* [PATCH][v2] btrfs-progs: add a --check-bg-usage option to fsck
@ 2019-08-02 16:09 Josef Bacik
2019-10-01 18:30 ` David Sterba
0 siblings, 1 reply; 2+ messages in thread
From: Josef Bacik @ 2019-08-02 16:09 UTC (permalink / raw)
To: linux-btrfs, kernel-team
Sometimes when messing with the chunk allocator code we can end up
over-allocating chunks. Generally speaking I'll notice this when a
random xfstest fails with ENOSPC when it shouldn't, but I'm super
worried that I won't catch a problem until somebody has a fs completely
filled up with empty block groups. Add a fsck option to check for too
many empty block groups. This way I can set FSCK_OPTIONS="-B" to catch
cases where we're too aggressive with the chunk allocator but not so
aggressive that it causes problems in xfstests.
Thankfully this doesn't trip up currently, so this will just keep me
from regressing us. Thanks,
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
---
v1->v2:
- tested with my alloc chunk ioctl, realized the chunk checker removes the bg
recs from the list, so this wasn't actually doing anything. Moved the check
so now it properly fails on a bad fs.
btrfsck.h | 1 +
check/main.c | 53 ++++++++++++++++++++++++++++++++++++++++++++++++++--
2 files changed, 52 insertions(+), 2 deletions(-)
diff --git a/btrfsck.h b/btrfsck.h
index ac7f5d48..5e779075 100644
--- a/btrfsck.h
+++ b/btrfsck.h
@@ -44,6 +44,7 @@ struct block_group_record {
u64 offset;
u64 flags;
+ u64 used;
};
struct block_group_tree {
diff --git a/check/main.c b/check/main.c
index 0cc6fdba..ca2ace10 100644
--- a/check/main.c
+++ b/check/main.c
@@ -62,6 +62,7 @@ int no_holes = 0;
static int is_free_space_tree = 0;
int init_extent_tree = 0;
int check_data_csum = 0;
+int check_bg_usage = 0;
struct btrfs_fs_info *global_info;
struct task_ctx ctx = { 0 };
struct cache_tree *roots_info_cache = NULL;
@@ -5126,6 +5127,7 @@ btrfs_new_block_group_record(struct extent_buffer *leaf, struct btrfs_key *key,
ptr = btrfs_item_ptr(leaf, slot, struct btrfs_block_group_item);
rec->flags = btrfs_disk_block_group_flags(leaf, ptr);
+ rec->used = btrfs_disk_block_group_used(leaf, ptr);
INIT_LIST_HEAD(&rec->list);
@@ -8522,6 +8524,41 @@ out:
return ret;
}
+static int check_block_group_usage(struct block_group_tree *block_group_cache)
+{
+ struct block_group_record *bg_rec;
+ int empty_data = 0, empty_metadata = 0, empty_system = 0;
+ int ret = 0;
+
+ list_for_each_entry(bg_rec, &block_group_cache->block_groups, list) {
+ if (bg_rec->used)
+ continue;
+ if (bg_rec->flags & BTRFS_BLOCK_GROUP_DATA)
+ empty_data++;
+ else if (bg_rec->flags & BTRFS_BLOCK_GROUP_METADATA)
+ empty_metadata++;
+ else
+ empty_system++;
+ }
+
+ if (empty_data > 1) {
+ ret = -EINVAL;
+ fprintf(stderr, "Too many empty data block groups: %d\n",
+ empty_data);
+ }
+ if (empty_metadata > 1) {
+ ret = -EINVAL;
+ fprintf(stderr, "Too many empty metadata block groups: %d\n",
+ empty_metadata);
+ }
+ if (empty_system > 1) {
+ ret = -EINVAL;
+ fprintf(stderr, "Too many empty system block groups: %d\n",
+ empty_system);
+ }
+ return ret;
+}
+
static int check_chunks_and_extents(struct btrfs_fs_info *fs_info)
{
struct rb_root dev_cache;
@@ -8622,6 +8659,13 @@ again:
goto out;
}
+ if (check_bg_usage) {
+ ret = check_block_group_usage(&block_group_cache);
+ if (ret)
+ err = ret;
+ goto out;
+ }
+
ret = check_chunks(&chunk_cache, &block_group_cache,
&dev_extent_cache, NULL, NULL, NULL, 0);
if (ret) {
@@ -9810,6 +9854,7 @@ static const char * const cmd_check_usage[] = {
" -E|--subvol-extents <subvolid>",
" print subvolume extents and sharing state",
" -p|--progress indicate progress",
+ " -B|--check-bg-usage check for too many empty block groups",
NULL
};
@@ -9841,7 +9886,7 @@ static int cmd_check(const struct cmd_struct *cmd, int argc, char **argv)
GETOPT_VAL_INIT_EXTENT, GETOPT_VAL_CHECK_CSUM,
GETOPT_VAL_READONLY, GETOPT_VAL_CHUNK_TREE,
GETOPT_VAL_MODE, GETOPT_VAL_CLEAR_SPACE_CACHE,
- GETOPT_VAL_FORCE };
+ GETOPT_VAL_FORCE};
static const struct option long_options[] = {
{ "super", required_argument, NULL, 's' },
{ "repair", no_argument, NULL, GETOPT_VAL_REPAIR },
@@ -9864,10 +9909,11 @@ static int cmd_check(const struct cmd_struct *cmd, int argc, char **argv)
{ "clear-space-cache", required_argument, NULL,
GETOPT_VAL_CLEAR_SPACE_CACHE},
{ "force", no_argument, NULL, GETOPT_VAL_FORCE },
+ { "check-bg-usage", no_argument, NULL, 'B' },
{ NULL, 0, NULL, 0}
};
- c = getopt_long(argc, argv, "as:br:pEQ", long_options, NULL);
+ c = getopt_long(argc, argv, "as:br:pEQB", long_options, NULL);
if (c < 0)
break;
switch(c) {
@@ -9875,6 +9921,9 @@ static int cmd_check(const struct cmd_struct *cmd, int argc, char **argv)
case 'b':
ctree_flags |= OPEN_CTREE_BACKUP_ROOT;
break;
+ case 'B':
+ check_bg_usage = 1;
+ break;
case 's':
num = arg_strtou64(optarg);
if (num >= BTRFS_SUPER_MIRROR_MAX) {
--
2.21.0
^ permalink raw reply related [flat|nested] 2+ messages in thread* Re: [PATCH][v2] btrfs-progs: add a --check-bg-usage option to fsck
2019-08-02 16:09 [PATCH][v2] btrfs-progs: add a --check-bg-usage option to fsck Josef Bacik
@ 2019-10-01 18:30 ` David Sterba
0 siblings, 0 replies; 2+ messages in thread
From: David Sterba @ 2019-10-01 18:30 UTC (permalink / raw)
To: Josef Bacik; +Cc: linux-btrfs, kernel-team
On Fri, Aug 02, 2019 at 12:09:53PM -0400, Josef Bacik wrote:
> Sometimes when messing with the chunk allocator code we can end up
> over-allocating chunks. Generally speaking I'll notice this when a
> random xfstest fails with ENOSPC when it shouldn't, but I'm super
> worried that I won't catch a problem until somebody has a fs completely
> filled up with empty block groups. Add a fsck option to check for too
> many empty block groups. This way I can set FSCK_OPTIONS="-B" to catch
> cases where we're too aggressive with the chunk allocator but not so
> aggressive that it causes problems in xfstests.
>
> Thankfully this doesn't trip up currently, so this will just keep me
> from regressing us. Thanks,
>
> Signed-off-by: Josef Bacik <josef@toxicpanda.com>
> ---
> v1->v2:
> - tested with my alloc chunk ioctl, realized the chunk checker removes the bg
> recs from the list, so this wasn't actually doing anything. Moved the check
> so now it properly fails on a bad fs.
I sent some comments to v1 that seem to apply to v2 as well.
> static int is_free_space_tree = 0;
> int init_extent_tree = 0;
> int check_data_csum = 0;
> +int check_bg_usage = 0;
> struct btrfs_fs_info *global_info;
> struct task_ctx ctx = { 0 };
> struct cache_tree *roots_info_cache = NULL;
> @@ -5126,6 +5127,7 @@ btrfs_new_block_group_record(struct extent_buffer *leaf, struct btrfs_key *key,
>
> ptr = btrfs_item_ptr(leaf, slot, struct btrfs_block_group_item);
> rec->flags = btrfs_disk_block_group_flags(leaf, ptr);
> + rec->used = btrfs_disk_block_group_used(leaf, ptr);
>
> INIT_LIST_HEAD(&rec->list);
>
> @@ -8522,6 +8524,41 @@ out:
> return ret;
> }
>
> +static int check_block_group_usage(struct block_group_tree *block_group_cache)
> +{
> + struct block_group_record *bg_rec;
> + int empty_data = 0, empty_metadata = 0, empty_system = 0;
> + int ret = 0;
> +
> + list_for_each_entry(bg_rec, &block_group_cache->block_groups, list) {
> + if (bg_rec->used)
> + continue;
> + if (bg_rec->flags & BTRFS_BLOCK_GROUP_DATA)
> + empty_data++;
> + else if (bg_rec->flags & BTRFS_BLOCK_GROUP_METADATA)
> + empty_metadata++;
> + else
> + empty_system++;
> + }
> +
> + if (empty_data > 1) {
> + ret = -EINVAL;
> + fprintf(stderr, "Too many empty data block groups: %d\n",
> + empty_data);
> + }
> + if (empty_metadata > 1) {
> + ret = -EINVAL;
> + fprintf(stderr, "Too many empty metadata block groups: %d\n",
> + empty_metadata);
> + }
> + if (empty_system > 1) {
> + ret = -EINVAL;
> + fprintf(stderr, "Too many empty system block groups: %d\n",
> + empty_system);
Can you please add images that trigger each of the above problems?
> + }
> + return ret;
> +}
> +
> static int check_chunks_and_extents(struct btrfs_fs_info *fs_info)
> {
> struct rb_root dev_cache;
> @@ -8622,6 +8659,13 @@ again:
> goto out;
> }
>
> + if (check_bg_usage) {
> + ret = check_block_group_usage(&block_group_cache);
> + if (ret)
> + err = ret;
> + goto out;
> + }
> +
> ret = check_chunks(&chunk_cache, &block_group_cache,
> &dev_extent_cache, NULL, NULL, NULL, 0);
> if (ret) {
> @@ -9810,6 +9854,7 @@ static const char * const cmd_check_usage[] = {
> " -E|--subvol-extents <subvolid>",
> " print subvolume extents and sharing state",
> " -p|--progress indicate progress",
> + " -B|--check-bg-usage check for too many empty block groups",
> NULL
> };
>
> @@ -9841,7 +9886,7 @@ static int cmd_check(const struct cmd_struct *cmd, int argc, char **argv)
> GETOPT_VAL_INIT_EXTENT, GETOPT_VAL_CHECK_CSUM,
> GETOPT_VAL_READONLY, GETOPT_VAL_CHUNK_TREE,
> GETOPT_VAL_MODE, GETOPT_VAL_CLEAR_SPACE_CACHE,
> - GETOPT_VAL_FORCE };
> + GETOPT_VAL_FORCE};
> static const struct option long_options[] = {
> { "super", required_argument, NULL, 's' },
> { "repair", no_argument, NULL, GETOPT_VAL_REPAIR },
> @@ -9864,10 +9909,11 @@ static int cmd_check(const struct cmd_struct *cmd, int argc, char **argv)
> { "clear-space-cache", required_argument, NULL,
> GETOPT_VAL_CLEAR_SPACE_CACHE},
> { "force", no_argument, NULL, GETOPT_VAL_FORCE },
> + { "check-bg-usage", no_argument, NULL, 'B' },
The option name does not match the description, it could be something
like --check-empty-bg-count and certainly should not use the short
option. That are reserved for most common usecases, I don't think
this is the case.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2019-10-01 18:30 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2019-08-02 16:09 [PATCH][v2] btrfs-progs: add a --check-bg-usage option to fsck Josef Bacik
2019-10-01 18:30 ` David Sterba
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).