From mboxrd@z Thu Jan 1 00:00:00 1970 From: Varun Sethi Subject: RE: [PATCH v2 01/26] iommu/fsl: Sprinkle some __init* annotations Date: Wed, 11 Feb 2015 13:47:29 +0000 Message-ID: References: <1422455698-3074-1-git-send-email-Emilian.Medve@Freescale.com> <1422455698-3074-2-git-send-email-Emilian.Medve@Freescale.com> <54D92ABC.3030802@Freescale.com> <54DB37DA.2040500@Freescale.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Content-Language: en-US List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: iommu-bounces-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org Errors-To: iommu-bounces-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org To: Emilian Medve , "iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org" , "joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org" , "jroedel-l3A5Bk7waGM@public.gmane.org" List-Id: iommu@lists.linux-foundation.org Hi Emil, > -----Original Message----- > From: Sethi Varun-B16395 > Sent: Wednesday, February 11, 2015 5:00 PM > To: 'Emil Medve'; iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org; joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org; > jroedel-l3A5Bk7waGM@public.gmane.org > Subject: RE: [PATCH v2 01/26] iommu/fsl: Sprinkle some __init* annotations > > Hi Emil, > > > -----Original Message----- > > From: Emil Medve [mailto:Emilian.Medve-eDlz3WWmN0ll57MIdRCFDg@public.gmane.org] > > Sent: Wednesday, February 11, 2015 4:37 PM > > To: Sethi Varun-B16395; iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org; > > joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org; jroedel-l3A5Bk7waGM@public.gmane.org > > Subject: Re: [PATCH v2 01/26] iommu/fsl: Sprinkle some __init* > > annotations > > > > Hello Varun, > > > > > > On 02/09/2015 08:26 PM, Sethi Varun-B16395 wrote: > > > Hi Emil, > > > > > > > > >> -----Original Message----- > > >> From: Emil Medve [mailto:Emilian.Medve-eDlz3WWmN0ll57MIdRCFDg@public.gmane.org] > > >> Sent: Tuesday, February 10, 2015 3:17 AM > > >> To: iommu-cunTk1MwBs9QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org; joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org; > > >> jroedel-l3A5Bk7waGM@public.gmane.org; Sethi Varun-B16395 > > >> Subject: Re: [PATCH v2 01/26] iommu/fsl: Sprinkle some __init* > > >> annotations > > >> > > >> Hello Joerg, > > >> > > >> > > >> On 01/28/2015 08:34 AM, Emil Medve wrote: > > >>> Signed-off-by: Emil Medve > > >>> --- > > >>> drivers/iommu/fsl_pamu.c | 12 ++++++------ > > >>> drivers/iommu/fsl_pamu_domain.c | 2 +- > > >>> 2 files changed, 7 insertions(+), 7 deletions(-) > > >> > > >> Please don't apply this patch as it's wrong. It's based on > > >> https://git.kernel.org/cgit/linux/kernel/git/joro/iommu.git/commit/ > > >> ?h > > >> =ppc/ > > >> pamu&id=0f1fb99b62ce226f8d818852f812c5d79071ce58 > > >> which is wrong as well and its causing semi-random crashes upon > > >> more extensive testing. I will follow-up with details shortly > > > > > > Are you seeing a crash with > > > https://git.kernel.org/cgit/linux/kernel/git/joro/iommu.git/commit/? > > > h= > > > ppc/pamu&id=0f1fb99b62ce226f8d818852f812c5d79071ce58 > > > ? At, what point do you see crashes? > > > > platform_driver.device_driver gets added to a (platform) bus klist of > drivers. > > (There are also a handful of other pointers that get set to > > fsl_of_pamu_driver) Once fsl_of_pamu_driver gets released that klist > > becomes "corrupted" and, depending on how said memory gets re-used, > > traversing it for any reason will end up into a crash > > > > Anyway, I identified easier ways to produce kernel crashes related to > > the PAMU driver via sysfs (with or without my __init* annotation > > patches). For example a "unbind, bind" sequence will cause a crash > > with the driver as is in > > v3.19 > > > > BTW, why is the PAMU an "early" driver? > > > > I'll send out some patches to properly fix all these __init* > > annotations > Yes, the fsl_of_pamu_driver structure shouldn't be marked init data, it could > actually corrupt the driver list. PAMU driver should initialize before the DPAA > driver initialization also, to ensure the default DMA window is setup for all > devices. > I think that we can revert this patch, the section mismatch isn't really an issue. The pamu driver probe routine wouldn't get called during platform bus probe. -Varun