From mboxrd@z Thu Jan 1 00:00:00 1970 Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.remote-ip=209.85.128.45 ARC-Seal: i=1; d=subspace.kernel.org; s=arc20261015; a=rsa-sha256; cv=none; t=1793863196; b=KfZZNbCP8yuYKxGfQr8QWs+GaIlDMrDUKnW7sb+tGTRTRKHWLlg+ACFGQaaKF9Kornrt PhLDd/3q55MyVJe/TyvijYRgwfQWTh71FBgoCq6BVf0rgB7Xmt0PsH8Au0aymhavXlcN3 Fa5VzcokQ17t5gu8YmFzfHam1QEnfghEdfKLw/tyaEQBtTE616PQlg7rkuOYnKNksy5mt BCBlBVDCoUivSMrg+vZiSRjw5Za39OMvLkPkwnNFXgTgVbQ/Ezy3a4gQzm9SjMl2x68ln 849yleOlv77LJsbUkXnh2iAK+YtSP77zo+9QpPaOhjzqjnjM+jkmafHNv0vjM+LSu7w== ARC-Message-Signature: i=1; d=subspace.kernel.org; s=arc20261015; a=rsa-sha256; c=relaxed/relaxed; t=1793863196; h=DMARC-Filter:Received:DKIM-Signature:X-Google-DKIM-Signature: X-Forwarded-Encrypted:X-Gm-Message-State:X-Gm-Gg:X-Received:Received: From:To:Cc:Subject:Date:Message-ID:X-Mailer:In-Reply-To:References: MIME-Version:Content-Transfer-Encoding; bh=0+TsRuLTi5B5q+QjcNoQnJZ1NbXIYbMbWvn6J0IrHwc=; b=5o11m6/y5I64Lm7G6kx48Jg9PSRnfK09BmQKyAOjnYVjPhlYTKdeAI3gCp0zlZZFnIET wmaEmBH/gYOuNThlakJ9HnTXBJUpZTmoLwEoohrcJXQQo75/adyqgvVOLUqfU5sP1DS4b TYJFOPWNSFTJwvKNL0POX62q+Nb/hqHPi2b6KbXi1/5yzs0dABalwODdV9WK4od6/Xth3 zAGiQSsnElJf6o3QRw8dDlmnj3a6n5Gy3ZE//hJcYHkjqiKqi3PT5b/Bs1PzeW4QOixZM 8rLZyO62zF6BaR81tkR/RpPZ1lz3PhpZRCWwVgff9BRzAxMq8aohcm2DjC/5Z6oNzzg== ARC-Authentication-Results: i=1; smtp.subspace.kernel.org; dmarc=pass header.from=suse.com; spf=pass smtp.mailfrom=suse.com; dkim=pass header.d=suse.com header.i=@suse.com header.a=rsa-sha256 header.s=google header.b=EOCJpwTq; arc=none smtp.remote-ip=209.85.128.45 Received: from mail-wm1-f45.google.com (mail-wm1-f45.google.com [209.85.128.45]) (using TLSv1.3 with cipher TLS_AES_128_GCM_SHA256 (128/128 bits) key-exchange X25519 server-signature ECDSA (prime256v1) server-digest SHA256) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id 29963606A329 for ; Thu, 5 Nov 2026 07:19:55 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.4.2 smtp.subspace.kernel.org 29963606A329 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=quarantine dis=none) header.from=suse.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=suse.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key, unprotected) header.d=suse.com header.i=@suse.com header.a=rsa-sha256 header.s=google header.b=EOCJpwTq Received: by mail-wm1-f45.google.com with SMTP id 5b1f17b1804b1-4a56e05568aso4917815e9.1 for ; Wed, 04 Nov 2026 23:19:55 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=suse.com; s=google; t=1793863194; x=1794467994; darn=vger.kernel.org; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:from:to:cc:subject:date :message-id:reply-to:content-type; bh=0+TsRuLTi5B5q+QjcNoQnJZ1NbXIYbMbWvn6J0IrHwc=; b=EOCJpwTqJugiAIixF2RaT3bl8cvPVzHgmoi1YcLh+oPtm9BslUd6JsnvqaFHBZvcVb fMzy/cZNF9p2VXPPgU3QUEqmmnZhtPnGWsI+RP68sLKlKyhGS8iEfldUGRKdkYH16AhS wsj7fy+RqoIAm8xcGg8gdasDRP9HK0gmsOTrEGpe539X7GtEkaT1fqANqlNIxdCkssyR pPVD+k1dydIxS171WxFb25rFI1mw8NGaG8fVY4IbH29WAFLZ+sjvJJUV9fx+dmXwSzbw p4xJ44bq47A2Vm2b1txSDYdnDqEBxbzIiuvrtiA2HX73w8wjJ9TpRgu9ETK62+IT3hkB KM+w== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20260707; t=1793863194; x=1794467994; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:x-gm-gg:x-gm-message-state:from :to:cc:subject:date:message-id:reply-to:content-type; bh=0+TsRuLTi5B5q+QjcNoQnJZ1NbXIYbMbWvn6J0IrHwc=; b=YrLg8VFZJ40BdwAICva5MoFgxZgJxGa1ce51c0Oy4OekT2WFKJnXJht0SB5fznk3Ja IAXEvZAc9dVv/ka0DKSVH2+TMbOEL2h/qXNawJkqQ+iDed5/e8h5nMBtPSj0F+kQ/2JF EoW5RT0ITPWmALBaQyvEM+F9R2zpGgHWihnisoaiZuNxUIVDTqe3OVNcjAIs/ZV6p0GM DgUaLb48DphykO2lkQQbTs0TL77N+7hbrJgFlnk5ni7M3E5DGe7mIw/ILZa8Z74iYUWP c+dTNkHxRHwCIkibbxQtgGHT5bGkneLOJAcgSONNtInGzKGBr6AdboujHB95sDA2rdwJ nr8A== X-Forwarded-Encrypted: i=1; AKwUvBxO0LY35f/AOtPhyKU4W/eRR4EyQePs1RyXfdcGSqUTUMFPcgsirBJWN2o2XKHBY/ua0XShEYLkyXgSw9k=@vger.kernel.org X-Gm-Message-State: APuU4Z/qjD48yhK3SiUrHHbsw6xGTSTSeu3ybzB/77Pt0toNoCWzAxSK EHsEz6+4c2i0wcTY+I7xPX3yxcmx4wmt3iiaO2eYRdm89ZYchoiR2Lyhx1pildrCw0s= X-Gm-Gg: AYBFou21xgPTK5orDwgohxYjmPBnqkXLF0/miF2cYcTI4cGc03ZRlIF5BdIKuZkWUpv gzj0OpxYC38zKnNdlsu/G0My689zXW8m/6VttkEzj8tAdMAxu9ni1o0XsNzZdpzWC2QfdTGAjFy bCmfm9i22IMA1YBr2uYYBiULbYGKyFtqiqSmnr+E/7zaRSwoE6ZCSNtR/QZfjyvxQz2MoGKRsbf tvf1+1TxA1Uwd+vvnGueIfDwRV9CmiCBKcMnNqqIA74C6LSA8YohIxUj69ZXW785z0ix7/pAIUM 1l6QQwJpajfsJQvskjFZrYPbeAVINBOs6aKy9aM2WdgV1HG4IOOSAA7dOzUbuCstFuee4i8GmsB ZBM+A5ZAhnmH2rDl5/kc1q1rpZu2e9nttiYlnbAZBmOAYlDIaifv7uZh3mCKzaFzEkYcz8rttZE CKGDAaz71aAyi4CRwadvomvwuuvD3k9MHHnAUnihPO6UEiF3XUEuFbeZ2d4WZWCkpvlr+XqdbEE 4dX3qGRfipRuZkpoNggRsS3ecu/MomcO7ROGSZuKmj/Y1lwov1mKIwxpfH6MZe/mhMYvipaMbUD pKose/qBenFPFDERYWL4UMAPzMoHLsEdQoTR4w/BIe0= X-Received: by 2002:a7b:ce0d:0:b0:4a1:715d:1913 with SMTP id 5b1f17b1804b1-4ac177538c5mr304358505e9.1.1793863194273; Wed, 04 Nov 2026 23:19:54 -0800 (PST) Received: from localhost.localdomain ([202.127.77.110]) by smtp.gmail.com with ESMTPSA id a92af1059eb24-1b245b86fbesm14687084c88.1.2026.11.04.23.19.49 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Wed, 04 Nov 2026 23:19:53 -0800 (PST) From: Heming Zhao To: joseph.qi@linux.alibaba.com, mark@fasheh.com, jlbec@evilplan.org, hch@lst.de Cc: Heming Zhao , ocfs2-devel@lists.linux.dev, linux-kernel@vger.kernel.org Subject: [RFC PATCH v5 3/4] ocfs2: switch dio write path from buffer_head to iomap Date: Thu, 5 Nov 2026 15:19:27 +0800 Message-ID: <20261105071930.16399-4-heming.zhao@suse.com> X-Mailer: git-send-email 2.55.0 In-Reply-To: <20261105071930.16399-1-heming.zhao@suse.com> References: <20261105071930.16399-1-heming.zhao@suse.com> X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Post: List-Help: List-Subscribe: List-Unsubscribe: List-Owner: Precedence: list MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Converts OCFS2's DIO write path from the legacy buffer_head infrastructure to the modern iomap framework. Key modifications and designs are as follows: Introduce ocfs2_dio_wr_map_blocks() to allocate and map direct write blocks. This function is similar to the buffer_head path ocfs2_dio_wr_get_block(), which still allocates one cluster per call. Implement ocfs2_dio_write_end_io() and ocfs2_iomap_dio_end_io_write() to handle IO completion. They convert UNWRITTEN extents, updates inode size (EOF). In the buffer_head path, ocfs2_dio_wr_get_block() stores dwc in bh->b_private, enabling it to detect the first call for an IO chunk and invoke ocfs2_add_inode_to_orphan() to protect the extent allocation. However, iomap doesn't provide a similar private field. Therefore, moves ocfs2_add_inode_to_orphan() from ocfs2_dio_wr_map_blocks(the buffer_head counterpart to ocfs2_dio_wr_get_block()) to ocfs2_file_write_iter(). It is also impossible t pass The dwc to iomap_dio_ops.end_io, the iomap ->end_io callback cannot effectively distinguish between I/Os initiated by different threads concurrently. For example, when multiple DIO write threads operate on the same inode concurrently, the end_io callback without the dwc context may incorrectly delete an orphan inode created by another thread, leading to a crash. The fix is to move the orphan inode deletion from end_io into ocfs2_file_write_iter(). In ocfs2_file_write_iter(), inode_lock() serializes the write threads, and IOMAP_DIO_FORCE_WAIT ensures that the function only returns after the bio completes. This guarantees that i_size has already been updated when ocfs2_dio_del_orphan() is called. Doesn't like ocfs2_dio_wr_get_block(), move the zero operations from ocfs2_dio_wr_map_blocks() to ocfs2_file_write_iter(). If do the zeroing in ocfs2_dio_wr_map_blocks() would zero blocks that earlier mappings of this same dio have already written, as i_size is only updated at dio completion. this will makes xfstests generic/538 failure. Assisted-by: LLM Co-developed-by: Joseph Qi Signed-off-by: Joseph Qi Signed-off-by: Heming Zhao --- fs/ocfs2/aops.c | 418 +++++++++++++++++++++++++++++++++++++++++++++--- fs/ocfs2/aops.h | 3 + fs/ocfs2/file.c | 137 +++++++++++++--- 3 files changed, 521 insertions(+), 37 deletions(-) diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 8e4d31cb85db..d5b471bdaec2 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -199,6 +199,19 @@ int ocfs2_map_blocks(struct inode *inode, struct ocfs2_map_block *map, return err; } +static int ocfs2_lock_map_blocks(struct inode *inode, + struct ocfs2_map_block *map, int flags) +{ + int ret = 0; + struct ocfs2_inode_info *oi = OCFS2_I(inode); + + down_read(&oi->ip_alloc_sem); + ret = ocfs2_map_blocks(inode, map, flags); + up_read(&oi->ip_alloc_sem); + + return ret; +} + int ocfs2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create) { @@ -2336,6 +2349,158 @@ static int ocfs2_dio_wr_get_block(struct inode *inode, sector_t iblock, return ret; } +/* + * Prepare the area between i_size and a direct write that starts or ends + * beyond it. Must be called once per dio, before any io is issued, with + * i_rwsem and the rw lock held. + */ +int ocfs2_dio_zero_eof(struct inode *inode, loff_t pos, size_t count) +{ + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct buffer_head *di_bh = NULL; + int ret; + + if (pos + count <= i_size_read(inode)) + return 0; + + ret = ocfs2_inode_lock(inode, &di_bh, 1); + if (ret) { + mlog_errno(ret); + return ret; + } + + down_write(&oi->ip_alloc_sem); + if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb))) + ret = ocfs2_zero_tail(inode, di_bh, pos); + else + ret = ocfs2_expand_nonsparse_inode(inode, di_bh, pos, count, + NULL); + if (ret < 0) + mlog_errno(ret); + up_write(&oi->ip_alloc_sem); + + ocfs2_inode_unlock(inode, 1); + brelse(di_bh); + return ret; +} + +/* copy from ocfs2_dio_wr_get_block */ +static int ocfs2_dio_wr_map_blocks(struct inode *inode, + struct ocfs2_map_block *map, int create) +{ + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct ocfs2_write_ctxt *wc; + struct ocfs2_write_cluster_desc *desc = NULL; + struct buffer_head *di_bh = NULL; + u64 p_blkno; + sector_t iblock = map->lblk; + unsigned int i_blkbits = inode->i_sb->s_blocksize_bits; + loff_t pos = iblock << i_blkbits; + sector_t endblk = (i_size_read(inode) - 1) >> i_blkbits; + unsigned int len, total_len = map->len << i_blkbits; + int ret = 0; + + len = osb->s_clustersize - (pos & (osb->s_clustersize - 1)); + len = min(total_len, len); + /* + * map->len is count in ocfs2_iomap_begin according to write + * "pos" and "end", we need map twice to return different buffer state: + * 1. area in file size, not set NEW; + * 2. area out file size, set NEW. + * + * iblock endblk + * |--------|---------|---------|--------- + * |<-------area in file------->| + */ + if ((iblock <= endblk) && + ((iblock + ((len - 1) >> i_blkbits)) > endblk)) + len = (endblk - iblock + 1) << i_blkbits; + + mlog(0, "get block of %llu at %llu:%u req %u\n", + inode->i_ino, pos, len, total_len); + + /* + * Because we need to change file size in ocfs2_iomap_dio_end_io_write(), + * or we may need to add it to orphan dir. So can not fall to fast path + * while file size will be changed. + */ + if (pos + total_len <= i_size_read(inode)) { + /* This is the fast path for re-write. */ + ret = ocfs2_lock_map_blocks(inode, map, OCFS2_GET_BLOCKS_CREATE); + if ((map->flags & OCFS2_MAP_MAPPED) && + !(map->flags & OCFS2_MAP_NEW) && + ret == 0) + goto out; + + /* Clear state set by ocfs2_map_blocks. */ + map->flags = 0; + } + + ret = ocfs2_inode_lock(inode, &di_bh, 1); + if (ret) { + mlog_errno(ret); + goto out; + } + + down_write(&oi->ip_alloc_sem); + + /* + * Note, Zeroing the gap between i_size and the write is done + * once per dio by ocfs2_dio_zero_eof() before the io is issued. + */ + ret = ocfs2_write_begin_nolock(inode->i_mapping, pos, len, + OCFS2_WRITE_DIRECT, NULL, + (void **)&wc, di_bh, NULL); + if (ret) { + mlog_errno(ret); + goto unlock; + } + + desc = &wc->w_desc[0]; + + p_blkno = ocfs2_clusters_to_blocks(inode->i_sb, desc->c_phys); + if (p_blkno == 0) { + ret = -EIO; + mlog_errno(ret); + goto unlock; + } + p_blkno += iblock & (u64)(ocfs2_clusters_to_blocks(inode->i_sb, 1) - 1); + + map->pblk = p_blkno; + /* + * Only report the mapping as UNWRITTEN when the cluster is newly + * allocated or was already unwritten (desc->c_needs_zero). + */ + map->flags |= desc->c_needs_zero ? + OCFS2_MAP_UNWRITTEN : OCFS2_MAP_MAPPED; + map->len = len >> i_blkbits; + + if (desc->c_needs_zero) + map->flags |= OCFS2_MAP_NEW; + + if (iblock > endblk) + map->flags |= OCFS2_MAP_NEW; + + ocfs2_free_unwritten_list(inode, &wc->w_unwritten_list); + /* Here returns only two types: len (success), errno (failure). */ + ret = ocfs2_write_end_nolock(inode->i_mapping, pos, len, len, wc); + if (ret < 0) { + mlog_errno(ret); + goto unlock; + } + WARN_ON_ONCE(ret != len); + ret = 0; + +unlock: + up_write(&oi->ip_alloc_sem); + ocfs2_inode_unlock(inode, 1); + brelse(di_bh); + +out: + return ret; +} + static int ocfs2_dio_end_io_write(struct inode *inode, struct ocfs2_dio_write_ctxt *dwc, loff_t offset, @@ -2542,6 +2707,7 @@ static int ocfs2_iomap_begin(struct inode *inode, loff_t offset, loff_t length, struct ocfs2_map_block map; struct ocfs2_inode_info *oi = OCFS2_I(inode); u8 blkbits = inode->i_blkbits; + unsigned int orig_mlen; if ((offset >> blkbits) > OCFS2_MAX_LOGICAL_BLOCK) return -EINVAL; @@ -2553,9 +2719,10 @@ static int ocfs2_iomap_begin(struct inode *inode, loff_t offset, loff_t length, map.len = min_t(loff_t, (offset + length - 1) >> blkbits, OCFS2_MAX_LOGICAL_BLOCK) - map.lblk + 1; map.flags = 0; + orig_mlen = map.len; if (flags & IOMAP_WRITE) { - return -EOPNOTSUPP; + ret = ocfs2_dio_wr_map_blocks(inode, &map, 1); } else { down_read(&oi->ip_alloc_sem); ret = ocfs2_map_blocks(inode, &map, 0); @@ -2605,6 +2772,217 @@ const struct iomap_ops ocfs2_iomap_ops = { .iomap_begin = ocfs2_iomap_begin, }; +/* copy from ocfs2_dio_end_io_write */ +static int ocfs2_iomap_dio_end_io_write(struct inode *inode, + loff_t offset, + ssize_t bytes) +{ + struct ocfs2_cached_dealloc_ctxt dealloc; + struct ocfs2_extent_tree et; + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct ocfs2_inode_info *oi = OCFS2_I(inode); + struct buffer_head *di_bh = NULL; + struct ocfs2_dinode *di; + struct ocfs2_alloc_context *data_ac = NULL; + struct ocfs2_alloc_context *meta_ac = NULL; + handle_t *handle = NULL; + loff_t end = offset + bytes; + int ret = 0, credits = 0, batch = 0; + struct ocfs2_map_block map; + unsigned int blkbits = inode->i_blkbits; + unsigned int ue_cpos = 0, ue_phys = 0, ue_len = 0; + unsigned int curr_lblk, end_lblk; + u32 extents_to_split; + + map.lblk = offset >> blkbits; + + ocfs2_init_dealloc_ctxt(&dealloc); + + ret = ocfs2_inode_lock(inode, &di_bh, 1); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + + down_write(&oi->ip_alloc_sem); + di = (struct ocfs2_dinode *)di_bh->b_data; + + ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh); + + /* Attach dealloc with extent tree in case that we may reuse extents + * which are already unlinked from current extent tree due to extent + * rotation and merging. + */ + et.et_dealloc = &dealloc; + + /* + * Currently, the slow path of ocfs2_dio_wr_map_blocks() only handles + * single-cluster cases. Because kiocb->private is cleared before the + * iomap_dio_ops end_io callback is invoked, this function cannot + * obtain the exact number of unwritten extents via dwc, unlike the + * buffered path function ocfs2_dio_end_io_write(). Consequently, the + * algorithm here may overestimate the actual number of unwritten + * extents; however, this only incurs additional allocator overhead and + * does not compromise the correctness of the results. + */ + extents_to_split = DIV_ROUND_UP(bytes, 1 << osb->s_clustersize_bits); + ret = ocfs2_lock_allocators(inode, &et, 0, extents_to_split * 2, + &data_ac, &meta_ac); + if (ret) { + mlog_errno(ret); + goto unlock; + } + + credits = ocfs2_calc_extend_credits(inode->i_sb, &di->id2.i_list); + + + curr_lblk = offset >> blkbits; + /* + * Round the end up so the final partial block (sub-block direct I/O) + * is included; otherwise the last, partially-written cluster is left + * unwritten and reads back as zero. + */ + end_lblk = (offset + bytes + (1 << blkbits) - 1) >> blkbits; + while (ret >= 0 && curr_lblk < end_lblk) { + memset(&map, 0, sizeof(map)); + map.lblk = curr_lblk; + map.len = end_lblk - curr_lblk; + + if (!handle) { + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto unlock; + } + ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, + OCFS2_JOURNAL_ACCESS_WRITE); + if (ret) { + mlog_errno(ret); + goto commit; + } + } + ret = ocfs2_assure_trans_credits(handle, credits); + if (ret < 0) { + mlog_errno(ret); + goto commit; + } + + ret = ocfs2_map_blocks(inode, &map, 0); + if (ret < 0) { + mlog_errno(ret); + goto commit; + } + if (map.len == 0) + goto commit; + + if (map.flags & OCFS2_MAP_UNWRITTEN) { + ue_cpos = ocfs2_blocks_to_clusters(inode->i_sb, map.lblk); + ue_phys = ocfs2_blocks_to_clusters(inode->i_sb, map.pblk); + ue_len = ocfs2_clusters_for_blocks(inode->i_sb, + map.lblk + map.len) - ue_cpos; + + ret = ocfs2_mark_extent_written(inode, &et, handle, + ue_cpos, ue_len, + ue_phys, + meta_ac, &dealloc); + if (ret < 0) { + mlog_errno(ret); + goto commit; + } + } + curr_lblk += map.len; + if (++batch == OCFS2_DIO_MARK_EXTENT_BATCH) { + ocfs2_commit_trans(osb, handle); + handle = NULL; + batch = 0; + } + } + + if (end > i_size_read(inode)) { + if (!handle) { + handle = ocfs2_start_trans(osb, credits); + if (IS_ERR(handle)) { + ret = PTR_ERR(handle); + mlog_errno(ret); + goto unlock; + } + } + ret = ocfs2_set_inode_size(handle, inode, di_bh, end); + if (ret < 0) + mlog_errno(ret); + } + +commit: + if (handle) + ocfs2_commit_trans(osb, handle); + +unlock: + up_write(&oi->ip_alloc_sem); + + if (data_ac) { + ocfs2_free_alloc_context(data_ac); + data_ac = NULL; + } + if (meta_ac) { + ocfs2_free_alloc_context(meta_ac); + meta_ac = NULL; + } + + /* + * The orphan entry is removed by ocfs2_file_write_iter(), by the writer + * that added it. Other dios on this inode may complete here while it + * is still in flight, so they must not touch it. + */ + ocfs2_inode_unlock(inode, 1); + brelse(di_bh); + +out: + ocfs2_run_deallocs(osb, &dealloc); + return ret; +} + +/* See the comments from ocfs2_iomap_dio_end_io_r_pr(). */ +static int ocfs2_dio_write_end_io(struct kiocb *iocb, ssize_t size, + int error, unsigned int flags, int level) +{ + struct inode *inode = file_inode(iocb->ki_filp); + loff_t offset = iocb->ki_pos; + int ret = 0; + + if (error) + mlog_ratelimited(ML_ERROR, "Direct IO failed, bytes = %lld errno:%d", + (long long)size, error); + + /* bounced back to buffered I/O, leave the rw_lock held */ + if (size == 0 && error == 0) + return 0; + + if (size && error == 0 && + ((flags & IOMAP_DIO_UNWRITTEN) || + (offset + size > i_size_read(inode)))) { + ret = ocfs2_iomap_dio_end_io_write(inode, offset, size); + if (ret < 0) + mlog_errno(ret); + } + + ocfs2_rw_unlock(inode, level); + + return ret ? ret : error; +} + +static int ocfs2_iomap_dio_end_io_w_pr(struct kiocb *iocb, ssize_t size, + int error, unsigned int flags) +{ + return ocfs2_dio_write_end_io(iocb, size, error, flags, 0); +} + +static int ocfs2_iomap_dio_end_io_w_ex(struct kiocb *iocb, ssize_t size, + int error, unsigned int flags) +{ + return ocfs2_dio_write_end_io(iocb, size, error, flags, 1); +} + /* * Direct I/O completion. Like the old ocfs2_dio_end_io(), we use the rw_lock * DLM lock to protect io on one node from truncation on another, and drop it @@ -2612,38 +2990,40 @@ const struct iomap_ops ocfs2_iomap_ops = { * including on the buffered-fallback (-ENOTBLK) path, so the rw_lock taken in * ocfs2_file_{read,write}_iter() is released exactly once. */ -static int ocfs2_iomap_dio_end_io(struct kiocb *iocb, ssize_t size, int error, - int level) +static int ocfs2_iomap_dio_end_io_r_pr(struct kiocb *iocb, ssize_t size, + int error, unsigned int flags) { struct inode *inode = file_inode(iocb->ki_filp); + if (error) + mlog_ratelimited(ML_ERROR, "Direct IO failed, bytes = %lld errno:%d", + (long long)size, error); + /* - * When iomap_dio_complete() calls us with size==0 and error==0, it - * means the mapping bounced back to buffered I/O. In that case leave - * the rw_lock ocfs2_file_{read,write}_iter() keeps it for the buffered - * retry and releases it in its own cleanup path. + * iomap_dio_complete() calls us even when no I/O was issued because the + * mapping bounced us back to buffered I/O (reported as size == 0 and + * error == 0). In that case leave the rw_lock held so + * ocfs2_file_{read,write}_iter() keeps it for the buffered retry and + * releases it in its own cleanup path. */ if (size == 0 && error == 0) return 0; - ocfs2_rw_unlock(inode, level); + ocfs2_rw_unlock(inode, 0); + return error; } -static int ocfs2_dio_end_io_r_pr(struct kiocb *iocb, ssize_t size, - int error, unsigned int flags) -{ - if (error) - mlog_ratelimited(ML_ERROR, "Direct IO failed, bytes = %lld errno:%d", - (long long)size, error); - - error = ocfs2_iomap_dio_end_io(iocb, size, error, 0); +const struct iomap_dio_ops ocfs2_iomap_dio_ops_w_pr = { + .end_io = ocfs2_iomap_dio_end_io_w_pr, +}; - return error; -} +const struct iomap_dio_ops ocfs2_iomap_dio_ops_w_ex = { + .end_io = ocfs2_iomap_dio_end_io_w_ex, +}; const struct iomap_dio_ops ocfs2_iomap_dio_ops_r_pr = { - .end_io = ocfs2_dio_end_io_r_pr, + .end_io = ocfs2_iomap_dio_end_io_r_pr, }; const struct address_space_operations ocfs2_aops = { .dirty_folio = block_dirty_folio, diff --git a/fs/ocfs2/aops.h b/fs/ocfs2/aops.h index 2fe3b39e02f7..3bfce26ac3ab 100644 --- a/fs/ocfs2/aops.h +++ b/fs/ocfs2/aops.h @@ -11,6 +11,8 @@ extern const struct iomap_ops ocfs2_iomap_ops; extern const struct iomap_dio_ops ocfs2_iomap_dio_ops_r_pr; +extern const struct iomap_dio_ops ocfs2_iomap_dio_ops_w_pr; +extern const struct iomap_dio_ops ocfs2_iomap_dio_ops_w_ex; int ocfs2_map_folio_blocks(struct folio *folio, u64 *p_blkno, struct inode *inode, unsigned int from, @@ -46,6 +48,7 @@ int ocfs2_size_fits_inline_data(struct buffer_head *di_bh, u64 new_size); int ocfs2_get_block(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create); +int ocfs2_dio_zero_eof(struct inode *inode, loff_t pos, size_t count); int ocfs2_map_blocks(struct inode *inode, struct ocfs2_map_block *map, int flags); /* all ocfs2_dio_end_io()'s fault */ diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c index a62cb68f246b..f92172673d87 100644 --- a/fs/ocfs2/file.c +++ b/fs/ocfs2/file.c @@ -48,6 +48,7 @@ #include "quota.h" #include "refcounttree.h" #include "ocfs2_trace.h" +#include "namei.h" #include "buffer_head_io.h" @@ -2403,6 +2404,8 @@ static int ocfs2_prepare_inode_for_write(struct file *file, static bool ocfs2_should_use_dio(struct kiocb *iocb, struct iov_iter *iter, struct inode *inode) { + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + /* * Fallback to buffered I/O if we see an inode without * extents. @@ -2410,9 +2413,43 @@ static bool ocfs2_should_use_dio(struct kiocb *iocb, struct iov_iter *iter, if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) return false; + /* Fallback to buffered I/O if we do not support append dio. */ + if (iocb->ki_pos + iter->count > i_size_read(inode) && + !ocfs2_supports_append_dio(osb)) + return false; + return true; } +/* + * Remove the inode from the orphan dir after an appending direct write. + * Only the writer that added the inode (under i_rwsem) may remove it; + */ +static int ocfs2_dio_del_orphan(struct inode *inode) +{ + struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); + struct buffer_head *di_bh = NULL; + struct ocfs2_dinode *di; + int ret; + + ret = ocfs2_inode_lock(inode, &di_bh, 1); + if (ret < 0) { + mlog_errno(ret); + return ret; + } + + di = (struct ocfs2_dinode *)di_bh->b_data; + if (di->i_flags & cpu_to_le32(OCFS2_DIO_ORPHANED_FL)) { + ret = ocfs2_del_inode_from_orphan(osb, inode, di_bh, 0, 0); + if (ret < 0) + mlog_errno(ret); + } + + ocfs2_inode_unlock(inode, 1); + brelse(di_bh); + return ret; +} + static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, struct iov_iter *from) { @@ -2430,6 +2467,8 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, i_size_read(inode) ? 1 : 0); int direct_io = iocb->ki_flags & IOCB_DIRECT ? 1 : 0; int nowait = iocb->ki_flags & IOCB_NOWAIT ? 1 : 0; + int dio_flags = 0; + int orphaned = 0; if (trace_ocfs2_file_write_iter_enabled()) { struct dentry *dentry = file->f_path.dentry; @@ -2455,8 +2494,6 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, } else inode_lock(inode); - ocfs2_iocb_init_rw_locked(iocb); - /* * Concurrent O_DIRECT writes are allowed with * mount_option "coherency=buffered". @@ -2512,6 +2549,10 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, goto out; } + ret = kiocb_modified(iocb); + if (ret < 0) + goto out; + if (direct_io && !is_sync_kiocb(iocb) && ocfs2_is_io_unaligned(inode, count, iocb->ki_pos)) { /* @@ -2520,26 +2561,79 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, saved_ki_complete = xchg(&iocb->ki_complete, NULL); } - /* communicate with ocfs2_dio_end_io */ - ocfs2_iocb_set_rw_locked(iocb, rw_level); + if (direct_io && ocfs2_should_use_dio(iocb, from, inode)) { + if (ocfs2_clusters_for_bytes(inode->i_sb, iocb->ki_pos + count) > + ocfs2_clusters_for_bytes(inode->i_sb, i_size_read(inode))) { + /* + * when we are going to alloc extents beyond file size, add the + * inode to orphan dir, so we can recall those spaces when + * system crashed during write. + */ + ret = ocfs2_add_inode_to_orphan(osb, inode); + if (ret < 0) { + mlog_errno(ret); + goto out; + } + orphaned = 1; + dio_flags = IOMAP_DIO_FORCE_WAIT; + } - written = __generic_file_write_iter(iocb, from); - /* buffered aio wouldn't have proper lock coverage today */ - BUG_ON(written == -EIOCBQUEUED && !direct_io); + ret = ocfs2_dio_zero_eof(inode, iocb->ki_pos, count); + if (ret < 0) + goto out; - /* - * deep in g_f_a_w_n()->ocfs2_direct_IO we pass in a ocfs2_dio_end_io - * function pointer which is called when o_direct io completes so that - * it can unlock our rw lock. - * Unfortunately there are error cases which call end_io and others - * that don't. so we don't have to unlock the rw_lock if either an - * async dio is going to do it in the future or an end_io after an - * error has already done it. - */ - if ((written == -EIOCBQUEUED) || (!ocfs2_iocb_is_rw_locked(iocb))) { - rw_level = -1; + const struct iomap_dio_ops *dops = rw_level ? + &ocfs2_iomap_dio_ops_w_ex : &ocfs2_iomap_dio_ops_w_pr; + struct iomap_dio *dio; + + dio = __iomap_dio_rw(iocb, from, &ocfs2_iomap_ops, + dops, dio_flags, NULL, 0); + if (dio == NULL) { + /* No I/O issued; rw_lock still held. */ + written = 0; + } else if (IS_ERR(dio)) { + written = PTR_ERR(dio); + /* + * -EIOCBQUEUED: async I/O; end_io will drop the lock on + * completion. Any other error means no I/O started and + * end_io was not called, so the cleanup path below drops + * the lock. + */ + if (written == -EIOCBQUEUED) + rw_level = -1; + } else { + /* + * Synchronous completion. iomap_dio_complete() runs + * end_io, which drops the lock unless the mapping + * bounced us back to buffered I/O (returns 0, lock + * retained). + */ + written = iomap_dio_complete(dio); + if (written != 0) + rw_level = -1; + } + + /* + * __iomap_dio_rw() returns -ENOTBLK when it could not invalidate + * the page cache for the DIO range (e.g. racing buffered/mmap + * I/O on the same file, as exercised by generic/095). Fall back + * to a buffered write for the remaining data, the same way + * ext4/xfs do, instead of leaking -ENOTBLK to userspace. + */ + if (written == -ENOTBLK) + written = 0; + if (written == 0 && iov_iter_count(from)) { + direct_write_fallback(iocb, from, ret, + generic_perform_write(iocb, from)); + } + } else { + iocb->ki_flags &= ~IOCB_DIRECT; + written = __generic_file_write_iter(iocb, from); } + /* buffered aio wouldn't have proper lock coverage today */ + BUG_ON(written == -EIOCBQUEUED && !direct_io); + if (unlikely(written <= 0)) goto out; @@ -2564,6 +2658,13 @@ static ssize_t ocfs2_file_write_iter(struct kiocb *iocb, } out: + /* + * IOMAP_DIO_FORCE_WAIT made the dio complete synchronously, so its + * end_io has already updated i_size. + */ + if (orphaned) + ocfs2_dio_del_orphan(inode); + if (saved_ki_complete) xchg(&iocb->ki_complete, saved_ki_complete); -- 2.55.0