Re: [f2fs-dev] [PATCH] f2fs: avoid fi->i_gc_rwsem[WRITE] lock in f2fs_gc

From: Chao Yu
Date: Tue Aug 14 2018 - 09:27:20 EST

Next message: Chao Yu: "Re: [f2fs-dev] [PATCH v3] f2fs: fix performance issue observed with multi-thread sequential read"
Previous message: syzbot: "possible deadlock in rhashtable_lookup_insert_fast"
In reply to: Jaegeuk Kim: "Re: [f2fs-dev] [PATCH] f2fs: avoid fi->i_gc_rwsem[WRITE] lock in f2fs_gc"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On 2018/8/14 4:12, Jaegeuk Kim wrote:
> On 08/12, Chao Yu wrote:
>> On 2018/8/4 10:31, Chao Yu wrote:
>>> How about keep lock order as:
>>>
>>> - inode_lock
>>> - i_mmap_sem
>>> - lock_all()
>>> - unlock_all()
>>> - i_gc_rwsem[WRITE]
>>> - lock_op()
>>
>> I got below warning when testing last dev-test:
>>
>> - f2fs_direct_IO current lock dependency
>> - i_gc_rwsem[WRITE]
>> - i_mmap_sem
>> - do_blockdev_direct_IO
>> - i_mmap_sem
>> - i_gc_rwsem[WRITE]
>>
>
> Yeah, it seems it's true.
> How about this?

It looks good to me, anyway, let me check this patch with fstests again.

Thanks,

>
> ---
> fs/f2fs/data.c | 4 ++--
> fs/f2fs/file.c | 43 +++++++++++++++++++++++--------------------
> 2 files changed, 25 insertions(+), 22 deletions(-)
>
> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
> index f09231b1cc74..021923dc666b 100644
> --- a/fs/f2fs/data.c
> +++ b/fs/f2fs/data.c
> @@ -2208,14 +2208,14 @@ static void f2fs_write_failed(struct address_space *mapping, loff_t to)
> loff_t i_size = i_size_read(inode);
>
> if (to > i_size) {
> - down_write(&F2FS_I(inode)->i_mmap_sem);
> down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> + down_write(&F2FS_I(inode)->i_mmap_sem);
>
> truncate_pagecache(inode, i_size);
> f2fs_truncate_blocks(inode, i_size, true);
>
> - up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> up_write(&F2FS_I(inode)->i_mmap_sem);
> + up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> }
> }
>
> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
> index 560751adba01..8b13afb23734 100644
> --- a/fs/f2fs/file.c
> +++ b/fs/f2fs/file.c
> @@ -798,8 +798,8 @@ int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
> if (attr->ia_valid & ATTR_SIZE) {
> bool to_smaller = (attr->ia_size <= i_size_read(inode));
>
> - down_write(&F2FS_I(inode)->i_mmap_sem);
> down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> + down_write(&F2FS_I(inode)->i_mmap_sem);
>
> truncate_setsize(inode, attr->ia_size);
>
> @@ -809,8 +809,8 @@ int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
> * do not trim all blocks after i_size if target size is
> * larger than i_size.
> */
> - up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> up_write(&F2FS_I(inode)->i_mmap_sem);
> + up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>
> if (err)
> return err;
> @@ -963,8 +963,8 @@ static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
> blk_start = (loff_t)pg_start << PAGE_SHIFT;
> blk_end = (loff_t)pg_end << PAGE_SHIFT;
>
> - down_write(&F2FS_I(inode)->i_mmap_sem);
> down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> + down_write(&F2FS_I(inode)->i_mmap_sem);
>
> truncate_inode_pages_range(mapping, blk_start,
> blk_end - 1);
> @@ -973,8 +973,8 @@ static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
> ret = f2fs_truncate_hole(inode, pg_start, pg_end);
> f2fs_unlock_op(sbi);
>
> - up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> up_write(&F2FS_I(inode)->i_mmap_sem);
> + up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> }
> }
>
> @@ -1201,6 +1201,7 @@ static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
>
> /* avoid gc operation during block exchange */
> down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> + down_write(&F2FS_I(inode)->i_mmap_sem);
>
> f2fs_lock_op(sbi);
> f2fs_drop_extent_tree(inode);
> @@ -1208,6 +1209,7 @@ static int f2fs_do_collapse(struct inode *inode, loff_t offset, loff_t len)
> ret = __exchange_data_block(inode, inode, end, start, nrpages - end, true);
> f2fs_unlock_op(sbi);
>
> + up_write(&F2FS_I(inode)->i_mmap_sem);
> up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> return ret;
> }
> @@ -1228,17 +1230,17 @@ static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
> if (ret)
> return ret;
>
> - down_write(&F2FS_I(inode)->i_mmap_sem);
> /* write out all dirty pages from offset */
> ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
> if (ret)
> - goto out_unlock;
> + return ret;
>
> ret = f2fs_do_collapse(inode, offset, len);
> if (ret)
> - goto out_unlock;
> + return ret;
>
> /* write out all moved pages, if possible */
> + down_write(&F2FS_I(inode)->i_mmap_sem);
> filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
> truncate_pagecache(inode, offset);
>
> @@ -1246,10 +1248,9 @@ static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
> truncate_pagecache(inode, new_size);
>
> ret = f2fs_truncate_blocks(inode, new_size, true);
> + up_write(&F2FS_I(inode)->i_mmap_sem);
> if (!ret)
> f2fs_i_size_write(inode, new_size);
> -out_unlock:
> - up_write(&F2FS_I(inode)->i_mmap_sem);
> return ret;
> }
>
> @@ -1315,10 +1316,9 @@ static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
> if (ret)
> return ret;
>
> - down_write(&F2FS_I(inode)->i_mmap_sem);
> ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
> if (ret)
> - goto out_sem;
> + return ret;
>
> pg_start = ((unsigned long long) offset) >> PAGE_SHIFT;
> pg_end = ((unsigned long long) offset + len) >> PAGE_SHIFT;
> @@ -1330,7 +1330,7 @@ static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
> ret = fill_zero(inode, pg_start, off_start,
> off_end - off_start);
> if (ret)
> - goto out_sem;
> + return ret;
>
> new_size = max_t(loff_t, new_size, offset + len);
> } else {
> @@ -1338,7 +1338,7 @@ static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
> ret = fill_zero(inode, pg_start++, off_start,
> PAGE_SIZE - off_start);
> if (ret)
> - goto out_sem;
> + return ret;
>
> new_size = max_t(loff_t, new_size,
> (loff_t)pg_start << PAGE_SHIFT);
> @@ -1350,6 +1350,7 @@ static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
> pgoff_t end;
>
> down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> + down_write(&F2FS_I(inode)->i_mmap_sem);
>
> truncate_pagecache_range(inode,
> (loff_t)index << PAGE_SHIFT,
> @@ -1361,6 +1362,7 @@ static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
> ret = f2fs_get_dnode_of_data(&dn, index, ALLOC_NODE);
> if (ret) {
> f2fs_unlock_op(sbi);
> + up_write(&F2FS_I(inode)->i_mmap_sem);
> up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> goto out;
> }
> @@ -1372,6 +1374,7 @@ static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
> f2fs_put_dnode(&dn);
>
> f2fs_unlock_op(sbi);
> + up_write(&F2FS_I(inode)->i_mmap_sem);
> up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>
> f2fs_balance_fs(sbi, dn.node_changed);
> @@ -1400,9 +1403,6 @@ static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
> else
> f2fs_i_size_write(inode, new_size);
> }
> -out_sem:
> - up_write(&F2FS_I(inode)->i_mmap_sem);
> -
> return ret;
> }
>
> @@ -1433,13 +1433,14 @@ static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
>
> down_write(&F2FS_I(inode)->i_mmap_sem);
> ret = f2fs_truncate_blocks(inode, i_size_read(inode), true);
> + up_write(&F2FS_I(inode)->i_mmap_sem);
> if (ret)
> - goto out;
> + return ret;
>
> /* write out all dirty pages from offset */
> ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
> if (ret)
> - goto out;
> + return ret;
>
> pg_start = offset >> PAGE_SHIFT;
> pg_end = (offset + len) >> PAGE_SHIFT;
> @@ -1448,6 +1449,7 @@ static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
>
> /* avoid gc operation during block exchange */
> down_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
> + down_write(&F2FS_I(inode)->i_mmap_sem);
> truncate_pagecache(inode, offset);
>
> while (!ret && idx > pg_start) {
> @@ -1463,16 +1465,17 @@ static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
> idx + delta, nr, false);
> f2fs_unlock_op(sbi);
> }
> + up_write(&F2FS_I(inode)->i_mmap_sem);
> up_write(&F2FS_I(inode)->i_gc_rwsem[WRITE]);
>
> /* write out all moved pages, if possible */
> + down_write(&F2FS_I(inode)->i_mmap_sem);
> filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
> truncate_pagecache(inode, offset);
> + up_write(&F2FS_I(inode)->i_mmap_sem);
>
> if (!ret)
> f2fs_i_size_write(inode, new_size);
> -out:
> - up_write(&F2FS_I(inode)->i_mmap_sem);
> return ret;
> }
>
>

Next message: Chao Yu: "Re: [f2fs-dev] [PATCH v3] f2fs: fix performance issue observed with multi-thread sequential read"
Previous message: syzbot: "possible deadlock in rhashtable_lookup_insert_fast"
In reply to: Jaegeuk Kim: "Re: [f2fs-dev] [PATCH] f2fs: avoid fi->i_gc_rwsem[WRITE] lock in f2fs_gc"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]