| Message ID | 1520908964-70345-1-git-send-email-cgxu519@gmx.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, Mar 13, 2018 at 10:42 AM, Chengguang Xu <cgxu519@gmx.com> wrote: > In current code, regular file and directory use same struct > ceph_file_info to store fs specific data so the struct has to > include some fields which are only used for directory > (e.g., readdir related info), when having plenty of regular files, > it will lead to memory waste. > > This patch introduces dedicated ceph_dir_file_info cache for > directory and delete readdir related thins from ceph_file_info, > so that regular file does not include those unused fields anymore. > Also, chagned to manipulate fscache after reuglar file acquires > ceph_file_info successfully. > Applied with some modifications. I remove fscache related change, pass 'dfi' directly to readdir helper functions. Regards Yan, Zheng > Signed-off-by: Chengguang Xu <cgxu519@gmx.com> > --- > Changes since v2: > - Rebase to testing tree. > - Introduce dfi only for readdir related things. > > Changes since v1: > - Modify ceph_dir_file_info to include ceph_file_info instead of pointing to it. > > fs/ceph/dir.c | 161 +++++++++++++++++++++++-------------------- > fs/ceph/file.c | 91 ++++++++++++++++-------- > fs/ceph/super.c | 8 +++ > fs/ceph/super.h | 9 +++ > include/linux/ceph/libceph.h | 1 + > 5 files changed, 169 insertions(+), 101 deletions(-) > > diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c > index 51d09a3..be485c1 100644 > --- a/fs/ceph/dir.c > +++ b/fs/ceph/dir.c > @@ -105,15 +105,17 @@ static int fpos_cmp(loff_t l, loff_t r) > static int note_last_dentry(struct ceph_file_info *fi, const char *name, > int len, unsigned int next_offset) > { > + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); > + > char *buf = kmalloc(len+1, GFP_KERNEL); > if (!buf) > return -ENOMEM; > - kfree(fi->last_name); > - fi->last_name = buf; > - memcpy(fi->last_name, name, len); > - fi->last_name[len] = 0; > - fi->next_offset = next_offset; > - dout("note_last_dentry '%s'\n", fi->last_name); > + kfree(dfi->last_name); > + dfi->last_name = buf; > + memcpy(dfi->last_name, name, len); > + dfi->last_name[len] = 0; > + dfi->next_offset = next_offset; > + dout("note_last_dentry '%s'\n", dfi->last_name); > return 0; > } > > @@ -176,6 +178,7 @@ static int __dcache_readdir(struct file *file, struct dir_context *ctx, > int shared_gen) > { > struct ceph_file_info *fi = file->private_data; > + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); > struct dentry *parent = file->f_path.dentry; > struct inode *dir = d_inode(parent); > struct dentry *dentry, *last = NULL; > @@ -280,9 +283,9 @@ static int __dcache_readdir(struct file *file, struct dir_context *ctx, > err = ret; > dput(last); > /* last_name no longer match cache index */ > - if (fi->readdir_cache_idx >= 0) { > - fi->readdir_cache_idx = -1; > - fi->dir_release_count = 0; > + if (dfi->readdir_cache_idx >= 0) { > + dfi->readdir_cache_idx = -1; > + dfi->dir_release_count = 0; > } > } > return err; > @@ -290,17 +293,20 @@ static int __dcache_readdir(struct file *file, struct dir_context *ctx, > > static bool need_send_readdir(struct ceph_file_info *fi, loff_t pos) > { > - if (!fi->last_readdir) > + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); > + > + if (!dfi->last_readdir) > return true; > if (is_hash_order(pos)) > - return !ceph_frag_contains_value(fi->frag, fpos_hash(pos)); > + return !ceph_frag_contains_value(dfi->frag, fpos_hash(pos)); > else > - return fi->frag != fpos_frag(pos); > + return dfi->frag != fpos_frag(pos); > } > > static int ceph_readdir(struct file *file, struct dir_context *ctx) > { > struct ceph_file_info *fi = file->private_data; > + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); > struct inode *inode = file_inode(file); > struct ceph_inode_info *ci = ceph_inode(inode); > struct ceph_fs_client *fsc = ceph_inode_to_client(inode); > @@ -358,9 +364,9 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) > CEPH_MDS_OP_LSSNAP : CEPH_MDS_OP_READDIR; > > /* discard old result, if any */ > - if (fi->last_readdir) { > - ceph_mdsc_put_request(fi->last_readdir); > - fi->last_readdir = NULL; > + if (dfi->last_readdir) { > + ceph_mdsc_put_request(dfi->last_readdir); > + dfi->last_readdir = NULL; > } > > if (is_hash_order(ctx->pos)) { > @@ -374,7 +380,7 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) > } > > dout("readdir fetching %llx.%llx frag %x offset '%s'\n", > - ceph_vinop(inode), frag, fi->last_name); > + ceph_vinop(inode), frag, dfi->last_name); > req = ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS); > if (IS_ERR(req)) > return PTR_ERR(req); > @@ -390,8 +396,8 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) > __set_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags); > req->r_inode_drop = CEPH_CAP_FILE_EXCL; > } > - if (fi->last_name) { > - req->r_path2 = kstrdup(fi->last_name, GFP_KERNEL); > + if (dfi->last_name) { > + req->r_path2 = kstrdup(dfi->last_name, GFP_KERNEL); > if (!req->r_path2) { > ceph_mdsc_put_request(req); > return -ENOMEM; > @@ -401,10 +407,10 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) > cpu_to_le32(fpos_hash(ctx->pos)); > } > > - req->r_dir_release_cnt = fi->dir_release_count; > - req->r_dir_ordered_cnt = fi->dir_ordered_count; > - req->r_readdir_cache_idx = fi->readdir_cache_idx; > - req->r_readdir_offset = fi->next_offset; > + req->r_dir_release_cnt = dfi->dir_release_count; > + req->r_dir_ordered_cnt = dfi->dir_ordered_count; > + req->r_readdir_cache_idx = dfi->readdir_cache_idx; > + req->r_readdir_offset = dfi->next_offset; > req->r_args.readdir.frag = cpu_to_le32(frag); > req->r_args.readdir.flags = > cpu_to_le16(CEPH_READDIR_REPLY_BITFLAGS); > @@ -428,35 +434,35 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) > if (le32_to_cpu(rinfo->dir_dir->frag) != frag) { > frag = le32_to_cpu(rinfo->dir_dir->frag); > if (!rinfo->hash_order) { > - fi->next_offset = req->r_readdir_offset; > + dfi->next_offset = req->r_readdir_offset; > /* adjust ctx->pos to beginning of frag */ > ctx->pos = ceph_make_fpos(frag, > - fi->next_offset, > + dfi->next_offset, > false); > } > } > > - fi->frag = frag; > - fi->last_readdir = req; > + dfi->frag = frag; > + dfi->last_readdir = req; > > if (test_bit(CEPH_MDS_R_DID_PREPOPULATE, &req->r_req_flags)) { > - fi->readdir_cache_idx = req->r_readdir_cache_idx; > - if (fi->readdir_cache_idx < 0) { > + dfi->readdir_cache_idx = req->r_readdir_cache_idx; > + if (dfi->readdir_cache_idx < 0) { > /* preclude from marking dir ordered */ > - fi->dir_ordered_count = 0; > + dfi->dir_ordered_count = 0; > } else if (ceph_frag_is_leftmost(frag) && > - fi->next_offset == 2) { > + dfi->next_offset == 2) { > /* note dir version at start of readdir so > * we can tell if any dentries get dropped */ > - fi->dir_release_count = req->r_dir_release_cnt; > - fi->dir_ordered_count = req->r_dir_ordered_cnt; > + dfi->dir_release_count = req->r_dir_release_cnt; > + dfi->dir_ordered_count = req->r_dir_ordered_cnt; > } > } else { > dout("readdir !did_prepopulate\n"); > /* disable readdir cache */ > - fi->readdir_cache_idx = -1; > + dfi->readdir_cache_idx = -1; > /* preclude from marking dir complete */ > - fi->dir_release_count = 0; > + dfi->dir_release_count = 0; > } > > /* note next offset and last dentry name */ > @@ -470,14 +476,14 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) > if (err) > return err; > } else if (req->r_reply_info.dir_end) { > - fi->next_offset = 2; > + dfi->next_offset = 2; > /* keep last name */ > } > } > > - rinfo = &fi->last_readdir->r_reply_info; > + rinfo = &dfi->last_readdir->r_reply_info; > dout("readdir frag %x num %d pos %llx chunk first %llx\n", > - fi->frag, rinfo->dir_nr, ctx->pos, > + dfi->frag, rinfo->dir_nr, ctx->pos, > rinfo->dir_nr ? rinfo->dir_entries[0].offset : 0LL); > > i = 0; > @@ -521,27 +527,28 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) > ctx->pos++; > } > > - ceph_mdsc_put_request(fi->last_readdir); > - fi->last_readdir = NULL; > + ceph_mdsc_put_request(dfi->last_readdir); > + dfi->last_readdir = NULL; > > - if (fi->next_offset > 2) { > - frag = fi->frag; > + if (dfi->next_offset > 2) { > + frag = dfi->frag; > goto more; > } > > /* more frags? */ > - if (!ceph_frag_is_rightmost(fi->frag)) { > - frag = ceph_frag_next(fi->frag); > + if (!ceph_frag_is_rightmost(dfi->frag)) { > + frag = ceph_frag_next(dfi->frag); > if (is_hash_order(ctx->pos)) { > loff_t new_pos = ceph_make_fpos(ceph_frag_value(frag), > - fi->next_offset, true); > + dfi->next_offset, true); > if (new_pos > ctx->pos) > ctx->pos = new_pos; > /* keep last_name */ > } else { > - ctx->pos = ceph_make_fpos(frag, fi->next_offset, false); > - kfree(fi->last_name); > - fi->last_name = NULL; > + ctx->pos = ceph_make_fpos(frag, dfi->next_offset, > + false); > + kfree(dfi->last_name); > + dfi->last_name = NULL; > } > dout("readdir next frag is %x\n", frag); > goto more; > @@ -553,20 +560,22 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) > * were released during the whole readdir, and we should have > * the complete dir contents in our cache. > */ > - if (atomic64_read(&ci->i_release_count) == fi->dir_release_count) { > + if (atomic64_read(&ci->i_release_count) == > + dfi->dir_release_count) { > spin_lock(&ci->i_ceph_lock); > - if (fi->dir_ordered_count == atomic64_read(&ci->i_ordered_count)) { > + if (dfi->dir_ordered_count == > + atomic64_read(&ci->i_ordered_count)) { > dout(" marking %p complete and ordered\n", inode); > /* use i_size to track number of entries in > * readdir cache */ > - BUG_ON(fi->readdir_cache_idx < 0); > - i_size_write(inode, fi->readdir_cache_idx * > + BUG_ON(dfi->readdir_cache_idx < 0); > + i_size_write(inode, dfi->readdir_cache_idx * > sizeof(struct dentry*)); > } else { > dout(" marking %p complete\n", inode); > } > - __ceph_dir_set_complete(ci, fi->dir_release_count, > - fi->dir_ordered_count); > + __ceph_dir_set_complete(ci, dfi->dir_release_count, > + dfi->dir_ordered_count); > spin_unlock(&ci->i_ceph_lock); > } > > @@ -576,15 +585,17 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) > > static void reset_readdir(struct ceph_file_info *fi) > { > - if (fi->last_readdir) { > - ceph_mdsc_put_request(fi->last_readdir); > - fi->last_readdir = NULL; > + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); > + > + if (dfi->last_readdir) { > + ceph_mdsc_put_request(dfi->last_readdir); > + dfi->last_readdir = NULL; > } > - kfree(fi->last_name); > - fi->last_name = NULL; > - fi->dir_release_count = 0; > - fi->readdir_cache_idx = -1; > - fi->next_offset = 2; /* compensate for . and .. */ > + kfree(dfi->last_name); > + dfi->last_name = NULL; > + dfi->dir_release_count = 0; > + dfi->readdir_cache_idx = -1; > + dfi->next_offset = 2; /* compensate for . and .. */ > fi->flags &= ~CEPH_F_ATEND; > } > > @@ -594,6 +605,7 @@ static void reset_readdir(struct ceph_file_info *fi) > */ > static bool need_reset_readdir(struct ceph_file_info *fi, loff_t new_pos) > { > + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); > struct ceph_mds_reply_info_parsed *rinfo; > loff_t chunk_offset; > if (new_pos == 0) > @@ -601,10 +613,10 @@ static bool need_reset_readdir(struct ceph_file_info *fi, loff_t new_pos) > if (is_hash_order(new_pos)) { > /* no need to reset last_name for a forward seek when > * dentries are sotred in hash order */ > - } else if (fi->frag != fpos_frag(new_pos)) { > + } else if (dfi->frag != fpos_frag(new_pos)) { > return true; > } > - rinfo = fi->last_readdir ? &fi->last_readdir->r_reply_info : NULL; > + rinfo = dfi->last_readdir ? &dfi->last_readdir->r_reply_info : NULL; > if (!rinfo || !rinfo->dir_nr) > return true; > chunk_offset = rinfo->dir_entries[0].offset; > @@ -615,6 +627,7 @@ static bool need_reset_readdir(struct ceph_file_info *fi, loff_t new_pos) > static loff_t ceph_dir_llseek(struct file *file, loff_t offset, int whence) > { > struct ceph_file_info *fi = file->private_data; > + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); > struct inode *inode = file->f_mapping->host; > loff_t retval; > > @@ -638,8 +651,8 @@ static loff_t ceph_dir_llseek(struct file *file, loff_t offset, int whence) > } else if (is_hash_order(offset) && offset > file->f_pos) { > /* for hash offset, we don't know if a forward seek > * is within same frag */ > - fi->dir_release_count = 0; > - fi->readdir_cache_idx = -1; > + dfi->dir_release_count = 0; > + dfi->readdir_cache_idx = -1; > } > > if (offset != file->f_pos) { > @@ -1370,7 +1383,7 @@ static void ceph_d_prune(struct dentry *dentry) > static ssize_t ceph_read_dir(struct file *file, char __user *buf, size_t size, > loff_t *ppos) > { > - struct ceph_file_info *fi = file->private_data; > + struct ceph_dir_file_info *dfi = file->private_data; > struct inode *inode = file_inode(file); > struct ceph_inode_info *ci = ceph_inode(inode); > int left; > @@ -1379,12 +1392,12 @@ static ssize_t ceph_read_dir(struct file *file, char __user *buf, size_t size, > if (!ceph_test_mount_opt(ceph_sb_to_client(inode->i_sb), DIRSTAT)) > return -EISDIR; > > - if (!fi->dir_info) { > - fi->dir_info = kmalloc(bufsize, GFP_KERNEL); > - if (!fi->dir_info) > + if (!dfi->dir_info) { > + dfi->dir_info = kmalloc(bufsize, GFP_KERNEL); > + if (!dfi->dir_info) > return -ENOMEM; > - fi->dir_info_len = > - snprintf(fi->dir_info, bufsize, > + dfi->dir_info_len = > + snprintf(dfi->dir_info, bufsize, > "entries: %20lld\n" > " files: %20lld\n" > " subdirs: %20lld\n" > @@ -1404,10 +1417,10 @@ static ssize_t ceph_read_dir(struct file *file, char __user *buf, size_t size, > (long)ci->i_rctime.tv_nsec); > } > > - if (*ppos >= fi->dir_info_len) > + if (*ppos >= dfi->dir_info_len) > return 0; > - size = min_t(unsigned int, size, fi->dir_info_len-*ppos); > - left = copy_to_user(buf, fi->dir_info + *ppos, size); > + size = min_t(unsigned int, size, dfi->dir_info_len-*ppos); > + left = copy_to_user(buf, dfi->dir_info + *ppos, size); > if (left == size) > return -EFAULT; > *ppos += (size - left); > diff --git a/fs/ceph/file.c b/fs/ceph/file.c > index 2de7031..c782beb 100644 > --- a/fs/ceph/file.c > +++ b/fs/ceph/file.c > @@ -95,36 +95,61 @@ static __le32 ceph_flags_sys2wire(u32 flags) > return req; > } > > +static int ceph_init_file_info(struct inode *inode, struct file *file, > + int fmode, bool isdir) > +{ > + struct ceph_file_info *fi; > + struct ceph_dir_file_info *dfi; > + > + dout("%s %p %p 0%o (%s)\n", __func__, inode, file, > + inode->i_mode, isdir ? "dir" : "regular"); > + BUG_ON(inode->i_fop->release != ceph_release); > + > + if (isdir) { > + dfi = kmem_cache_zalloc(ceph_dir_file_cachep, GFP_KERNEL); > + if (!dfi) { > + ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */ > + return -ENOMEM; > + } > + > + file->private_data = dfi; > + fi = &dfi->file_info; > + dfi->next_offset = 2; > + dfi->readdir_cache_idx = -1; > + } else { > + fi = kmem_cache_zalloc(ceph_file_cachep, GFP_KERNEL); > + if (!fi) { > + ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */ > + return -ENOMEM; > + } > + > + file->private_data = fi; > + ceph_fscache_register_inode_cookie(inode); > + ceph_fscache_file_set_cookie(inode, file); > + } > + > + fi->fmode = fmode; > + spin_lock_init(&fi->rw_contexts_lock); > + INIT_LIST_HEAD(&fi->rw_contexts); > + > + return 0; > +} > + > /* > * initialize private struct file data. > * if we fail, clean up by dropping fmode reference on the ceph_inode > */ > static int ceph_init_file(struct inode *inode, struct file *file, int fmode) > { > - struct ceph_file_info *fi; > int ret = 0; > > switch (inode->i_mode & S_IFMT) { > case S_IFREG: > - ceph_fscache_register_inode_cookie(inode); > - ceph_fscache_file_set_cookie(inode, file); > case S_IFDIR: > - dout("init_file %p %p 0%o (regular)\n", inode, file, > - inode->i_mode); > - fi = kmem_cache_zalloc(ceph_file_cachep, GFP_KERNEL); > - if (!fi) { > - ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */ > - return -ENOMEM; > - } > - fi->fmode = fmode; > - > - spin_lock_init(&fi->rw_contexts_lock); > - INIT_LIST_HEAD(&fi->rw_contexts); > - > - fi->next_offset = 2; > - fi->readdir_cache_idx = -1; > - file->private_data = fi; > - BUG_ON(inode->i_fop->release != ceph_release); > + ret = ceph_init_file_info(inode, file, fmode, > + S_ISDIR(inode->i_mode)); > + if (ret) > + return ret; > break; > > case S_IFLNK: > @@ -399,15 +424,27 @@ int ceph_release(struct inode *inode, struct file *file) > { > struct ceph_inode_info *ci = ceph_inode(inode); > struct ceph_file_info *fi = file->private_data; > + struct ceph_dir_file_info *dfi; > > - dout("release inode %p file %p\n", inode, file); > - ceph_put_fmode(ci, fi->fmode); > - if (fi->last_readdir) > - ceph_mdsc_put_request(fi->last_readdir); > - kfree(fi->last_name); > - kfree(fi->dir_info); > - WARN_ON(!list_empty(&fi->rw_contexts)); > - kmem_cache_free(ceph_file_cachep, fi); > + if (S_ISDIR(inode->i_mode)) { > + dout("release inode %p dir file %p\n", inode, file); > + WARN_ON(!list_empty(&fi->rw_contexts)); > + > + dfi = CEPH_DFI(fi); > + ceph_put_fmode(ci, fi->fmode); > + > + if (dfi->last_readdir) > + ceph_mdsc_put_request(dfi->last_readdir); > + kfree(dfi->last_name); > + kfree(dfi->dir_info); > + kmem_cache_free(ceph_dir_file_cachep, dfi); > + } else { > + dout("release inode %p regular file %p\n", inode, file); > + WARN_ON(!list_empty(&fi->rw_contexts)); > + > + ceph_put_fmode(ci, fi->fmode); > + kmem_cache_free(ceph_file_cachep, fi); > + } > > /* wake up anyone waiting for caps on this inode */ > wake_up_all(&ci->i_cap_wq); > diff --git a/fs/ceph/super.c b/fs/ceph/super.c > index 880e83e..286e8db 100644 > --- a/fs/ceph/super.c > +++ b/fs/ceph/super.c > @@ -699,6 +699,7 @@ static void destroy_fs_client(struct ceph_fs_client *fsc) > struct kmem_cache *ceph_cap_flush_cachep; > struct kmem_cache *ceph_dentry_cachep; > struct kmem_cache *ceph_file_cachep; > +struct kmem_cache *ceph_dir_file_cachep; > > static void ceph_inode_init_once(void *foo) > { > @@ -735,6 +736,10 @@ static int __init init_caches(void) > if (!ceph_file_cachep) > goto bad_file; > > + ceph_dir_file_cachep = KMEM_CACHE(ceph_dir_file_info, SLAB_MEM_SPREAD); > + if (!ceph_dir_file_cachep) > + goto bad_dir_file; > + > error = ceph_fscache_register(); > if (error) > goto bad_fscache; > @@ -742,6 +747,8 @@ static int __init init_caches(void) > return 0; > > bad_fscache: > + kmem_cache_destroy(ceph_dir_file_cachep); > +bad_dir_file: > kmem_cache_destroy(ceph_file_cachep); > bad_file: > kmem_cache_destroy(ceph_dentry_cachep); > @@ -767,6 +774,7 @@ static void destroy_caches(void) > kmem_cache_destroy(ceph_cap_flush_cachep); > kmem_cache_destroy(ceph_dentry_cachep); > kmem_cache_destroy(ceph_file_cachep); > + kmem_cache_destroy(ceph_dir_file_cachep); > > ceph_fscache_unregister(); > } > diff --git a/fs/ceph/super.h b/fs/ceph/super.h > index 4e619de..2e5ff4a 100644 > --- a/fs/ceph/super.h > +++ b/fs/ceph/super.h > @@ -678,6 +678,10 @@ struct ceph_file_info { > > spinlock_t rw_contexts_lock; > struct list_head rw_contexts; > +}; > + > +struct ceph_dir_file_info { > + struct ceph_file_info file_info; > > /* readdir: position within the dir */ > u32 frag; > @@ -695,6 +699,11 @@ struct ceph_file_info { > int dir_info_len; > }; > > +static inline struct ceph_dir_file_info *CEPH_DFI(struct ceph_file_info *fi) > +{ > + return container_of(fi, struct ceph_dir_file_info, file_info); > +} > + > struct ceph_rw_context { > struct list_head list; > struct task_struct *thread; > diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h > index c2ec44c..49c93b9 100644 > --- a/include/linux/ceph/libceph.h > +++ b/include/linux/ceph/libceph.h > @@ -262,6 +262,7 @@ static inline int calc_pages_for(u64 off, u64 len) > extern struct kmem_cache *ceph_cap_flush_cachep; > extern struct kmem_cache *ceph_dentry_cachep; > extern struct kmem_cache *ceph_file_cachep; > +extern struct kmem_cache *ceph_dir_file_cachep; > > /* ceph_common.c */ > extern bool libceph_compatible(void *data); > -- > 1.8.3.1 > > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" 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/fs/ceph/dir.c b/fs/ceph/dir.c index 51d09a3..be485c1 100644 --- a/fs/ceph/dir.c +++ b/fs/ceph/dir.c @@ -105,15 +105,17 @@ static int fpos_cmp(loff_t l, loff_t r) static int note_last_dentry(struct ceph_file_info *fi, const char *name, int len, unsigned int next_offset) { + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); + char *buf = kmalloc(len+1, GFP_KERNEL); if (!buf) return -ENOMEM; - kfree(fi->last_name); - fi->last_name = buf; - memcpy(fi->last_name, name, len); - fi->last_name[len] = 0; - fi->next_offset = next_offset; - dout("note_last_dentry '%s'\n", fi->last_name); + kfree(dfi->last_name); + dfi->last_name = buf; + memcpy(dfi->last_name, name, len); + dfi->last_name[len] = 0; + dfi->next_offset = next_offset; + dout("note_last_dentry '%s'\n", dfi->last_name); return 0; } @@ -176,6 +178,7 @@ static int __dcache_readdir(struct file *file, struct dir_context *ctx, int shared_gen) { struct ceph_file_info *fi = file->private_data; + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); struct dentry *parent = file->f_path.dentry; struct inode *dir = d_inode(parent); struct dentry *dentry, *last = NULL; @@ -280,9 +283,9 @@ static int __dcache_readdir(struct file *file, struct dir_context *ctx, err = ret; dput(last); /* last_name no longer match cache index */ - if (fi->readdir_cache_idx >= 0) { - fi->readdir_cache_idx = -1; - fi->dir_release_count = 0; + if (dfi->readdir_cache_idx >= 0) { + dfi->readdir_cache_idx = -1; + dfi->dir_release_count = 0; } } return err; @@ -290,17 +293,20 @@ static int __dcache_readdir(struct file *file, struct dir_context *ctx, static bool need_send_readdir(struct ceph_file_info *fi, loff_t pos) { - if (!fi->last_readdir) + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); + + if (!dfi->last_readdir) return true; if (is_hash_order(pos)) - return !ceph_frag_contains_value(fi->frag, fpos_hash(pos)); + return !ceph_frag_contains_value(dfi->frag, fpos_hash(pos)); else - return fi->frag != fpos_frag(pos); + return dfi->frag != fpos_frag(pos); } static int ceph_readdir(struct file *file, struct dir_context *ctx) { struct ceph_file_info *fi = file->private_data; + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); struct inode *inode = file_inode(file); struct ceph_inode_info *ci = ceph_inode(inode); struct ceph_fs_client *fsc = ceph_inode_to_client(inode); @@ -358,9 +364,9 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) CEPH_MDS_OP_LSSNAP : CEPH_MDS_OP_READDIR; /* discard old result, if any */ - if (fi->last_readdir) { - ceph_mdsc_put_request(fi->last_readdir); - fi->last_readdir = NULL; + if (dfi->last_readdir) { + ceph_mdsc_put_request(dfi->last_readdir); + dfi->last_readdir = NULL; } if (is_hash_order(ctx->pos)) { @@ -374,7 +380,7 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) } dout("readdir fetching %llx.%llx frag %x offset '%s'\n", - ceph_vinop(inode), frag, fi->last_name); + ceph_vinop(inode), frag, dfi->last_name); req = ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS); if (IS_ERR(req)) return PTR_ERR(req); @@ -390,8 +396,8 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) __set_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags); req->r_inode_drop = CEPH_CAP_FILE_EXCL; } - if (fi->last_name) { - req->r_path2 = kstrdup(fi->last_name, GFP_KERNEL); + if (dfi->last_name) { + req->r_path2 = kstrdup(dfi->last_name, GFP_KERNEL); if (!req->r_path2) { ceph_mdsc_put_request(req); return -ENOMEM; @@ -401,10 +407,10 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) cpu_to_le32(fpos_hash(ctx->pos)); } - req->r_dir_release_cnt = fi->dir_release_count; - req->r_dir_ordered_cnt = fi->dir_ordered_count; - req->r_readdir_cache_idx = fi->readdir_cache_idx; - req->r_readdir_offset = fi->next_offset; + req->r_dir_release_cnt = dfi->dir_release_count; + req->r_dir_ordered_cnt = dfi->dir_ordered_count; + req->r_readdir_cache_idx = dfi->readdir_cache_idx; + req->r_readdir_offset = dfi->next_offset; req->r_args.readdir.frag = cpu_to_le32(frag); req->r_args.readdir.flags = cpu_to_le16(CEPH_READDIR_REPLY_BITFLAGS); @@ -428,35 +434,35 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) if (le32_to_cpu(rinfo->dir_dir->frag) != frag) { frag = le32_to_cpu(rinfo->dir_dir->frag); if (!rinfo->hash_order) { - fi->next_offset = req->r_readdir_offset; + dfi->next_offset = req->r_readdir_offset; /* adjust ctx->pos to beginning of frag */ ctx->pos = ceph_make_fpos(frag, - fi->next_offset, + dfi->next_offset, false); } } - fi->frag = frag; - fi->last_readdir = req; + dfi->frag = frag; + dfi->last_readdir = req; if (test_bit(CEPH_MDS_R_DID_PREPOPULATE, &req->r_req_flags)) { - fi->readdir_cache_idx = req->r_readdir_cache_idx; - if (fi->readdir_cache_idx < 0) { + dfi->readdir_cache_idx = req->r_readdir_cache_idx; + if (dfi->readdir_cache_idx < 0) { /* preclude from marking dir ordered */ - fi->dir_ordered_count = 0; + dfi->dir_ordered_count = 0; } else if (ceph_frag_is_leftmost(frag) && - fi->next_offset == 2) { + dfi->next_offset == 2) { /* note dir version at start of readdir so * we can tell if any dentries get dropped */ - fi->dir_release_count = req->r_dir_release_cnt; - fi->dir_ordered_count = req->r_dir_ordered_cnt; + dfi->dir_release_count = req->r_dir_release_cnt; + dfi->dir_ordered_count = req->r_dir_ordered_cnt; } } else { dout("readdir !did_prepopulate\n"); /* disable readdir cache */ - fi->readdir_cache_idx = -1; + dfi->readdir_cache_idx = -1; /* preclude from marking dir complete */ - fi->dir_release_count = 0; + dfi->dir_release_count = 0; } /* note next offset and last dentry name */ @@ -470,14 +476,14 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) if (err) return err; } else if (req->r_reply_info.dir_end) { - fi->next_offset = 2; + dfi->next_offset = 2; /* keep last name */ } } - rinfo = &fi->last_readdir->r_reply_info; + rinfo = &dfi->last_readdir->r_reply_info; dout("readdir frag %x num %d pos %llx chunk first %llx\n", - fi->frag, rinfo->dir_nr, ctx->pos, + dfi->frag, rinfo->dir_nr, ctx->pos, rinfo->dir_nr ? rinfo->dir_entries[0].offset : 0LL); i = 0; @@ -521,27 +527,28 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) ctx->pos++; } - ceph_mdsc_put_request(fi->last_readdir); - fi->last_readdir = NULL; + ceph_mdsc_put_request(dfi->last_readdir); + dfi->last_readdir = NULL; - if (fi->next_offset > 2) { - frag = fi->frag; + if (dfi->next_offset > 2) { + frag = dfi->frag; goto more; } /* more frags? */ - if (!ceph_frag_is_rightmost(fi->frag)) { - frag = ceph_frag_next(fi->frag); + if (!ceph_frag_is_rightmost(dfi->frag)) { + frag = ceph_frag_next(dfi->frag); if (is_hash_order(ctx->pos)) { loff_t new_pos = ceph_make_fpos(ceph_frag_value(frag), - fi->next_offset, true); + dfi->next_offset, true); if (new_pos > ctx->pos) ctx->pos = new_pos; /* keep last_name */ } else { - ctx->pos = ceph_make_fpos(frag, fi->next_offset, false); - kfree(fi->last_name); - fi->last_name = NULL; + ctx->pos = ceph_make_fpos(frag, dfi->next_offset, + false); + kfree(dfi->last_name); + dfi->last_name = NULL; } dout("readdir next frag is %x\n", frag); goto more; @@ -553,20 +560,22 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) * were released during the whole readdir, and we should have * the complete dir contents in our cache. */ - if (atomic64_read(&ci->i_release_count) == fi->dir_release_count) { + if (atomic64_read(&ci->i_release_count) == + dfi->dir_release_count) { spin_lock(&ci->i_ceph_lock); - if (fi->dir_ordered_count == atomic64_read(&ci->i_ordered_count)) { + if (dfi->dir_ordered_count == + atomic64_read(&ci->i_ordered_count)) { dout(" marking %p complete and ordered\n", inode); /* use i_size to track number of entries in * readdir cache */ - BUG_ON(fi->readdir_cache_idx < 0); - i_size_write(inode, fi->readdir_cache_idx * + BUG_ON(dfi->readdir_cache_idx < 0); + i_size_write(inode, dfi->readdir_cache_idx * sizeof(struct dentry*)); } else { dout(" marking %p complete\n", inode); } - __ceph_dir_set_complete(ci, fi->dir_release_count, - fi->dir_ordered_count); + __ceph_dir_set_complete(ci, dfi->dir_release_count, + dfi->dir_ordered_count); spin_unlock(&ci->i_ceph_lock); } @@ -576,15 +585,17 @@ static int ceph_readdir(struct file *file, struct dir_context *ctx) static void reset_readdir(struct ceph_file_info *fi) { - if (fi->last_readdir) { - ceph_mdsc_put_request(fi->last_readdir); - fi->last_readdir = NULL; + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); + + if (dfi->last_readdir) { + ceph_mdsc_put_request(dfi->last_readdir); + dfi->last_readdir = NULL; } - kfree(fi->last_name); - fi->last_name = NULL; - fi->dir_release_count = 0; - fi->readdir_cache_idx = -1; - fi->next_offset = 2; /* compensate for . and .. */ + kfree(dfi->last_name); + dfi->last_name = NULL; + dfi->dir_release_count = 0; + dfi->readdir_cache_idx = -1; + dfi->next_offset = 2; /* compensate for . and .. */ fi->flags &= ~CEPH_F_ATEND; } @@ -594,6 +605,7 @@ static void reset_readdir(struct ceph_file_info *fi) */ static bool need_reset_readdir(struct ceph_file_info *fi, loff_t new_pos) { + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); struct ceph_mds_reply_info_parsed *rinfo; loff_t chunk_offset; if (new_pos == 0) @@ -601,10 +613,10 @@ static bool need_reset_readdir(struct ceph_file_info *fi, loff_t new_pos) if (is_hash_order(new_pos)) { /* no need to reset last_name for a forward seek when * dentries are sotred in hash order */ - } else if (fi->frag != fpos_frag(new_pos)) { + } else if (dfi->frag != fpos_frag(new_pos)) { return true; } - rinfo = fi->last_readdir ? &fi->last_readdir->r_reply_info : NULL; + rinfo = dfi->last_readdir ? &dfi->last_readdir->r_reply_info : NULL; if (!rinfo || !rinfo->dir_nr) return true; chunk_offset = rinfo->dir_entries[0].offset; @@ -615,6 +627,7 @@ static bool need_reset_readdir(struct ceph_file_info *fi, loff_t new_pos) static loff_t ceph_dir_llseek(struct file *file, loff_t offset, int whence) { struct ceph_file_info *fi = file->private_data; + struct ceph_dir_file_info *dfi = CEPH_DFI(fi); struct inode *inode = file->f_mapping->host; loff_t retval; @@ -638,8 +651,8 @@ static loff_t ceph_dir_llseek(struct file *file, loff_t offset, int whence) } else if (is_hash_order(offset) && offset > file->f_pos) { /* for hash offset, we don't know if a forward seek * is within same frag */ - fi->dir_release_count = 0; - fi->readdir_cache_idx = -1; + dfi->dir_release_count = 0; + dfi->readdir_cache_idx = -1; } if (offset != file->f_pos) { @@ -1370,7 +1383,7 @@ static void ceph_d_prune(struct dentry *dentry) static ssize_t ceph_read_dir(struct file *file, char __user *buf, size_t size, loff_t *ppos) { - struct ceph_file_info *fi = file->private_data; + struct ceph_dir_file_info *dfi = file->private_data; struct inode *inode = file_inode(file); struct ceph_inode_info *ci = ceph_inode(inode); int left; @@ -1379,12 +1392,12 @@ static ssize_t ceph_read_dir(struct file *file, char __user *buf, size_t size, if (!ceph_test_mount_opt(ceph_sb_to_client(inode->i_sb), DIRSTAT)) return -EISDIR; - if (!fi->dir_info) { - fi->dir_info = kmalloc(bufsize, GFP_KERNEL); - if (!fi->dir_info) + if (!dfi->dir_info) { + dfi->dir_info = kmalloc(bufsize, GFP_KERNEL); + if (!dfi->dir_info) return -ENOMEM; - fi->dir_info_len = - snprintf(fi->dir_info, bufsize, + dfi->dir_info_len = + snprintf(dfi->dir_info, bufsize, "entries: %20lld\n" " files: %20lld\n" " subdirs: %20lld\n" @@ -1404,10 +1417,10 @@ static ssize_t ceph_read_dir(struct file *file, char __user *buf, size_t size, (long)ci->i_rctime.tv_nsec); } - if (*ppos >= fi->dir_info_len) + if (*ppos >= dfi->dir_info_len) return 0; - size = min_t(unsigned int, size, fi->dir_info_len-*ppos); - left = copy_to_user(buf, fi->dir_info + *ppos, size); + size = min_t(unsigned int, size, dfi->dir_info_len-*ppos); + left = copy_to_user(buf, dfi->dir_info + *ppos, size); if (left == size) return -EFAULT; *ppos += (size - left); diff --git a/fs/ceph/file.c b/fs/ceph/file.c index 2de7031..c782beb 100644 --- a/fs/ceph/file.c +++ b/fs/ceph/file.c @@ -95,36 +95,61 @@ static __le32 ceph_flags_sys2wire(u32 flags) return req; } +static int ceph_init_file_info(struct inode *inode, struct file *file, + int fmode, bool isdir) +{ + struct ceph_file_info *fi; + struct ceph_dir_file_info *dfi; + + dout("%s %p %p 0%o (%s)\n", __func__, inode, file, + inode->i_mode, isdir ? "dir" : "regular"); + BUG_ON(inode->i_fop->release != ceph_release); + + if (isdir) { + dfi = kmem_cache_zalloc(ceph_dir_file_cachep, GFP_KERNEL); + if (!dfi) { + ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */ + return -ENOMEM; + } + + file->private_data = dfi; + fi = &dfi->file_info; + dfi->next_offset = 2; + dfi->readdir_cache_idx = -1; + } else { + fi = kmem_cache_zalloc(ceph_file_cachep, GFP_KERNEL); + if (!fi) { + ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */ + return -ENOMEM; + } + + file->private_data = fi; + ceph_fscache_register_inode_cookie(inode); + ceph_fscache_file_set_cookie(inode, file); + } + + fi->fmode = fmode; + spin_lock_init(&fi->rw_contexts_lock); + INIT_LIST_HEAD(&fi->rw_contexts); + + return 0; +} + /* * initialize private struct file data. * if we fail, clean up by dropping fmode reference on the ceph_inode */ static int ceph_init_file(struct inode *inode, struct file *file, int fmode) { - struct ceph_file_info *fi; int ret = 0; switch (inode->i_mode & S_IFMT) { case S_IFREG: - ceph_fscache_register_inode_cookie(inode); - ceph_fscache_file_set_cookie(inode, file); case S_IFDIR: - dout("init_file %p %p 0%o (regular)\n", inode, file, - inode->i_mode); - fi = kmem_cache_zalloc(ceph_file_cachep, GFP_KERNEL); - if (!fi) { - ceph_put_fmode(ceph_inode(inode), fmode); /* clean up */ - return -ENOMEM; - } - fi->fmode = fmode; - - spin_lock_init(&fi->rw_contexts_lock); - INIT_LIST_HEAD(&fi->rw_contexts); - - fi->next_offset = 2; - fi->readdir_cache_idx = -1; - file->private_data = fi; - BUG_ON(inode->i_fop->release != ceph_release); + ret = ceph_init_file_info(inode, file, fmode, + S_ISDIR(inode->i_mode)); + if (ret) + return ret; break; case S_IFLNK: @@ -399,15 +424,27 @@ int ceph_release(struct inode *inode, struct file *file) { struct ceph_inode_info *ci = ceph_inode(inode); struct ceph_file_info *fi = file->private_data; + struct ceph_dir_file_info *dfi; - dout("release inode %p file %p\n", inode, file); - ceph_put_fmode(ci, fi->fmode); - if (fi->last_readdir) - ceph_mdsc_put_request(fi->last_readdir); - kfree(fi->last_name); - kfree(fi->dir_info); - WARN_ON(!list_empty(&fi->rw_contexts)); - kmem_cache_free(ceph_file_cachep, fi); + if (S_ISDIR(inode->i_mode)) { + dout("release inode %p dir file %p\n", inode, file); + WARN_ON(!list_empty(&fi->rw_contexts)); + + dfi = CEPH_DFI(fi); + ceph_put_fmode(ci, fi->fmode); + + if (dfi->last_readdir) + ceph_mdsc_put_request(dfi->last_readdir); + kfree(dfi->last_name); + kfree(dfi->dir_info); + kmem_cache_free(ceph_dir_file_cachep, dfi); + } else { + dout("release inode %p regular file %p\n", inode, file); + WARN_ON(!list_empty(&fi->rw_contexts)); + + ceph_put_fmode(ci, fi->fmode); + kmem_cache_free(ceph_file_cachep, fi); + } /* wake up anyone waiting for caps on this inode */ wake_up_all(&ci->i_cap_wq); diff --git a/fs/ceph/super.c b/fs/ceph/super.c index 880e83e..286e8db 100644 --- a/fs/ceph/super.c +++ b/fs/ceph/super.c @@ -699,6 +699,7 @@ static void destroy_fs_client(struct ceph_fs_client *fsc) struct kmem_cache *ceph_cap_flush_cachep; struct kmem_cache *ceph_dentry_cachep; struct kmem_cache *ceph_file_cachep; +struct kmem_cache *ceph_dir_file_cachep; static void ceph_inode_init_once(void *foo) { @@ -735,6 +736,10 @@ static int __init init_caches(void) if (!ceph_file_cachep) goto bad_file; + ceph_dir_file_cachep = KMEM_CACHE(ceph_dir_file_info, SLAB_MEM_SPREAD); + if (!ceph_dir_file_cachep) + goto bad_dir_file; + error = ceph_fscache_register(); if (error) goto bad_fscache; @@ -742,6 +747,8 @@ static int __init init_caches(void) return 0; bad_fscache: + kmem_cache_destroy(ceph_dir_file_cachep); +bad_dir_file: kmem_cache_destroy(ceph_file_cachep); bad_file: kmem_cache_destroy(ceph_dentry_cachep); @@ -767,6 +774,7 @@ static void destroy_caches(void) kmem_cache_destroy(ceph_cap_flush_cachep); kmem_cache_destroy(ceph_dentry_cachep); kmem_cache_destroy(ceph_file_cachep); + kmem_cache_destroy(ceph_dir_file_cachep); ceph_fscache_unregister(); } diff --git a/fs/ceph/super.h b/fs/ceph/super.h index 4e619de..2e5ff4a 100644 --- a/fs/ceph/super.h +++ b/fs/ceph/super.h @@ -678,6 +678,10 @@ struct ceph_file_info { spinlock_t rw_contexts_lock; struct list_head rw_contexts; +}; + +struct ceph_dir_file_info { + struct ceph_file_info file_info; /* readdir: position within the dir */ u32 frag; @@ -695,6 +699,11 @@ struct ceph_file_info { int dir_info_len; }; +static inline struct ceph_dir_file_info *CEPH_DFI(struct ceph_file_info *fi) +{ + return container_of(fi, struct ceph_dir_file_info, file_info); +} + struct ceph_rw_context { struct list_head list; struct task_struct *thread; diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h index c2ec44c..49c93b9 100644 --- a/include/linux/ceph/libceph.h +++ b/include/linux/ceph/libceph.h @@ -262,6 +262,7 @@ static inline int calc_pages_for(u64 off, u64 len) extern struct kmem_cache *ceph_cap_flush_cachep; extern struct kmem_cache *ceph_dentry_cachep; extern struct kmem_cache *ceph_file_cachep; +extern struct kmem_cache *ceph_dir_file_cachep; /* ceph_common.c */ extern bool libceph_compatible(void *data);
In current code, regular file and directory use same struct ceph_file_info to store fs specific data so the struct has to include some fields which are only used for directory (e.g., readdir related info), when having plenty of regular files, it will lead to memory waste. This patch introduces dedicated ceph_dir_file_info cache for directory and delete readdir related thins from ceph_file_info, so that regular file does not include those unused fields anymore. Also, chagned to manipulate fscache after reuglar file acquires ceph_file_info successfully. Signed-off-by: Chengguang Xu <cgxu519@gmx.com> --- Changes since v2: - Rebase to testing tree. - Introduce dfi only for readdir related things. Changes since v1: - Modify ceph_dir_file_info to include ceph_file_info instead of pointing to it. fs/ceph/dir.c | 161 +++++++++++++++++++++++-------------------- fs/ceph/file.c | 91 ++++++++++++++++-------- fs/ceph/super.c | 8 +++ fs/ceph/super.h | 9 +++ include/linux/ceph/libceph.h | 1 + 5 files changed, 169 insertions(+), 101 deletions(-)