Re: [PATCH] mm: kill kmemcheck
Hi Sasha, On 3/11/15 1:43 PM, Sasha Levin wrote: As discussed on LSF/MM, kill kmemcheck. KASan is a replacement that is able to work without the limitation of kmemcheck (single CPU, slow). KASan is already upstream. We are also not aware of any users of kmemcheck (or users who don't consider KASan as a suitable replacement). I've build tested it using all[yes,no,mod]config and fuzzed a bit with this patch applied, didn't notice any bad behaviour. Signed-off-by: Sasha Levin sasha.le...@oracle.com Can you elaborate on what exactly was discussed at LSF/MM? Preferably in the commit log. ;-) - Pekka ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH] mm: kill kmemcheck
On Thu, 12 Mar 2015 14:49:52 +1100 Michael Ellerman m...@ellerman.id.au wrote: From Documentation/kasan.txt: therefore you will need a certain version of GCC 4.9.2 AFAIK gcc 4.9.3 hasn't been released yet. (Or does it mean = 4.9.2 ?) It means 4.9.2. I also feel that this makes it too premature to remove kmemcheck, as not everyone (including myself) has the latest and greatest gcc on their systems. Can we perhaps wait until there is a released version of GCC that supports KASan? And maybe then a touch longer so folks can test it works on their platforms? Note, I replied to this cutting out the Cc's to the individuals so that it can have a wider audience! (figure out what I mean by that). You can see my reply on LKML (hint, you wont see the original email, or this current thread for that matter). -- Steve ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH] mm: kill kmemcheck
On 03/12/2015 03:07 AM, Pekka Enberg wrote: Hi Sasha, On 3/11/15 1:43 PM, Sasha Levin wrote: As discussed on LSF/MM, kill kmemcheck. KASan is a replacement that is able to work without the limitation of kmemcheck (single CPU, slow). KASan is already upstream. We are also not aware of any users of kmemcheck (or users who don't consider KASan as a suitable replacement). I've build tested it using all[yes,no,mod]config and fuzzed a bit with this patch applied, didn't notice any bad behaviour. Signed-off-by: Sasha Levin sasha.le...@oracle.com Can you elaborate on what exactly was discussed at LSF/MM? Preferably in the commit log. ;-) There wasn't a long discussion about removing kmemcheck, it just followed up a KASan topic and the question of whether kmemcheck can be deprecated now that KASan is merged came up. No one at the room didn't use it, knew a user of it, or couldn't describe a usecase where kmemcheck was superior to KASan - so the conclusion was to try and replace it. The only thing I can really put in the changelog is a reference to the KASan docs and to ask folks to complain loudly if we missed a usecase. Thanks, Sasha ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH] mm: kill kmemcheck
On 03/11/2015 11:49 PM, Michael Ellerman wrote: On Wed, 2015-03-11 at 07:43 -0400, Sasha Levin wrote: As discussed on LSF/MM, kill kmemcheck. KASan is a replacement that is able to work without the limitation of kmemcheck (single CPU, slow). KASan is already upstream. We are also not aware of any users of kmemcheck (or users who don't consider KASan as a suitable replacement). From Documentation/kasan.txt: therefore you will need a certain version of GCC 4.9.2 AFAIK gcc 4.9.3 hasn't been released yet. (Or does it mean = 4.9.2 ?) Can we perhaps wait until there is a released version of GCC that supports KASan? And maybe then a touch longer so folks can test it works on their platforms? I think this is just an off-by-one in the documentation. The cover letter for the KASan patchset stated: KASAN uses compile-time instrumentation for checking every memory access, therefore you will need a fresh GCC = v4.9.2 Thanks, Sasha ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH] mm: kill kmemcheck
Sasha Levin sasha.le...@oracle.com writes: As discussed on LSF/MM, kill kmemcheck. Damn: I literally added CONFIG_KMEMCHECK support to virtio yesterday! Will try Kasan now. Thanks for the heads-up, Rusty. ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev
Re: [PATCH] mm: kill kmemcheck
On Wed, 2015-03-11 at 07:43 -0400, Sasha Levin wrote: As discussed on LSF/MM, kill kmemcheck. KASan is a replacement that is able to work without the limitation of kmemcheck (single CPU, slow). KASan is already upstream. We are also not aware of any users of kmemcheck (or users who don't consider KASan as a suitable replacement). From Documentation/kasan.txt: therefore you will need a certain version of GCC 4.9.2 AFAIK gcc 4.9.3 hasn't been released yet. (Or does it mean = 4.9.2 ?) Can we perhaps wait until there is a released version of GCC that supports KASan? And maybe then a touch longer so folks can test it works on their platforms? cheers ___ Linuxppc-dev mailing list Linuxppc-dev@lists.ozlabs.org https://lists.ozlabs.org/listinfo/linuxppc-dev