| Message ID | 1442943979-1931-1-git-send-email-ross.zwisler@linux.intel.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
On Tue, Sep 22, 2015 at 10:46 AM, Ross Zwisler <ross.zwisler@linux.intel.com> wrote: > The following commit: > > commit 46c043ede471 ("mm: take i_mmap_lock in unmap_mapping_range() for > DAX") > > moved some code in __dax_pmd_fault() that was responsible for zeroing > newly allocated PMD pages. The new location didn't properly set up > 'kaddr', though, so when run this code resulted in a NULL pointer BUG. > > Fix this by getting the correct 'kaddr' via bdev_direct_access(), and > only make the second call to bdev_direct_access() if we don't already > have a PFN from the first call. > > Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> > Reported-by: Dan Williams <dan.j.williams@intel.com> > --- > fs/dax.c | 31 ++++++++++++++++++++++--------- > 1 file changed, 22 insertions(+), 9 deletions(-) > > diff --git a/fs/dax.c b/fs/dax.c > index 7ae6df7..08ac2bd 100644 > --- a/fs/dax.c > +++ b/fs/dax.c > @@ -532,7 +532,7 @@ int __dax_pmd_fault(struct vm_area_struct *vma, unsigned long address, > void __pmem *kaddr; > pgoff_t size, pgoff; > sector_t block, sector; > - unsigned long pfn; > + unsigned long pfn = 0; > int result = 0; > > /* Fall back to PTEs if we're going to COW */ > @@ -569,8 +569,20 @@ int __dax_pmd_fault(struct vm_area_struct *vma, unsigned long address, > if (!buffer_size_valid(&bh) || bh.b_size < PMD_SIZE) > goto fallback; > > + sector = bh.b_blocknr << (blkbits - 9); > + > if (buffer_unwritten(&bh) || buffer_new(&bh)) { > int i; > + > + length = bdev_direct_access(bh.b_bdev, sector, &kaddr, &pfn, > + bh.b_size); > + if (length < 0) { > + result = VM_FAULT_SIGBUS; > + goto out; > + } > + if ((length < PMD_SIZE) || (pfn & PG_PMD_COLOUR)) > + goto fallback; > + > for (i = 0; i < PTRS_PER_PMD; i++) > clear_pmem(kaddr + i * PAGE_SIZE, PAGE_SIZE); > wmb_pmem(); > @@ -623,15 +635,16 @@ int __dax_pmd_fault(struct vm_area_struct *vma, unsigned long address, > result = VM_FAULT_NOPAGE; > spin_unlock(ptl); > } else { > - sector = bh.b_blocknr << (blkbits - 9); > - length = bdev_direct_access(bh.b_bdev, sector, &kaddr, &pfn, > - bh.b_size); > - if (length < 0) { > - result = VM_FAULT_SIGBUS; > - goto out; > + if (pfn == 0) { > + length = bdev_direct_access(bh.b_bdev, sector, &kaddr, > + &pfn, bh.b_size); bdev_direct_access isn't that expensive just call it twice unconditionally. We'll be going that direction anyways when we fix the lifetime of 'kaddr' race in DAX.
diff --git a/fs/dax.c b/fs/dax.c index 7ae6df7..08ac2bd 100644 --- a/fs/dax.c +++ b/fs/dax.c @@ -532,7 +532,7 @@ int __dax_pmd_fault(struct vm_area_struct *vma, unsigned long address, void __pmem *kaddr; pgoff_t size, pgoff; sector_t block, sector; - unsigned long pfn; + unsigned long pfn = 0; int result = 0; /* Fall back to PTEs if we're going to COW */ @@ -569,8 +569,20 @@ int __dax_pmd_fault(struct vm_area_struct *vma, unsigned long address, if (!buffer_size_valid(&bh) || bh.b_size < PMD_SIZE) goto fallback; + sector = bh.b_blocknr << (blkbits - 9); + if (buffer_unwritten(&bh) || buffer_new(&bh)) { int i; + + length = bdev_direct_access(bh.b_bdev, sector, &kaddr, &pfn, + bh.b_size); + if (length < 0) { + result = VM_FAULT_SIGBUS; + goto out; + } + if ((length < PMD_SIZE) || (pfn & PG_PMD_COLOUR)) + goto fallback; + for (i = 0; i < PTRS_PER_PMD; i++) clear_pmem(kaddr + i * PAGE_SIZE, PAGE_SIZE); wmb_pmem(); @@ -623,15 +635,16 @@ int __dax_pmd_fault(struct vm_area_struct *vma, unsigned long address, result = VM_FAULT_NOPAGE; spin_unlock(ptl); } else { - sector = bh.b_blocknr << (blkbits - 9); - length = bdev_direct_access(bh.b_bdev, sector, &kaddr, &pfn, - bh.b_size); - if (length < 0) { - result = VM_FAULT_SIGBUS; - goto out; + if (pfn == 0) { + length = bdev_direct_access(bh.b_bdev, sector, &kaddr, + &pfn, bh.b_size); + if (length < 0) { + result = VM_FAULT_SIGBUS; + goto out; + } + if ((length < PMD_SIZE) || (pfn & PG_PMD_COLOUR)) + goto fallback; } - if ((length < PMD_SIZE) || (pfn & PG_PMD_COLOUR)) - goto fallback; result |= vmf_insert_pfn_pmd(vma, address, pmd, pfn, write); }
The following commit: commit 46c043ede471 ("mm: take i_mmap_lock in unmap_mapping_range() for DAX") moved some code in __dax_pmd_fault() that was responsible for zeroing newly allocated PMD pages. The new location didn't properly set up 'kaddr', though, so when run this code resulted in a NULL pointer BUG. Fix this by getting the correct 'kaddr' via bdev_direct_access(), and only make the second call to bdev_direct_access() if we don't already have a PFN from the first call. Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com> Reported-by: Dan Williams <dan.j.williams@intel.com> --- fs/dax.c | 31 ++++++++++++++++++++++--------- 1 file changed, 22 insertions(+), 9 deletions(-)