Microblaze noMMU break - mm/nommu.h
Subsystems (2)
A message that is not in the archive (sent off-list or not yet received): [email protected]
Hi Michal,
Quoted 5 lines · Michal Simek
Your commit
58fa879e1e640a1856f736b418984ebeccee1c95break noMMU Microblaze kernel.should contain different change in
mm/internal.h. - patch is below
I would swear that I testbuilt that for nommu, but clearly I did not:
one fix to nommu.c in the patch below...
Quoted 1 line
Please fix it. If someone reported it, please ignore my email
Yours is the first I've heard: thanks a lot for your quick report, and my apologies for the nuisance - I seem to have fallen into a nasty habit of breaking one config or another with my patches.
Quoted 4 lines
Your commit
03f6462a3ae78f36eb1f0ee8b4d5ae2f7859c1d5moveshighest_memmap_pfnfrompage_alloc.ctomemory.cbutmemory.cis not compiled for noMMU kernels - which caused fault too.
And that one I just didn't think of nommu, more's the shame:
another fix to nommu.c in the patch below...
Quoted 2 lines
I haven't seen that patches in linux-next which I test too. :-(
Thank you for testing linux-next, but these kind of inadvertent breaks to CONFIG_NOMMU are probably most often caused by changes coming in via Andrew's mmotm - where most (liable to break nomm) mm development occurs.
linux-next was started really as a service to make Andrew's life with mmotm slightly easier, but now, yes, he would like to move over to getting his mm series into linux-next... but it's not a trivial task to turn that around.
If it's possible for you to test mmotms, your input on breakages would be much appreciated; but I can quite understand that linux-next is enough work for you, and that I appear to be bleating at you to do something different, when it's me who needs to learn to be more careful and build more configs.
The patch below fixes both for me: if you're able to check it on Microblaze now, please let us know if they're good for you; but if not, I'll send direct to Linus in a few hours time, better to have as narrow a gap of build breakage there as we can.
Thanks, and sorry,
Hugh
[PATCH] nommu: fix two build breakages
My 58fa879e1e640a1856f736b418984ebeccee1c95 "mm: FOLL flags for GUP flags" broke CONFIG_NOMMU build by forgetting to update nommu.c foll_flags type:mm/nommu.c:171: error: conflicting types for __get_user_pages' mm/internal.h:254: error: previous declaration of __get_user_pages' was here make[1]: *** [mm/nommu.o] Error 1
My 03f6462a3ae78f36eb1f0ee8b4d5ae2f7859c1d5 "mm: move highest_memmap_pfn" broke CONFIG_NOMMU build by forgetting to add a nommu.c highest_memmap_pfn:mm/built-in.o: In function `memmap_init_zone':
(.meminit.text+0x326): undefined reference to highest_memmap_pfn' mm/built-in.o: In function memmap_init_zone':
(.meminit.text+0x32d): undefined reference to `highest_memmap_pfn'
Fix both breakages, and give myself 30 lashes (ouch!)
---
mm/nommu.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
mm/nommu.c +2 -1
@@ -61,6 +61,7 @@ void *high_memory;struct page *mem_map;unsigned long max_mapnr;unsigned long num_physpages; +unsigned long highest_memmap_pfn;struct percpu_counter vm_committed_as;int sysctl_overcommit_memory =OVERCOMMIT_GUESS ;/* heuristic overcommit */ int sysctl_overcommit_ratio =50 ;/* default is 50% */
@@ -169,7 +170,7 @@ unsigned int kobjsize(const void *objp) }int __get_user_pages (struct task_struct *tsk,struct mm_struct *mm, -unsigned long start,int nr_pages,int foll_flags, +unsigned long start,int nr_pages,unsigned int foll_flags,struct page **pages,struct vm_area_struct **vmas) {struct vm_area_struct *vma;
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 23 lines · Michal Simek
2009/9/23 Hugh Dickins <[email protected]>
Quoted 6 lines
Thank you for testing linux-next, but these kind of inadvertent breaks to
CONFIG_NOMMUare probably most often caused by changes coming in via Andrew's mmotm - where most (liable to break nomm) mm development occurs.you are welcome.
Quoted 6 lines
linux-next was started really as a service to make Andrew's life with mmotm slightly easier, but now, yes, he would like to move over to getting his mm series into linux-next... but it's not a trivial task to turn that around.
Quoted 6 lines
If it's possible for you to test mmotms, your input on breakages would be much appreciated; but I can quite understand that linux-next is enough work for you, and that I appear to be bleating at you to do something different, when it's me who needs to learn to be more careful and build more configs.
Quoted 3 lines
I can easily test any git tree - just give me link. + info how is this tree maintained
will be helpful too. (If is used rebase, if yes - on which tree, etc)
Thank you.
git://git.zen-sources.org/zen/mmotm.git
At first I was going to say that mmotm is maintained as a quilt tree, a tarfile of patches against the last -rcN (including one origin.patch of the changes in Linus's tree, and the linux-next.patch from that), which you would apply using quilt.
But you're git-oriented, so wouldn't be wanting to spend time on that:
it is also made available nowadays as a git tree, which should suit you much better - but that is constructed from the original quilt patches, and I don't actually use it myself, so can't guarantee how well it works.
Here's what Andrew says in one of his mm-commits and lkml announcements (they're irregular, depending on many factors, principally Andrew's
load: roughly one or perhaps two per week). Sometimes an mmotm comes out just after Linus releases a new -rcN, but the mmotm would then be based on the previous -rcN - since it needs to include the prior linux-next, and be tested to some extent.
Quote:
The mm-of-the-moment snapshot 2009-09-17-18-00 has been uploaded to
http://userweb.kernel.org/~akpm/mmotm/
and will soon be available at
git://git.zen-sources.org/zen/mmotm.git
It contains the following patches against 2.6.31:
origin.patch # recent diff from -rcN to Linus -git
[urgent fix patches]
linux-next.patch
linux-next-git-rejects.patch
next-remove-localversion.patch
[misc patches akpm is holding]
mm.patch # which appends -mm1 to the -rcN
[mm and other patches ]
(when I write of an -rcN, well, at present it's just 2.6.31 itself, no "-rcN" at all.)
Quoted 2 lines
I tested it and works.
Thanks for reporting back: Linus was paying attention to the thread, and put the fix in directly without any more prompting.
Quoted 1 line
There is one new issue about pagewalk in different patch. :-(
Our sympathies.
Hugh