From 0378da0fda6edf5aaffda6f1248a78986bd955b5 Mon Sep 17 00:00:00 2001 From: Tao Ma Date: Thu, 12 Aug 2010 10:25:28 +0800 Subject: ocfs2: pass struct file* to ocfs2_write_begin_nolock. struct file * has file_ra_state to store the readahead state and data. So pass this to ocfs2_write_begin_nolock so that it can be used in ocfs2_refcount_cow. Signed-off-by: Tao Ma diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 0de69c9..e3efdd9 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -1642,7 +1642,8 @@ static int ocfs2_zero_tail(struct inode *inode, struct buffer_head *di_bh, return ret; } -int ocfs2_write_begin_nolock(struct address_space *mapping, +int ocfs2_write_begin_nolock(struct file *filp, + struct address_space *mapping, loff_t pos, unsigned len, unsigned flags, struct page **pagep, void **fsdata, struct buffer_head *di_bh, struct page *mmap_page) @@ -1854,7 +1855,7 @@ static int ocfs2_write_begin(struct file *file, struct address_space *mapping, */ down_write(&OCFS2_I(inode)->ip_alloc_sem); - ret = ocfs2_write_begin_nolock(mapping, pos, len, flags, pagep, + ret = ocfs2_write_begin_nolock(file, mapping, pos, len, flags, pagep, fsdata, di_bh, NULL); if (ret) { mlog_errno(ret); diff --git a/fs/ocfs2/aops.h b/fs/ocfs2/aops.h index c48e93f..7606f66 100644 --- a/fs/ocfs2/aops.h +++ b/fs/ocfs2/aops.h @@ -48,7 +48,8 @@ int ocfs2_write_end_nolock(struct address_space *mapping, loff_t pos, unsigned len, unsigned copied, struct page *page, void *fsdata); -int ocfs2_write_begin_nolock(struct address_space *mapping, +int ocfs2_write_begin_nolock(struct file *filp, + struct address_space *mapping, loff_t pos, unsigned len, unsigned flags, struct page **pagep, void **fsdata, struct buffer_head *di_bh, struct page *mmap_page); diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c index af2b8fe..b04d696 100644 --- a/fs/ocfs2/mmap.c +++ b/fs/ocfs2/mmap.c @@ -59,10 +59,11 @@ static int ocfs2_fault(struct vm_area_struct *area, struct vm_fault *vmf) return ret; } -static int __ocfs2_page_mkwrite(struct inode *inode, struct buffer_head *di_bh, +static int __ocfs2_page_mkwrite(struct file *file, struct buffer_head *di_bh, struct page *page) { int ret; + struct inode *inode = file->f_path.dentry->d_inode; struct address_space *mapping = inode->i_mapping; loff_t pos = page_offset(page); unsigned int len = PAGE_CACHE_SIZE; @@ -109,7 +110,7 @@ static int __ocfs2_page_mkwrite(struct inode *inode, struct buffer_head *di_bh, if (page->index == last_index) len = size & ~PAGE_CACHE_MASK; - ret = ocfs2_write_begin_nolock(mapping, pos, len, 0, &locked_page, + ret = ocfs2_write_begin_nolock(file, mapping, pos, len, 0, &locked_page, &fsdata, di_bh, page); if (ret) { if (ret != -ENOSPC) @@ -157,7 +158,7 @@ static int ocfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf) */ down_write(&OCFS2_I(inode)->ip_alloc_sem); - ret = __ocfs2_page_mkwrite(inode, di_bh, page); + ret = __ocfs2_page_mkwrite(vma->vm_file, di_bh, page); up_write(&OCFS2_I(inode)->ip_alloc_sem); -- cgit v0.10.2 From b890823635e022467b924e3d9da8c5166a4e349c Mon Sep 17 00:00:00 2001 From: Tao Ma Date: Thu, 12 Aug 2010 10:27:14 +0800 Subject: ocfs2: pass struct file* to ocfs2_prepare_inode_for_write. struct file * has file_ra_state to store the readahead state and data. So pass this to ocfs2_prepare_inode_for_write. so that it can be used in ocfs2_refcount_cow. Signed-off-by: Tao Ma diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c index 81296b4..5920159 100644 --- a/fs/ocfs2/file.c +++ b/fs/ocfs2/file.c @@ -2053,6 +2053,7 @@ out: } static int ocfs2_prepare_inode_for_refcount(struct inode *inode, + struct file *file, loff_t pos, size_t count, int *meta_level) { @@ -2078,7 +2079,7 @@ out: return ret; } -static int ocfs2_prepare_inode_for_write(struct dentry *dentry, +static int ocfs2_prepare_inode_for_write(struct file *file, loff_t *ppos, size_t count, int appending, @@ -2086,6 +2087,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry, int *has_refcount) { int ret = 0, meta_level = 0; + struct dentry *dentry = file->f_path.dentry; struct inode *inode = dentry->d_inode; loff_t saved_pos, end; @@ -2141,6 +2143,7 @@ static int ocfs2_prepare_inode_for_write(struct dentry *dentry, meta_level = -1; ret = ocfs2_prepare_inode_for_refcount(inode, + file, saved_pos, count, &meta_level); @@ -2255,7 +2258,7 @@ relock: } can_do_direct = direct_io; - ret = ocfs2_prepare_inode_for_write(file->f_path.dentry, ppos, + ret = ocfs2_prepare_inode_for_write(file, ppos, iocb->ki_left, appending, &can_do_direct, &has_refcount); if (ret < 0) { @@ -2385,7 +2388,7 @@ static int ocfs2_splice_to_file(struct pipe_inode_info *pipe, { int ret; - ret = ocfs2_prepare_inode_for_write(out->f_path.dentry, &sd->pos, + ret = ocfs2_prepare_inode_for_write(out, &sd->pos, sd->total_len, 0, NULL, NULL); if (ret < 0) { mlog_errno(ret); -- cgit v0.10.2 From 155027121fe52f9b4f25e9d156c22f2f5012e5fe Mon Sep 17 00:00:00 2001 From: Tao Ma Date: Thu, 12 Aug 2010 10:36:38 +0800 Subject: ocfs2: Add struct file to ocfs2_refcount_cow. Add a new parameter 'struct file *' to ocfs2_refcount_cow so that we can add readahead support later. Signed-off-by: Tao Ma diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index e3efdd9..7155c5a 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -1693,7 +1693,7 @@ int ocfs2_write_begin_nolock(struct file *filp, mlog_errno(ret); goto out; } else if (ret == 1) { - ret = ocfs2_refcount_cow(inode, di_bh, + ret = ocfs2_refcount_cow(inode, filp, di_bh, wc->w_cpos, wc->w_clen, UINT_MAX); if (ret) { mlog_errno(ret); diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c index 5920159..4331f57 100644 --- a/fs/ocfs2/file.c +++ b/fs/ocfs2/file.c @@ -361,7 +361,7 @@ static int ocfs2_cow_file_pos(struct inode *inode, if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) goto out; - return ocfs2_refcount_cow(inode, fe_bh, cpos, 1, cpos+1); + return ocfs2_refcount_cow(inode, NULL, fe_bh, cpos, 1, cpos+1); out: return status; @@ -904,8 +904,8 @@ static int ocfs2_zero_extend_get_range(struct inode *inode, zero_clusters = last_cpos - zero_cpos; if (needs_cow) { - rc = ocfs2_refcount_cow(inode, di_bh, zero_cpos, zero_clusters, - UINT_MAX); + rc = ocfs2_refcount_cow(inode, NULL, di_bh, zero_cpos, + zero_clusters, UINT_MAX); if (rc) { mlog_errno(rc); goto out; @@ -2071,7 +2071,7 @@ static int ocfs2_prepare_inode_for_refcount(struct inode *inode, *meta_level = 1; - ret = ocfs2_refcount_cow(inode, di_bh, cpos, clusters, UINT_MAX); + ret = ocfs2_refcount_cow(inode, file, di_bh, cpos, clusters, UINT_MAX); if (ret) mlog_errno(ret); out: diff --git a/fs/ocfs2/refcounttree.c b/fs/ocfs2/refcounttree.c index 3ac5aa7..66dab3c 100644 --- a/fs/ocfs2/refcounttree.c +++ b/fs/ocfs2/refcounttree.c @@ -3404,6 +3404,7 @@ static int ocfs2_replace_cow(struct ocfs2_cow_context *context) * unrefcounted extent. */ static int ocfs2_refcount_cow_hunk(struct inode *inode, + struct file *file, struct buffer_head *di_bh, u32 cpos, u32 write_len, u32 max_cpos) { @@ -3481,6 +3482,7 @@ out: * clusters between cpos and cpos+write_len are safe to modify. */ int ocfs2_refcount_cow(struct inode *inode, + struct file *file, struct buffer_head *di_bh, u32 cpos, u32 write_len, u32 max_cpos) { @@ -3500,7 +3502,7 @@ int ocfs2_refcount_cow(struct inode *inode, num_clusters = write_len; if (ext_flags & OCFS2_EXT_REFCOUNTED) { - ret = ocfs2_refcount_cow_hunk(inode, di_bh, cpos, + ret = ocfs2_refcount_cow_hunk(inode, file, di_bh, cpos, num_clusters, max_cpos); if (ret) { mlog_errno(ret); diff --git a/fs/ocfs2/refcounttree.h b/fs/ocfs2/refcounttree.h index 9983ba1..29cba0e 100644 --- a/fs/ocfs2/refcounttree.h +++ b/fs/ocfs2/refcounttree.h @@ -52,7 +52,8 @@ int ocfs2_prepare_refcount_change_for_del(struct inode *inode, u32 clusters, int *credits, int *ref_blocks); -int ocfs2_refcount_cow(struct inode *inode, struct buffer_head *di_bh, +int ocfs2_refcount_cow(struct inode *inode, + struct file *filep, struct buffer_head *di_bh, u32 cpos, u32 write_len, u32 max_cpos); typedef int (ocfs2_post_refcount_func)(struct inode *inode, -- cgit v0.10.2 From 7b61cf54a2445ad21df9dd44f0c8bb90154ddea8 Mon Sep 17 00:00:00 2001 From: Tao Ma Date: Thu, 17 Jun 2010 14:14:36 +0800 Subject: ocfs2: Add readahead support for CoW. Add a new function ocfs2_readahead_for_cow so that we start readahead before we start our CoW. Signed-off-by: Tao Ma diff --git a/fs/ocfs2/refcounttree.c b/fs/ocfs2/refcounttree.c index 66dab3c..ad08c11 100644 --- a/fs/ocfs2/refcounttree.c +++ b/fs/ocfs2/refcounttree.c @@ -3398,6 +3398,28 @@ static int ocfs2_replace_cow(struct ocfs2_cow_context *context) return ret; } +static void ocfs2_readahead_for_cow(struct inode *inode, + struct file *file, + u32 start, u32 len) +{ + struct address_space *mapping; + pgoff_t index; + unsigned long num_pages; + int cs_bits = OCFS2_SB(inode->i_sb)->s_clustersize_bits; + + if (!file) + return; + + mapping = file->f_mapping; + num_pages = (len << cs_bits) >> PAGE_CACHE_SHIFT; + if (!num_pages) + num_pages = 1; + + index = ((loff_t)start << cs_bits) >> PAGE_CACHE_SHIFT; + page_cache_sync_readahead(mapping, &file->f_ra, file, + index, num_pages); +} + /* * Starting at cpos, try to CoW write_len clusters. Don't CoW * past max_cpos. This will stop when it runs into a hole or an @@ -3433,6 +3455,8 @@ static int ocfs2_refcount_cow_hunk(struct inode *inode, BUG_ON(cow_len == 0); + ocfs2_readahead_for_cow(inode, file, cow_start, cow_len); + context = kzalloc(sizeof(struct ocfs2_cow_context), GFP_NOFS); if (!context) { ret = -ENOMEM; -- cgit v0.10.2 From 6ea4843f53282465f2bdbe5eedde7d8c3081dfdf Mon Sep 17 00:00:00 2001 From: Tao Ma Date: Thu, 12 Aug 2010 10:31:34 +0800 Subject: ocfs2: Add readhead during CoW. In CoW, when we meet with a readahead page, we know it is time to move the readahead window. So carry out a new readahead. Signed-off-by: Tao Ma diff --git a/fs/ocfs2/refcounttree.c b/fs/ocfs2/refcounttree.c index ad08c11..47549f6 100644 --- a/fs/ocfs2/refcounttree.c +++ b/fs/ocfs2/refcounttree.c @@ -49,6 +49,7 @@ struct ocfs2_cow_context { struct inode *inode; + struct file *file; u32 cow_start; u32 cow_len; struct ocfs2_extent_tree data_et; @@ -2922,13 +2923,16 @@ static int ocfs2_duplicate_clusters_by_page(handle_t *handle, u64 new_block = ocfs2_clusters_to_blocks(sb, new_cluster); struct page *page; pgoff_t page_index; - unsigned int from, to; + unsigned int from, to, readahead_pages; loff_t offset, end, map_end; struct address_space *mapping = context->inode->i_mapping; mlog(0, "old_cluster %u, new %u, len %u at offset %u\n", old_cluster, new_cluster, new_len, cpos); + readahead_pages = + (ocfs2_cow_contig_clusters(sb) << + OCFS2_SB(sb)->s_clustersize_bits) >> PAGE_CACHE_SHIFT; offset = ((loff_t)cpos) << OCFS2_SB(sb)->s_clustersize_bits; end = offset + (new_len << OCFS2_SB(sb)->s_clustersize_bits); /* @@ -2959,6 +2963,14 @@ static int ocfs2_duplicate_clusters_by_page(handle_t *handle, if (PAGE_CACHE_SIZE <= OCFS2_SB(sb)->s_clustersize) BUG_ON(PageDirty(page)); + if (PageReadahead(page) && context->file) { + page_cache_async_readahead(mapping, + &context->file->f_ra, + context->file, + page, page_index, + readahead_pages); + } + if (!PageUptodate(page)) { ret = block_read_full_page(page, ocfs2_get_block); if (ret) { @@ -3478,6 +3490,7 @@ static int ocfs2_refcount_cow_hunk(struct inode *inode, context->ref_root_bh = ref_root_bh; context->cow_duplicate_clusters = ocfs2_duplicate_clusters_by_page; context->get_clusters = ocfs2_di_get_clusters; + context->file = file; ocfs2_init_dinode_extent_tree(&context->data_et, INODE_CACHE(inode), di_bh); -- cgit v0.10.2