Thread (16 messages) 16 messages, 2 authors, 2021-09-14

Re: [PATCH 5/8] btrfs-progs: Add btrfs_uuid_tree_remove

From: Qu Wenruo <hidden>
Date: 2021-09-14 04:57:35


On 2021/9/14 下午12:49, Nikolay Borisov wrote:

On 14.09.21 г. 4:07, Qu Wenruo wrote:
quoted

On 2021/9/13 下午9:17, Nikolay Borisov wrote:
quoted
It will be used to clear received data on RW snapshots that were
received.

Signed-off-by: Nikolay Borisov <redacted>
---
   kernel-shared/ctree.h     |  3 ++
   kernel-shared/uuid-tree.c | 82 +++++++++++++++++++++++++++++++++++++++
   2 files changed, 85 insertions(+)
diff --git a/kernel-shared/ctree.h b/kernel-shared/ctree.h
index 91a85796a678..158281a9fd67 100644
--- a/kernel-shared/ctree.h
+++ b/kernel-shared/ctree.h
@@ -2860,6 +2860,9 @@ int btrfs_uuid_tree_add(struct
btrfs_trans_handle *trans, u8 *uuid, u8 type,
               u64 subvol_id_cpu);

   int btrfs_is_empty_uuid(u8 *uuid);
+int btrfs_uuid_tree_remove(struct btrfs_trans_handle *trans, u8
*uuid, u8 type,
+               u64 subid);
+
I have checked the callers in the incoming check patches, they are all
starting a new trans on uuid tree, thus there is no need to start a
trans out of this function.

Let's just pass the subvolume root into the function, and start a new
trans in side the function.

So that we only need one parameter @root, everything else can be fetched
from @root, and we can also modify the root item and uuid tree just in
one go.
I disagree, this function is a direct copy from the kernel with the only
adjustment that btrfs_uuid_to_key doesn't set key.type. If I change it
in progs then this means it won't be unified with the kernel when this
happens (if ever). I really rather keep shared functions as close as
possible.
OK, didn't notice it's really some shared function. (in kernel it's
implemented in uuid-tree.c, not ctree.c).

Then maybe it's better to introduce a btrfs-progs specific helper like
btrfs_remove_recevied_uuid(), other than calling the function directly.

In fact, both mode shares the same root item modification code, which is
really a duplication.

As in btrfs-progs we don't need the other call sites in kernels which is
shared for subvolume removal.

Thanks,
Qu
quoted
Thanks,
Qu
quoted
   static inline int is_fstree(u64 rootid)
   {
diff --git a/kernel-shared/uuid-tree.c b/kernel-shared/uuid-tree.c
index 51a7b5d9ff5d..0f0fbf667dda 100644
--- a/kernel-shared/uuid-tree.c
+++ b/kernel-shared/uuid-tree.c
@@ -120,3 +120,85 @@ int btrfs_is_empty_uuid(u8 *uuid)
       }
       return 1;
   }
+
+int btrfs_uuid_tree_remove(struct btrfs_trans_handle *trans, u8
*uuid, u8 type,
+        u64 subid)
+{
+    struct btrfs_fs_info *fs_info = trans->fs_info;
+    struct btrfs_root *uuid_root = fs_info->uuid_root;
+    int ret;
+    struct btrfs_path *path = NULL;
+    struct btrfs_key key;
+    struct extent_buffer *eb;
+    int slot;
+    unsigned long offset;
+    u32 item_size;
+    unsigned long move_dst;
+    unsigned long move_src;
+    unsigned long move_len;
+
+    if (!uuid_root) {
+        ret = -EINVAL;
+        goto out;
+    }
+
+    btrfs_uuid_to_key(uuid, &key);
+    key.type = type;
+
+    path = btrfs_alloc_path();
+    if (!path) {
+        ret = -ENOMEM;
+        goto out;
+    }
+
+    ret = btrfs_search_slot(trans, uuid_root, &key, path, -1, 1);
+    if (ret < 0) {
+        warning("error %d while searching for uuid item!", ret);
+        goto out;
+    }
+    if (ret > 0) {
+        ret = -ENOENT;
+        goto out;
+    }
+
+
+    eb = path->nodes[0];
+    slot = path->slots[0];
+    offset = btrfs_item_ptr_offset(eb, slot);
+    item_size = btrfs_item_size_nr(eb, slot);
+    if (!IS_ALIGNED(item_size, sizeof(u64))) {
+        warning("uuid item with illegal size %lu!",    (unsigned
long)item_size);
+        ret = -ENOENT;
+        goto out;
+    }
+    while (item_size) {
+        __le64 read_subid;
+
+        read_extent_buffer(eb, &read_subid, offset, sizeof(read_subid));
+        if (le64_to_cpu(read_subid) == subid)
+            break;
+        offset += sizeof(read_subid);
+        item_size -= sizeof(read_subid);
+    }
+
+    if (!item_size) {
+        ret = -ENOENT;
+        goto out;
+    }
+
+    item_size = btrfs_item_size_nr(eb, slot);
+    if (item_size == sizeof(subid)) {
+        ret = btrfs_del_item(trans, uuid_root, path);
+        goto out;
+    }
+
+    move_dst = offset;
+    move_src = offset + sizeof(subid);
+    move_len = item_size - (move_src - btrfs_item_ptr_offset(eb, slot));
+    memmove_extent_buffer(eb, move_dst, move_src, move_len);
+    btrfs_truncate_item(path, item_size - sizeof(subid), 1);
+
+out:
+    btrfs_free_path(path);
+    return ret;
+}
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help