Commit b1adaf65ba03([SCSI] block: add sg buffer copy helper functions)
introduces two sg buffer copy helpers, and calls flush_kernel_dcache_page()
on pages in SG list after these pages are written to.
Unfortunately, the commit may introduce a potential bug:
- Before sending some SCSI commands, kmalloc() buffer may be
passed to block layper, so flush_kernel_dcache_page() can
see a slab page finally
- According to cachetlb.txt, flush_kernel_dcache_page() is
only called on "a user page", which surely can't be a slab page.
- ARCH's implementation of flush_kernel_dcache_page() may
use page mapping information to do optimization so page_mapping()
will see the slab page, then VM_BUG_ON() is triggered.
Aaro Koskinen reported the bug on ARM/kirkwood when DEBUG_VM is enabled,
and this patch fixes the bug by adding test of '!PageSlab(miter->page)'
before calling flush_kernel_dcache_page().
Reported-by: Aaro Koskinen <[email protected]>
Cc: Russell King - ARM Linux <[email protected]>
Cc: [email protected]
Cc: Simon Baatz <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Aaro Koskinen <[email protected]>
Cc: Catalin Marinas <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: FUJITA Tomonori <[email protected]>
Cc: Tejun Heo <[email protected]>
Cc: "James E.J. Bottomley" <[email protected]>
Cc: Jens Axboe <[email protected]>
Signed-off-by: Ming Lei <[email protected]>
---
lib/scatterlist.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/lib/scatterlist.c b/lib/scatterlist.c
index a685c8a..d16fa29 100644
--- a/lib/scatterlist.c
+++ b/lib/scatterlist.c
@@ -577,7 +577,8 @@ void sg_miter_stop(struct sg_mapping_iter *miter)
miter->__offset += miter->consumed;
miter->__remaining -= miter->consumed;
- if (miter->__flags & SG_MITER_TO_SG)
+ if ((miter->__flags & SG_MITER_TO_SG) &&
+ !PageSlab(miter->page))
flush_kernel_dcache_page(miter->page);
if (miter->__flags & SG_MITER_ATOMIC) {
--
1.7.9.5
On Mon, Oct 28, 2013 at 04:20:05PM +0000, Ming Lei wrote:
> Commit b1adaf65ba03([SCSI] block: add sg buffer copy helper functions)
> introduces two sg buffer copy helpers, and calls flush_kernel_dcache_page()
> on pages in SG list after these pages are written to.
>
> Unfortunately, the commit may introduce a potential bug:
>
> - Before sending some SCSI commands, kmalloc() buffer may be
> passed to block layper, so flush_kernel_dcache_page() can
> see a slab page finally
>
> - According to cachetlb.txt, flush_kernel_dcache_page() is
> only called on "a user page", which surely can't be a slab page.
>
> - ARCH's implementation of flush_kernel_dcache_page() may
> use page mapping information to do optimization so page_mapping()
> will see the slab page, then VM_BUG_ON() is triggered.
>
> Aaro Koskinen reported the bug on ARM/kirkwood when DEBUG_VM is enabled,
> and this patch fixes the bug by adding test of '!PageSlab(miter->page)'
> before calling flush_kernel_dcache_page().
>
> Reported-by: Aaro Koskinen <[email protected]>
> Cc: Russell King - ARM Linux <[email protected]>
> Cc: [email protected]
> Cc: Simon Baatz <[email protected]>
> Cc: Will Deacon <[email protected]>
> Cc: Aaro Koskinen <[email protected]>
> Cc: Catalin Marinas <[email protected]>
> Cc: Andrew Morton <[email protected]>
> Cc: FUJITA Tomonori <[email protected]>
> Cc: Tejun Heo <[email protected]>
> Cc: "James E.J. Bottomley" <[email protected]>
> Cc: Jens Axboe <[email protected]>
> Signed-off-by: Ming Lei <[email protected]>
Acked-by: Catalin Marinas <[email protected]>
On Tue, Oct 29, 2013 at 12:20:05AM +0800, Ming Lei wrote:
> Commit b1adaf65ba03([SCSI] block: add sg buffer copy helper functions)
> introduces two sg buffer copy helpers, and calls flush_kernel_dcache_page()
> on pages in SG list after these pages are written to.
>
> Unfortunately, the commit may introduce a potential bug:
>
> - Before sending some SCSI commands, kmalloc() buffer may be
> passed to block layper, so flush_kernel_dcache_page() can
> see a slab page finally
>
> - According to cachetlb.txt, flush_kernel_dcache_page() is
> only called on "a user page", which surely can't be a slab page.
>
> - ARCH's implementation of flush_kernel_dcache_page() may
> use page mapping information to do optimization so page_mapping()
> will see the slab page, then VM_BUG_ON() is triggered.
>
> Aaro Koskinen reported the bug on ARM/kirkwood when DEBUG_VM is enabled,
> and this patch fixes the bug by adding test of '!PageSlab(miter->page)'
> before calling flush_kernel_dcache_page().
>
> Reported-by: Aaro Koskinen <[email protected]>
> Cc: Russell King - ARM Linux <[email protected]>
> Cc: [email protected]
> Cc: Simon Baatz <[email protected]>
> Cc: Will Deacon <[email protected]>
> Cc: Aaro Koskinen <[email protected]>
> Cc: Catalin Marinas <[email protected]>
> Cc: Andrew Morton <[email protected]>
> Cc: FUJITA Tomonori <[email protected]>
> Cc: Tejun Heo <[email protected]>
> Cc: "James E.J. Bottomley" <[email protected]>
> Cc: Jens Axboe <[email protected]>
> Signed-off-by: Ming Lei <[email protected]>
> ---
> lib/scatterlist.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/lib/scatterlist.c b/lib/scatterlist.c
> index a685c8a..d16fa29 100644
> --- a/lib/scatterlist.c
> +++ b/lib/scatterlist.c
> @@ -577,7 +577,8 @@ void sg_miter_stop(struct sg_mapping_iter *miter)
> miter->__offset += miter->consumed;
> miter->__remaining -= miter->consumed;
>
> - if (miter->__flags & SG_MITER_TO_SG)
> + if ((miter->__flags & SG_MITER_TO_SG) &&
> + !PageSlab(miter->page))
> flush_kernel_dcache_page(miter->page);
>
> if (miter->__flags & SG_MITER_ATOMIC) {
> --
> 1.7.9.5
>
Tested-by: Simon Baatz <[email protected]>
on top of 3.12-rc7 on Kirkwood with DEBUG_VM enabled.
- Simon
On Tue, 29 Oct 2013 00:20:05 +0800 Ming Lei <[email protected]> wrote:
> Commit b1adaf65ba03([SCSI] block: add sg buffer copy helper functions)
> introduces two sg buffer copy helpers, and calls flush_kernel_dcache_page()
> on pages in SG list after these pages are written to.
>
> Unfortunately, the commit may introduce a potential bug:
>
> - Before sending some SCSI commands, kmalloc() buffer may be
> passed to block layper, so flush_kernel_dcache_page() can
> see a slab page finally
>
> - According to cachetlb.txt, flush_kernel_dcache_page() is
> only called on "a user page", which surely can't be a slab page.
>
> - ARCH's implementation of flush_kernel_dcache_page() may
> use page mapping information to do optimization so page_mapping()
> will see the slab page, then VM_BUG_ON() is triggered.
>
> Aaro Koskinen reported the bug on ARM/kirkwood when DEBUG_VM is enabled,
> and this patch fixes the bug by adding test of '!PageSlab(miter->page)'
> before calling flush_kernel_dcache_page().
We should work out which kernel(s) need this patch. b1adaf65ba03 was
merged in 2008, so presumably some more recent patch has exposed the
problem, but I don't know what one that was.
Help me out here?
On Thu, Oct 31, 2013 at 03:27:45PM -0700, Andrew Morton wrote:
> On Tue, 29 Oct 2013 00:20:05 +0800 Ming Lei <[email protected]> wrote:
>
> > Commit b1adaf65ba03([SCSI] block: add sg buffer copy helper functions)
> > introduces two sg buffer copy helpers, and calls flush_kernel_dcache_page()
> > on pages in SG list after these pages are written to.
> >
> > Unfortunately, the commit may introduce a potential bug:
> >
> > - Before sending some SCSI commands, kmalloc() buffer may be
> > passed to block layper, so flush_kernel_dcache_page() can
> > see a slab page finally
> >
> > - According to cachetlb.txt, flush_kernel_dcache_page() is
> > only called on "a user page", which surely can't be a slab page.
> >
> > - ARCH's implementation of flush_kernel_dcache_page() may
> > use page mapping information to do optimization so page_mapping()
> > will see the slab page, then VM_BUG_ON() is triggered.
> >
> > Aaro Koskinen reported the bug on ARM/kirkwood when DEBUG_VM is enabled,
> > and this patch fixes the bug by adding test of '!PageSlab(miter->page)'
> > before calling flush_kernel_dcache_page().
>
> We should work out which kernel(s) need this patch. b1adaf65ba03 was
> merged in 2008, so presumably some more recent patch has exposed the
> problem, but I don't know what one that was.
>
> Help me out here?
1bc39742aab0 (ARM: 7755/1: handle user space mapped pages in
flush_kernel_dcache_page) in June 2013, which according to git describe
is:
v3.10-rc4-9-g1bc39742aab0
Presumably that means v3.10+ need it.
Hi,
On Thu, Oct 31, 2013 at 03:27:45PM -0700, Andrew Morton wrote:
> On Tue, 29 Oct 2013 00:20:05 +0800 Ming Lei <[email protected]> wrote:
>
> > Commit b1adaf65ba03([SCSI] block: add sg buffer copy helper functions)
> > introduces two sg buffer copy helpers, and calls flush_kernel_dcache_page()
> > on pages in SG list after these pages are written to.
> >
> > Unfortunately, the commit may introduce a potential bug:
> >
> > - Before sending some SCSI commands, kmalloc() buffer may be
> > passed to block layper, so flush_kernel_dcache_page() can
> > see a slab page finally
> >
> > - According to cachetlb.txt, flush_kernel_dcache_page() is
> > only called on "a user page", which surely can't be a slab page.
> >
> > - ARCH's implementation of flush_kernel_dcache_page() may
> > use page mapping information to do optimization so page_mapping()
> > will see the slab page, then VM_BUG_ON() is triggered.
> >
> > Aaro Koskinen reported the bug on ARM/kirkwood when DEBUG_VM is enabled,
> > and this patch fixes the bug by adding test of '!PageSlab(miter->page)'
> > before calling flush_kernel_dcache_page().
>
> We should work out which kernel(s) need this patch. b1adaf65ba03 was
> merged in 2008, so presumably some more recent patch has exposed the
> problem, but I don't know what one that was.
>
> Help me out here?
On ARM, this problem started to appear after
1bc39742aab09248169ef9d3727c9def3528b3f3 (ARM: 7755/1: handle user space
mapped pages in flush_kernel_dcache_page) and that was tagged for 3.2+
stable kernels.
A.
On Fri, 1 Nov 2013 00:49:27 +0200 Aaro Koskinen <[email protected]> wrote:
> Hi,
>
> On Thu, Oct 31, 2013 at 03:27:45PM -0700, Andrew Morton wrote:
> > On Tue, 29 Oct 2013 00:20:05 +0800 Ming Lei <[email protected]> wrote:
> >
> > > Commit b1adaf65ba03([SCSI] block: add sg buffer copy helper functions)
> > > introduces two sg buffer copy helpers, and calls flush_kernel_dcache_page()
> > > on pages in SG list after these pages are written to.
> > >
> > > Unfortunately, the commit may introduce a potential bug:
> > >
> > > - Before sending some SCSI commands, kmalloc() buffer may be
> > > passed to block layper, so flush_kernel_dcache_page() can
> > > see a slab page finally
> > >
> > > - According to cachetlb.txt, flush_kernel_dcache_page() is
> > > only called on "a user page", which surely can't be a slab page.
> > >
> > > - ARCH's implementation of flush_kernel_dcache_page() may
> > > use page mapping information to do optimization so page_mapping()
> > > will see the slab page, then VM_BUG_ON() is triggered.
> > >
> > > Aaro Koskinen reported the bug on ARM/kirkwood when DEBUG_VM is enabled,
> > > and this patch fixes the bug by adding test of '!PageSlab(miter->page)'
> > > before calling flush_kernel_dcache_page().
> >
> > We should work out which kernel(s) need this patch. b1adaf65ba03 was
> > merged in 2008, so presumably some more recent patch has exposed the
> > problem, but I don't know what one that was.
> >
> > Help me out here?
>
> On ARM, this problem started to appear after
> 1bc39742aab09248169ef9d3727c9def3528b3f3 (ARM: 7755/1: handle user space
> mapped pages in flush_kernel_dcache_page) and that was tagged for 3.2+
> stable kernels.
>
OK, thanks. I added
Cc: <[email protected]> [3.2+]
to the changelog.
I guess I should get this and a few other things into Linus today, as
3.12 is nigh.