| Message ID | 20160129151846.18752.48460.stgit@dwillia2-desk3.amr.corp.intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 01/29/2016 08:18 AM, Dan Williams wrote: > +unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p) > +{ > struct page *page; > > - page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_CACHE_SHIFT-9)), > - NULL); > + /* don't populate page cache for dax capable devices */ > + if (IS_DAX(bdev->bd_inode)) > + page = read_dax_sector(bdev, n); > + else > + page = read_pagecache_sector(bdev, n); > + Fall back to non-dax, if dax fails? > +struct page *read_dax_sector(struct block_device *bdev, sector_t n) > +{ > + struct page *page = __page_cache_alloc(GFP_KERNEL | __GFP_COLD); Why isn't that just alloc_pages()?
On Fri, Jan 29, 2016 at 9:46 AM, Jens Axboe <axboe@fb.com> wrote: > On 01/29/2016 08:18 AM, Dan Williams wrote: >> >> +unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, >> Sector *p) >> +{ >> struct page *page; >> >> - page = read_mapping_page(mapping, (pgoff_t)(n >> >> (PAGE_CACHE_SHIFT-9)), >> - NULL); >> + /* don't populate page cache for dax capable devices */ >> + if (IS_DAX(bdev->bd_inode)) >> + page = read_dax_sector(bdev, n); >> + else >> + page = read_pagecache_sector(bdev, n); >> + > > > Fall back to non-dax, if dax fails? I think we need to fail hard otherwise we're back to the original problem of confusing the dax code that expects to find an empty page cache. > >> +struct page *read_dax_sector(struct block_device *bdev, sector_t n) >> +{ >> + struct page *page = __page_cache_alloc(GFP_KERNEL | __GFP_COLD); > > > Why isn't that just alloc_pages()? Just for symmetry with the same allocation that the pagecache path makes, but alloc_pages() works too... -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Jan 29, 2016 at 07:18:46AM -0800, Dan Williams wrote: > Avoid populating pagecache when the block device is in DAX mode. > Otherwise these page cache entries collide with the fsync/msync > implementation and break data durability guarantees. > > Cc: Jan Kara <jack@suse.com> > Cc: Jeff Moyer <jmoyer@redhat.com> > Cc: Christoph Hellwig <hch@lst.de> > Cc: Dave Chinner <david@fromorbit.com> > Cc: Matthew Wilcox <willy@linux.intel.com> > Cc: Andrew Morton <akpm@linux-foundation.org> > Reported-by: Ross Zwisler <ross.zwisler@linux.intel.com> > Signed-off-by: Dan Williams <dan.j.williams@intel.com> This solves the problem for me, thanks! Tested-by: Ross Zwisler <ross.zwisler@linux.intel.com> -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Jan 29, 2016 at 07:18:46AM -0800, Dan Williams wrote: > diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h > index 29189aeace19..b1452c04f1a9 100644 > --- a/include/linux/blkdev.h > +++ b/include/linux/blkdev.h > @@ -1633,6 +1633,16 @@ struct blk_dax_ctl { > pfn_t pfn; > }; > > +#ifdef CONFIG_FS_DAX > +struct page *read_dax_sector(struct block_device *bdev, sector_t n); > +#else > +static inline struct page *read_dax_sector(struct block_device *bdev, > + sector_t n) > +{ > + return ERR_PTR(-ENXIO); > +} > +#endif > + Can you move this to include/linux/dax.h? I'd like to keep it that all functions in dax.c have a prototype in dax.h. With that change, Reviewed-by: Matthew Wilcox <willy@linux.intel.com> -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/block/partition-generic.c b/block/partition-generic.c index 746935a5973c..8e6fa1868249 100644 --- a/block/partition-generic.c +++ b/block/partition-generic.c @@ -16,6 +16,7 @@ #include <linux/kmod.h> #include <linux/ctype.h> #include <linux/genhd.h> +#include <linux/blkdev.h> #include <linux/blktrace_api.h> #include "partitions/check.h" @@ -550,13 +551,24 @@ int invalidate_partitions(struct gendisk *disk, struct block_device *bdev) return 0; } -unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p) +static struct page *read_pagecache_sector(struct block_device *bdev, sector_t n) { struct address_space *mapping = bdev->bd_inode->i_mapping; + + return read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_CACHE_SHIFT-9)), + NULL); +} + +unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p) +{ struct page *page; - page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_CACHE_SHIFT-9)), - NULL); + /* don't populate page cache for dax capable devices */ + if (IS_DAX(bdev->bd_inode)) + page = read_dax_sector(bdev, n); + else + page = read_pagecache_sector(bdev, n); + if (!IS_ERR(page)) { if (PageError(page)) goto fail; diff --git a/fs/dax.c b/fs/dax.c index 4fd6b0c5c6b5..227974adecb9 100644 --- a/fs/dax.c +++ b/fs/dax.c @@ -58,6 +58,26 @@ static void dax_unmap_atomic(struct block_device *bdev, blk_queue_exit(bdev->bd_queue); } +struct page *read_dax_sector(struct block_device *bdev, sector_t n) +{ + struct page *page = __page_cache_alloc(GFP_KERNEL | __GFP_COLD); + struct blk_dax_ctl dax = { + .size = PAGE_SIZE, + .sector = n & ~((((int) PAGE_SIZE) / 512) - 1), + }; + long rc; + + if (!page) + return ERR_PTR(-ENOMEM); + + rc = dax_map_atomic(bdev, &dax); + if (rc < 0) + return ERR_PTR(rc); + memcpy_from_pmem(page_address(page), dax.addr, PAGE_SIZE); + dax_unmap_atomic(bdev, &dax); + return page; +} + /* * dax_clear_blocks() is called from within transaction context from XFS, * and hence this means the stack from this point must follow GFP_NOFS diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 29189aeace19..b1452c04f1a9 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -1633,6 +1633,16 @@ struct blk_dax_ctl { pfn_t pfn; }; +#ifdef CONFIG_FS_DAX +struct page *read_dax_sector(struct block_device *bdev, sector_t n); +#else +static inline struct page *read_dax_sector(struct block_device *bdev, + sector_t n) +{ + return ERR_PTR(-ENXIO); +} +#endif + struct block_device_operations { int (*open) (struct block_device *, fmode_t); void (*release) (struct gendisk *, fmode_t);
Avoid populating pagecache when the block device is in DAX mode. Otherwise these page cache entries collide with the fsync/msync implementation and break data durability guarantees. Cc: Jan Kara <jack@suse.com> Cc: Jeff Moyer <jmoyer@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Dave Chinner <david@fromorbit.com> Cc: Matthew Wilcox <willy@linux.intel.com> Cc: Andrew Morton <akpm@linux-foundation.org> Reported-by: Ross Zwisler <ross.zwisler@linux.intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> --- block/partition-generic.c | 18 +++++++++++++++--- fs/dax.c | 20 ++++++++++++++++++++ include/linux/blkdev.h | 10 ++++++++++ 3 files changed, 45 insertions(+), 3 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html