MIPS32Â 34Kâ Core
It does not provide invalidate_kernel_vmap_range() /
flush_kernel_vmap_range() to deal with cache coherency problem
Therefore we provided dma_cache_inv() / dma_cache_wback_inv() , after
that we did not had any coherency problem
But as you saying it still has cache coherency problem, wrt xfstests 001 ?
diff -Nurp -X linux-2.6.35.11/Documentation/dontdiff
linux-2.6.35.11/fs/xfs/linux-2.6/xfs_buf.c
linux-2.6.35.11-dirty/fs/xfs/linux-2.6/xfs_buf.c
--- linux-2.6.35.11/fs/xfs/linux-2.6/xfs_buf.c 2011-02-07
04:04:07.000000000 +0900
+++ linux-2.6.35.11-dirty/fs/xfs/linux-2.6/xfs_buf.c 2011-03-22
18:29:09.000000000 +0900
@@ -1192,7 +1192,7 @@ xfs_buf_bio_end_io(
xfs_buf_ioerror(bp, -error);
if (!error && xfs_buf_is_vmapped(bp) && (bp->b_flags & XBF_READ))
- invalidate_kernel_vmap_range(bp->b_addr, xfs_buf_vmap_len(bp));
+ dma_cache_inv((unsigned long)bp->b_addr, xfs_buf_vmap_len(bp));
do {
struct page *page = bvec->bv_page;
@@ -1304,7 +1304,7 @@ next_chunk:
submit_io:
if (likely(bio->bi_size)) {
if (xfs_buf_is_vmapped(bp)) {
- flush_kernel_vmap_range(bp->b_addr,
+ dma_cache_wback_inv((unsigned long)bp->b_addr,
xfs_buf_vmap_len(bp));
}
submit_bio(rw, bio);
On Wed, Apr 27, 2011 at 10:41 PM, Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote:
> Does 001 also fail for say ext2 and ext3?
>
>
_______________________________________________
xfs mailing list
xfs@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/xfs
