memory itself > instead, to make sure that enough pages can be accessed. The application can > always lie about the length, so we can't rely on it anyway. According to the spec, it's expected that the application set 'length': "To enqueue a user pointer buffer applications set [...] length to its size." (Now that I say that, I realize I should only do this length check for USERPTR buffers.) If we don't at least sanity check it for the application, then it has no purpose at all on QBUF. If this is desirable, I would propose changing the spec. This patch was born of a mistake when my application set 624x480, which resulted in sizeimage=640x480=307200 but it used width & height to calculate the buffer size rather than sizeimage or even to take bytesperline into account. It was then honest with QBUF, confessing that it wasn't providing enough space, but QBUF just went ahead. What followed were random crashes while data was DMA'd into memory not set aside for the buffer, while I assumed that the buffer size was OK because QBUF had succeeded and was looking elsewhere in the program for the culprit. I think it makes sense to give the app an error on QBUF in this situation. > >> if (vbuf->memory == V4L2_MEMORY_USERPTR && >> vbuf->m.userptr != buf->vbuf.m.userptr) { >> isp_video_buffer_cleanup(buf); > MATRIX VISION GmbH, Talstrasse 16, DE-71570 Oppenweiler Registergericht: Amtsgericht Stuttgart, HRB 271090 Geschaeftsfuehrer: Gerhard Thullner, Werner Armingeon, Uwe Furtner, Erhard Meier -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Michael, On Friday 05 August 2011 13:41:54 Michael Jones wrote: > On 08/05/2011 10:59 AM, Laurent Pinchart wrote: > > Hi Michael, > > > > Thanks for the patch. > > > > On Thursday 04 August 2011 17:40:37 Michael Jones wrote: > >> Add buffer length to sanity checks for QBUF. > >> > >> Signed-off-by: Michael Jones <michael.jones@matrix-vision.de> > >> --- > >> > >> drivers/media/video/omap3isp/ispqueue.c | 3 +++ > >> 1 files changed, 3 insertions(+), 0 deletions(-) > >> > >> diff --git a/drivers/media/video/omap3isp/ispqueue.c > >> b/drivers/media/video/omap3isp/ispqueue.c index 9c31714..4f6876f 100644 > >> --- a/drivers/media/video/omap3isp/ispqueue.c > >> +++ b/drivers/media/video/omap3isp/ispqueue.c > >> @@ -867,6 +867,9 @@ int omap3isp_video_queue_qbuf(struct isp_video_queue > >> *queue, if (buf->state != ISP_BUF_STATE_IDLE) > >> > >> goto done; > >> > >> + if (vbuf->length < buf->vbuf.length) > >> + goto done; > >> + > > > > The vbuf->length value passed from userspace isn't used by the driver, so > > I'm not sure if verifying it is really useful. We verify the memory > > itself instead, to make sure that enough pages can be accessed. The > > application can always lie about the length, so we can't rely on it > > anyway. > > According to the spec, it's expected that the application set 'length': > "To enqueue a user pointer buffer applications set [...] length to its > size." (Now that I say that, I realize I should only do this length > check for USERPTR buffers.) If we don't at least sanity check it for the > application, then it has no purpose at all on QBUF. If this is > desirable, I would propose changing the spec. > > This patch was born of a mistake when my application set 624x480, which > resulted in sizeimage=640x480=307200 but it used width & height to > calculate the buffer size rather than sizeimage or even to take > bytesperline into account. It was then honest with QBUF, confessing that > it wasn't providing enough space, but QBUF just went ahead. What > followed were random crashes while data was DMA'd into memory not set > aside for the buffer, while I assumed that the buffer size was OK > because QBUF had succeeded and was looking elsewhere in the program for > the culprit. I think it makes sense to give the app an error on QBUF in > this situation. Right. This will help catching application errors without any drawback on the kernel side. Do you want to resubmit the patch with an additional USERPTR check, or should I write one ? > >> if (vbuf->memory == V4L2_MEMORY_USERPTR && > >> vbuf->m.userptr != buf->vbuf.m.userptr) { > >> isp_video_buffer_cleanup(buf);
Hi Laurent, On 08/08/2011 12:08 PM, Laurent Pinchart wrote: > > Hi Michael, > > On Friday 05 August 2011 13:41:54 Michael Jones wrote: >> On 08/05/2011 10:59 AM, Laurent Pinchart wrote: >>> Hi Michael, >>> >>> Thanks for the patch. >>> >>> On Thursday 04 August 2011 17:40:37 Michael Jones wrote: >>>> Add buffer length to sanity checks for QBUF. >>>> >>>> Signed-off-by: Michael Jones <michael.jones@matrix-vision.de> >>>> --- >>>> >>>> drivers/media/video/omap3isp/ispqueue.c | 3 +++ >>>> 1 files changed, 3 insertions(+), 0 deletions(-) >>>> >>>> diff --git a/drivers/media/video/omap3isp/ispqueue.c >>>> b/drivers/media/video/omap3isp/ispqueue.c index 9c31714..4f6876f 100644 >>>> --- a/drivers/media/video/omap3isp/ispqueue.c >>>> +++ b/drivers/media/video/omap3isp/ispqueue.c >>>> @@ -867,6 +867,9 @@ int omap3isp_video_queue_qbuf(struct isp_video_queue >>>> *queue, if (buf->state != ISP_BUF_STATE_IDLE) >>>> >>>> goto done; >>>> >>>> + if (vbuf->length < buf->vbuf.length) >>>> + goto done; >>>> + >>> >>> The vbuf->length value passed from userspace isn't used by the driver, so >>> I'm not sure if verifying it is really useful. We verify the memory >>> itself instead, to make sure that enough pages can be accessed. The >>> application can always lie about the length, so we can't rely on it >>> anyway. >> >> According to the spec, it's expected that the application set 'length': >> "To enqueue a user pointer buffer applications set [...] length to its >> size." (Now that I say that, I realize I should only do this length >> check for USERPTR buffers.) If we don't at least sanity check it for the >> application, then it has no purpose at all on QBUF. If this is >> desirable, I would propose changing the spec. >> >> This patch was born of a mistake when my application set 624x480, which >> resulted in sizeimage=640x480=307200 but it used width & height to >> calculate the buffer size rather than sizeimage or even to take >> bytesperline into account. It was then honest with QBUF, confessing that >> it wasn't providing enough space, but QBUF just went ahead. What >> followed were random crashes while data was DMA'd into memory not set >> aside for the buffer, while I assumed that the buffer size was OK >> because QBUF had succeeded and was looking elsewhere in the program for >> the culprit. I think it makes sense to give the app an error on QBUF in >> this situation. > > Right. This will help catching application errors without any drawback on the > kernel side. > > Do you want to resubmit the patch with an additional USERPTR check, or should > I write one ? Thanks for the review. I will resubmit the patch with the USERPTR check. > >>>> if (vbuf->memory == V4L2_MEMORY_USERPTR && >>>> vbuf->m.userptr != buf->vbuf.m.userptr) { >>>> isp_video_buffer_cleanup(buf); > MATRIX VISION GmbH, Talstrasse 16, DE-71570 Oppenweiler Registergericht: Amtsgericht Stuttgart, HRB 271090 Geschaeftsfuehrer: Gerhard Thullner, Werner Armingeon, Uwe Furtner, Erhard Meier -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/media/video/omap3isp/ispqueue.c b/drivers/media/video/omap3isp/ispqueue.c index 9c31714..4f6876f 100644 --- a/drivers/media/video/omap3isp/ispqueue.c +++ b/drivers/media/video/omap3isp/ispqueue.c @@ -867,6 +867,9 @@ int omap3isp_video_queue_qbuf(struct isp_video_queue *queue, if (buf->state != ISP_BUF_STATE_IDLE) goto done; + if (vbuf->length < buf->vbuf.length) + goto done; + if (vbuf->memory == V4L2_MEMORY_USERPTR && vbuf->m.userptr != buf->vbuf.m.userptr) { isp_video_buffer_cleanup(buf);