Re: [PATCH v2] btrfs: trigger cow fixup via dirty_folio()

在 2026/7/27 18:08, Qu Wenruo 写道:
[...]
>> +again:
>> +    folio_lock(folio);
>> +
>> +    if (!folio->mapping || !folio_test_fixup_pending(folio))
>> +        goto out;
>> +    if (ret)
>> +        goto out_error;
>> +
>> +    btrfs_lock_extent(&inode->io_tree, page_start, page_end, 
>> &cached_state);
>> +
>> +    for (bit = 0; bit < blocks_per_folio; bit++) {
>> +        struct btrfs_ordered_extent *ordered;
>> +        const u64 start = page_start + (bit << fs_info- 
>> >sectorsize_bits);
>> +
>> +        if (test_bit(bit, delalloc_bitmap))
> 
> Dead check. It's always false, as the delalloc_bitmap is initialized to 
> all zero.

I'm an idiot, we can retry from again label, so this is the correct way 
to handle it.

Please ignore this one.

Thanks,
Qu