| Message ID | 20171010075113.10718-2-quwenruo.btrfs@gmx.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 10.10.2017 10:51, Qu Wenruo wrote: > This patch introduces functions to repair device size related problems, > including: > 1) Unaligned total_bytes of dev_item > v4.14-rc kernel introduced total_bytes alignment checker. > However older kernel device add/shrink doesn't align these members. > This will cause kernel warning every time dev_item get updated. > > Although it can be fixed by shrinking device on latest kernel or > use manually aligned size on older kernel, a fallback method in > btrfs-progs won't hurt. > > 2) Mismatch super->total_bytes > There are some reports about unmountable fs, due to mismatched > super->total_bytes. > And normal kernel device shrink won't help since it only modify the > total_bytes by the size changed, not re-calculating it. > > The root cause is still under investigation, but at least to fix it > in btrfs-progs > > Fix all of them by manually rounding down total_bytes of each device, and > recalculate super->total_bytes using all existing devices. Thanks for doing this, we needed it. One minor nit in the comments but overall: Reviewed-by: Nikolay Borisov <nborisov@suse.com> > > Reported-by: Asif Youssuff <yoasif@gmail.com> > Reported-by: Rich Rauenzahn <rrauenza@gmail.com> > Signed-off-by: Qu Wenruo <quwenruo.btrfs@gmx.com>> --- > cmds-check.c | 191 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ > 1 file changed, 191 insertions(+) > > diff --git a/cmds-check.c b/cmds-check.c > index 0c08618ed701..007781fa5d1b 100644 > --- a/cmds-check.c > +++ b/cmds-check.c > @@ -12885,6 +12885,197 @@ close_out: > return ret; > } > > +/* > + * Return 0 if DEV_ITEM for @device is good > + * Return >0 if DEV_ITEM for @device has unaligned value and fixed > + * Return <0 if we failed to fix the unaligned value > + */ > +static int reset_one_dev_size(struct btrfs_fs_info *fs_info, > + struct btrfs_device *device) > +{ > + struct btrfs_trans_handle *trans; > + struct btrfs_key key; > + struct btrfs_path path; > + struct btrfs_root *chunk_root = fs_info->chunk_root; > + struct btrfs_dev_item *di; > + u64 old_bytes = device->total_bytes; > + int ret; > + > + if (IS_ALIGNED(device->total_bytes, fs_info->sectorsize)) > + return 0; > + > + /* Align the in-memory total_bytes first, and use it later */ > + device->total_bytes = round_down(device->total_bytes, > + fs_info->sectorsize); > + > + key.objectid = BTRFS_DEV_ITEMS_OBJECTID; > + key.type = BTRFS_DEV_ITEM_KEY; > + key.offset = device->devid; > + > + trans = btrfs_start_transaction(chunk_root, 1); > + if (IS_ERR(trans)) { > + ret = PTR_ERR(trans); > + error("error starting transaction: %d (%s)", > + ret, strerror(-ret)); > + return ret; > + } > + > + btrfs_init_path(&path); > + ret = btrfs_search_slot(trans, chunk_root, &key, &path, 0, 1); > + if (ret > 0) { > + error("failed to find DEV_ITEM for devid %llu", > + device->devid); > + ret = -ENOENT; > + goto err; > + } > + if (ret < 0) { > + error("failed to search chunk root: %d (%s)", > + ret, strerror(-ret)); > + goto err; > + } > + di = btrfs_item_ptr(path.nodes[0], path.slots[0], > + struct btrfs_dev_item); > + btrfs_set_device_total_bytes(path.nodes[0], di, device->total_bytes); > + btrfs_mark_buffer_dirty(path.nodes[0]); > + ret = btrfs_commit_transaction(trans, chunk_root); > + if (ret < 0) { > + error("failed to commit current transaction: %d (%s)", > + ret, strerror(-ret)); > + btrfs_release_path(&path); > + return ret; > + } > + btrfs_release_path(&path); > + printf("Fixed device size for devid %llu, old size: %llu new size: %llu\n", > + device->devid, old_bytes, device->total_bytes); > + return 1; > + > +err: > + /* We haven't modified anything, it's OK to commit current trans */ > + btrfs_commit_transaction(trans, chunk_root); > + btrfs_release_path(&path); > + return ret; > +} > + > +/* > + * Return 0 if super block total bytes matches with device total_bytes > + * Return >0 if super block has mismatch total_bytes and fixed > + * Return <0 if we failed to fix the mismatch total_bytes > + */ > +static int reset_super_total_bytes(struct btrfs_fs_info *fs_info) > +{ > + struct btrfs_trans_handle *trans; > + struct btrfs_device *device; > + struct list_head *dev_list = &fs_info->fs_devices->devices; > + u64 total_bytes = 0; > + u64 old_bytes = btrfs_super_total_bytes(fs_info->super_copy); > + int ret; > + > + list_for_each_entry(device, dev_list, dev_list) > + total_bytes += device->total_bytes; > + > + if (total_bytes == old_bytes) > + return 0; > + > + /* Just in case */ > + if (!IS_ALIGNED(total_bytes, fs_info->sectorsize)) { > + error("final total_bytes still not aligned to %u, please report a bug to btrfs mail list", > + fs_info->sectorsize); > + return -EUCLEAN; > + } > + > + btrfs_set_super_total_bytes(fs_info->super_copy, total_bytes); > + > + /* Commit transaction to update all super blocks */ > + trans = btrfs_start_transaction(fs_info->tree_root, 1); > + if (IS_ERR(trans)) { > + ret = PTR_ERR(trans); > + error("error starting transaction: %d (%s)", > + ret, strerror(-ret)); > + return ret; > + } > + ret = btrfs_commit_transaction(trans, fs_info->tree_root); > + if (ret < 0) { > + error("failed to commit current transaction: %d (%s)", > + ret, strerror(-ret)); > + return ret; > + } > + printf("Fixed super total bytes, old size: %llu new size: %llu\n", > + old_bytes, total_bytes); > + return 1; > +} > + > +/* > + * Repair device size related problems, including: > + * 1) Unaligned total_bytes of dev_item > + * v4.14-rc kernel introduced total_bytes alignment checker. > + * However older kernel device add/shrink doesn't restrictly align > + * these members. > + * This will cause kernel warning everytime dev_item get updated. > + * > + * Although it can be fixed by shrinking device on latest kernel or > + * use manually aligned size on older kernel, a fallback method in > + * btrfs-progs won't hurt. > + * > + * 2) Mismatch super->total_bytes > + * Due to similar situation, super->total_bytes can be mismatch with > + * total size of all devices. > + * This will cause the filesystem unable to be mounted with v4.14-rc kernel. The commit which introduced the code preventing mounting can be traced back to 4.7: 99e3ecfcb9f4 ("Btrfs: add more validation checks for superblock"). I'd say remove the version information or fix it to point to v4.7. > + * > + * Add such fixer in btrfs-progs as a fallback method. > + * > + * Fix all of them by manually rounding down total_bytes of each device, and > + * recalculate super->total_bytes using all existing devices. > + */ > +static int reset_devs_size(struct btrfs_fs_info *fs_info) > +{ > + struct btrfs_device *device; > + struct list_head *dev_list = &fs_info->fs_devices->devices; > + bool have_bad_value = false; > + int ret; > + > + /* Seed device is not support yet */ > + if (fs_info->fs_devices->seed) { > + error("resetting device size with seed device is not supported yet"); > + return -ENOTTY; > + } > + > + /* All devices must be on-line before reparing */ > + if (list_empty(dev_list)) { > + error("no device found"); > + return -ENODEV; > + } > + list_for_each_entry(device, dev_list, dev_list) { > + if (device->fd == -1 || !device->writeable) { > + error("device with devid %llu is missing or not writeable", > + device->devid); > + error("resetting device size needs all device(s) present and writeable"); > + return -ENODEV; > + } > + } > + > + /* Repair total_bytes of each device */ > + list_for_each_entry(device, dev_list, dev_list) { > + ret = reset_one_dev_size(fs_info, device); > + if (ret < 0) > + return ret; > + if (ret > 0) > + have_bad_value = true; > + } > + > + /* Repair super total_byte */ > + ret = reset_super_total_bytes(fs_info); > + if (ret > 0) > + have_bad_value = true; > + if (have_bad_value) { > + printf("Fixed unaligned/mismatch total_bytes for superblock and device item\n"); > + ret = 1; > + } else { > + printf("No device size related problem found\n"); > + ret = 0; > + } > + return ret; > +} > + > const char * const cmd_check_usage[] = { > "btrfs check [options] <device>", > "Check structural integrity of a filesystem (unmounted).", > -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" 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/cmds-check.c b/cmds-check.c index 0c08618ed701..007781fa5d1b 100644 --- a/cmds-check.c +++ b/cmds-check.c @@ -12885,6 +12885,197 @@ close_out: return ret; } +/* + * Return 0 if DEV_ITEM for @device is good + * Return >0 if DEV_ITEM for @device has unaligned value and fixed + * Return <0 if we failed to fix the unaligned value + */ +static int reset_one_dev_size(struct btrfs_fs_info *fs_info, + struct btrfs_device *device) +{ + struct btrfs_trans_handle *trans; + struct btrfs_key key; + struct btrfs_path path; + struct btrfs_root *chunk_root = fs_info->chunk_root; + struct btrfs_dev_item *di; + u64 old_bytes = device->total_bytes; + int ret; + + if (IS_ALIGNED(device->total_bytes, fs_info->sectorsize)) + return 0; + + /* Align the in-memory total_bytes first, and use it later */ + device->total_bytes = round_down(device->total_bytes, + fs_info->sectorsize); + + key.objectid = BTRFS_DEV_ITEMS_OBJECTID; + key.type = BTRFS_DEV_ITEM_KEY; + key.offset = device->devid; + + trans = btrfs_start_transaction(chunk_root, 1); + if (IS_ERR(trans)) { + ret = PTR_ERR(trans); + error("error starting transaction: %d (%s)", + ret, strerror(-ret)); + return ret; + } + + btrfs_init_path(&path); + ret = btrfs_search_slot(trans, chunk_root, &key, &path, 0, 1); + if (ret > 0) { + error("failed to find DEV_ITEM for devid %llu", + device->devid); + ret = -ENOENT; + goto err; + } + if (ret < 0) { + error("failed to search chunk root: %d (%s)", + ret, strerror(-ret)); + goto err; + } + di = btrfs_item_ptr(path.nodes[0], path.slots[0], + struct btrfs_dev_item); + btrfs_set_device_total_bytes(path.nodes[0], di, device->total_bytes); + btrfs_mark_buffer_dirty(path.nodes[0]); + ret = btrfs_commit_transaction(trans, chunk_root); + if (ret < 0) { + error("failed to commit current transaction: %d (%s)", + ret, strerror(-ret)); + btrfs_release_path(&path); + return ret; + } + btrfs_release_path(&path); + printf("Fixed device size for devid %llu, old size: %llu new size: %llu\n", + device->devid, old_bytes, device->total_bytes); + return 1; + +err: + /* We haven't modified anything, it's OK to commit current trans */ + btrfs_commit_transaction(trans, chunk_root); + btrfs_release_path(&path); + return ret; +} + +/* + * Return 0 if super block total bytes matches with device total_bytes + * Return >0 if super block has mismatch total_bytes and fixed + * Return <0 if we failed to fix the mismatch total_bytes + */ +static int reset_super_total_bytes(struct btrfs_fs_info *fs_info) +{ + struct btrfs_trans_handle *trans; + struct btrfs_device *device; + struct list_head *dev_list = &fs_info->fs_devices->devices; + u64 total_bytes = 0; + u64 old_bytes = btrfs_super_total_bytes(fs_info->super_copy); + int ret; + + list_for_each_entry(device, dev_list, dev_list) + total_bytes += device->total_bytes; + + if (total_bytes == old_bytes) + return 0; + + /* Just in case */ + if (!IS_ALIGNED(total_bytes, fs_info->sectorsize)) { + error("final total_bytes still not aligned to %u, please report a bug to btrfs mail list", + fs_info->sectorsize); + return -EUCLEAN; + } + + btrfs_set_super_total_bytes(fs_info->super_copy, total_bytes); + + /* Commit transaction to update all super blocks */ + trans = btrfs_start_transaction(fs_info->tree_root, 1); + if (IS_ERR(trans)) { + ret = PTR_ERR(trans); + error("error starting transaction: %d (%s)", + ret, strerror(-ret)); + return ret; + } + ret = btrfs_commit_transaction(trans, fs_info->tree_root); + if (ret < 0) { + error("failed to commit current transaction: %d (%s)", + ret, strerror(-ret)); + return ret; + } + printf("Fixed super total bytes, old size: %llu new size: %llu\n", + old_bytes, total_bytes); + return 1; +} + +/* + * Repair device size related problems, including: + * 1) Unaligned total_bytes of dev_item + * v4.14-rc kernel introduced total_bytes alignment checker. + * However older kernel device add/shrink doesn't restrictly align + * these members. + * This will cause kernel warning everytime dev_item get updated. + * + * Although it can be fixed by shrinking device on latest kernel or + * use manually aligned size on older kernel, a fallback method in + * btrfs-progs won't hurt. + * + * 2) Mismatch super->total_bytes + * Due to similar situation, super->total_bytes can be mismatch with + * total size of all devices. + * This will cause the filesystem unable to be mounted with v4.14-rc kernel. + * + * Add such fixer in btrfs-progs as a fallback method. + * + * Fix all of them by manually rounding down total_bytes of each device, and + * recalculate super->total_bytes using all existing devices. + */ +static int reset_devs_size(struct btrfs_fs_info *fs_info) +{ + struct btrfs_device *device; + struct list_head *dev_list = &fs_info->fs_devices->devices; + bool have_bad_value = false; + int ret; + + /* Seed device is not support yet */ + if (fs_info->fs_devices->seed) { + error("resetting device size with seed device is not supported yet"); + return -ENOTTY; + } + + /* All devices must be on-line before reparing */ + if (list_empty(dev_list)) { + error("no device found"); + return -ENODEV; + } + list_for_each_entry(device, dev_list, dev_list) { + if (device->fd == -1 || !device->writeable) { + error("device with devid %llu is missing or not writeable", + device->devid); + error("resetting device size needs all device(s) present and writeable"); + return -ENODEV; + } + } + + /* Repair total_bytes of each device */ + list_for_each_entry(device, dev_list, dev_list) { + ret = reset_one_dev_size(fs_info, device); + if (ret < 0) + return ret; + if (ret > 0) + have_bad_value = true; + } + + /* Repair super total_byte */ + ret = reset_super_total_bytes(fs_info); + if (ret > 0) + have_bad_value = true; + if (have_bad_value) { + printf("Fixed unaligned/mismatch total_bytes for superblock and device item\n"); + ret = 1; + } else { + printf("No device size related problem found\n"); + ret = 0; + } + return ret; +} + const char * const cmd_check_usage[] = { "btrfs check [options] <device>", "Check structural integrity of a filesystem (unmounted).",
This patch introduces functions to repair device size related problems, including: 1) Unaligned total_bytes of dev_item v4.14-rc kernel introduced total_bytes alignment checker. However older kernel device add/shrink doesn't align these members. This will cause kernel warning every time dev_item get updated. Although it can be fixed by shrinking device on latest kernel or use manually aligned size on older kernel, a fallback method in btrfs-progs won't hurt. 2) Mismatch super->total_bytes There are some reports about unmountable fs, due to mismatched super->total_bytes. And normal kernel device shrink won't help since it only modify the total_bytes by the size changed, not re-calculating it. The root cause is still under investigation, but at least to fix it in btrfs-progs Fix all of them by manually rounding down total_bytes of each device, and recalculate super->total_bytes using all existing devices. Reported-by: Asif Youssuff <yoasif@gmail.com> Reported-by: Rich Rauenzahn <rrauenza@gmail.com> Signed-off-by: Qu Wenruo <quwenruo.btrfs@gmx.com> --- cmds-check.c | 191 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 191 insertions(+)