btrfs: fix transaction commit blocking during trim of unallocated space

btrfs: fix transaction commit blocking during trim of unallocated space

When trimming unallocated space, btrfs_trim_fs() holds the device_list_mutex
for the entire duration while iterating through all devices. On large
filesystems with significant unallocated space, this operation can take
minutes to hours on large storage systems.

This causes a problem because btrfs_run_dev_stats(), which is called
during transaction commit, also requires device_list_mutex:

  btrfs_trim_fs()
    mutex_lock(&fs_devices->device_list_mutex)
    list_for_each_entry(device, ...)
      btrfs_trim_free_extents(device)
    mutex_unlock(&fs_devices->device_list_mutex)

  commit_transaction()
    btrfs_run_dev_stats()
      mutex_lock(&fs_devices->device_list_mutex)  // blocked!
      ...

While trim is running, all transaction commits are blocked waiting for
the mutex.

Fix this by refactoring btrfs_trim_free_extents() to process devices in
bounded chunks (up to 2GB per iteration) and release device_list_mutex
between chunks.

Signed-off-by: robbieko <robbieko@synology.com>
Signed-off-by: jinbaohong <jinbaohong@synology.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>