From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail.cn.fujitsu.com ([183.91.158.132]:52830 "EHLO heian.cn.fujitsu.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1751274AbeD3D2J (ORCPT ); Sun, 29 Apr 2018 23:28:09 -0400 Subject: Re: [PATCH 2/3] btrfs-progs: Allow tree to be printed without an fs_info To: Qu Wenruo , References: <20180430031545.29891-1-wqu@suse.com> <20180430031545.29891-2-wqu@suse.com> From: Su Yue Message-ID: <2ed59965-c7be-ca53-caca-29e2a5a96958@cn.fujitsu.com> Date: Mon, 30 Apr 2018 11:33:56 +0800 MIME-Version: 1.0 In-Reply-To: <20180430031545.29891-2-wqu@suse.com> Content-Type: text/plain; charset="utf-8"; format=flowed Sender: linux-btrfs-owner@vger.kernel.org List-ID: On 04/30/2018 11:15 AM, Qu Wenruo wrote: > For btrfs_print_tree() and btrfs_print_leaf(), the usage of fs_info is > mainly for nodesize and sectorsize. > > However for nodesize, we can get it from @eb->len without the need for > fs_info at all. > > For nodesize, introduce new helper BTRFS_NODEPTR_PER_EXTENT_BUFFER() to > get nodesize from @eb directly. > And with the help of previous modified btrfs_leaf_free_space(), > btrfs_print_tree() can live without fs_info at all. > > For btrfs_print_leaf(), we modify print_extent_csum() to accept NULL > fs_info by skipping csum length calculation. > > With all these modification, btrfs_print_tree/leaf() can be called > without accessing @fs_info at all, and make it more flex to handle > binary tree block dump, or inside gdb. > > Signed-off-by: Qu Wenruo Reviewed-by: Su Yue > --- > ctree.h | 7 +++++++ > print-tree.c | 15 +++++++++++++-- > 2 files changed, 20 insertions(+), 2 deletions(-) > > diff --git a/ctree.h b/ctree.h > index e291a7d48d14..80fd19612434 100644 > --- a/ctree.h > +++ b/ctree.h > @@ -1202,6 +1202,13 @@ static inline u32 BTRFS_NODEPTRS_PER_BLOCK(const struct btrfs_fs_info *info) > return BTRFS_LEAF_DATA_SIZE(info) / sizeof(struct btrfs_key_ptr); > } > > +static inline u32 > +BTRFS_NODEPTRS_PER_EXTENT_BUFFER(const struct extent_buffer *eb) > +{ > + BUG_ON(eb->fs_info && eb->fs_info->nodesize != eb->len); > + return __BTRFS_LEAF_DATA_SIZE(eb->len) / sizeof(struct btrfs_key_ptr); > +} > + > #define BTRFS_FILE_EXTENT_INLINE_DATA_START \ > (offsetof(struct btrfs_file_extent_item, disk_bytenr)) > static inline u32 BTRFS_MAX_INLINE_DATA_SIZE(const struct btrfs_fs_info *info) > diff --git a/print-tree.c b/print-tree.c > index 34a724d5f7ef..31a851ef4413 100644 > --- a/print-tree.c > +++ b/print-tree.c > @@ -1150,6 +1150,14 @@ static void print_extent_csum(struct extent_buffer *eb, > { > u32 size; > > + /* > + * If we don't have fs_info, only output its start position as we > + * don't have sectorsize for the calculation > + */ > + if (!fs_info) { > + printf("\t\trange start %llu\n", (unsigned long long)start); > + return; > + } > size = (item_size / btrfs_super_csum_size(fs_info->super_copy)) * > fs_info->sectorsize; > printf("\t\trange start %llu end %llu length %u\n", > @@ -1371,7 +1379,7 @@ void btrfs_print_tree(struct extent_buffer *eb, int follow) > printf("node %llu level %d items %d free %u generation %llu owner ", > (unsigned long long)eb->start, > btrfs_header_level(eb), nr, > - (u32)BTRFS_NODEPTRS_PER_BLOCK(fs_info) - nr, > + (u32)BTRFS_NODEPTRS_PER_EXTENT_BUFFER(eb) - nr, > (unsigned long long)btrfs_header_generation(eb)); > print_objectid(stdout, btrfs_header_owner(eb), 0); > printf("\n"); > @@ -1385,13 +1393,16 @@ void btrfs_print_tree(struct extent_buffer *eb, int follow) > btrfs_print_key(&disk_key); > printf(" block %llu (%llu) gen %llu\n", > (unsigned long long)blocknr, > - (unsigned long long)blocknr / fs_info->nodesize, > + (unsigned long long)blocknr / eb->len, > (unsigned long long)btrfs_node_ptr_generation(eb, i)); > fflush(stdout); > } > if (!follow) > return; > > + if (follow && !fs_info) > + return; > + > for (i = 0; i < nr; i++) { > next = read_tree_block(fs_info, > btrfs_node_blockptr(eb, i), >