From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Nelson Date: Fri, 20 Jun 2014 20:44:29 -0700 Subject: [Buildroot] [PATCH 1/5] freescale-imx: bump to version 3.10.17-1.0.0 In-Reply-To: <20140620210146.GB26745@free.fr> References: <1403150639-29382-1-git-send-email-bisson.gary@gmail.com> <1403150639-29382-2-git-send-email-bisson.gary@gmail.com> <20140619162009.GA3534@free.fr> <20140620210146.GB26745@free.fr> Message-ID: <53A4FF9D.2020909@boundarydevices.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hi Yann and Gary, On 06/20/2014 02:01 PM, Yann E. MORIN wrote: > Gary, All, > > On 2014-06-19 19:40 -0700, Gary Bisson spake thusly: >>>> diff --git >>> a/package/freescale-imx/imx-vpu/imx-vpu-0001-fix-IOSystemInit-failure.patch >>> b/package/freescale-imx/imx-vpu/imx-vpu-0001-fix-IOSystemInit-failure.patch >>>> new file mode 100644 >>>> index 0000000..b73a959 >>>> --- /dev/null >>>> +++ >>> b/package/freescale-imx/imx-vpu/imx-vpu-0001-fix-IOSystemInit-failure.patch >>>> @@ -0,0 +1,21 @@ >>>> +[PATCH] fix IOSystemInit failure >>> >>> Please provide a more detailed comit log. Why do we need this? >>> >> >> As I said in the first patch I used the nitrogen6x to do my testing, there >> is one "big" difference between freescale and boundary kernel which is the >> memory split (2G/2G for FSL, 3G/1G for Boundary). Clearly the VPU code has >> only been tested on Freescale kernels because there is no way it works on >> kernel with the latter splitting. The VPU init function returns an error if >> the allocated buffer address is < 0 which does not make any sense. Instead >> the test should be against MAP_FAILED (-1). Only after that patch I've been >> able to get the VPU decoding working. I will change the -1 into MAP_FAILED >> for the next patch. Nice catch. I do think this should be split into a separate patch though. A quick scan of this code shows that there are four other places that have the test for IOGetVirtMem() <= 0. > > OK, I see. You should also repeat this in the patch description, like: > > vpu-io: fix IOSystemInit failure > > There is one "big" difference between freescale and boundary kernel > which is the memory split (2G/2G for FSL, 3G/1G for Boundary). > > Clearly the VPU code has only been tested on Freescale kernels > because there is no way it works on kernel with the latter > splitting. The VPU init function returns an error if the allocated > buffer address is < 0, which happens with the 3G/1G split. > > Fix the test by checking against MAP_FAILED instead. > > Signed-off-by: You > > Thanks! :-) > > Regards, > Yann E. MORIN. >