Message-ID: <48b21671-19ae-0dbd-96cd-7300fd600c9b@huawei.com>
Date: Thu, 18 Apr 2024 14:12:39 +0800
From: Baokun Li <libaokun1@...wei.com>
To: Jingbo Xu <jefflexu@...ux.alibaba.com>
CC: <linux-erofs@...ts.ozlabs.org>, <xiang@...nel.org>, <chao@...nel.org>,
	<huyue2@...lpad.com>, <viro@...iv.linux.org.uk>, <brauner@...nel.org>,
	<linux-kernel@...r.kernel.org>, <yangerkun@...wei.com>, <houtao1@...wei.com>,
	Baokun Li <libaokun1@...wei.com>
Subject: Re: [PATCH v2] erofs: reliably distinguish block based and fscache
 mode

On 2024/4/18 13:50, Jingbo Xu wrote:
>
> On 4/18/24 11:36 AM, Baokun Li wrote:
>> On 2024/4/18 10:16, Jingbo Xu wrote:
>>> Hi Baokun,
>>>
>>> Thanks for catching this and move forward fixing this!
>> Hi Jingbo，
>>
>> Thanks for your review！
>>
>>> On 4/17/24 2:55 PM, Baokun Li wrote:
>>>> When erofs_kill_sb() is called in block dev based mode, s_bdev may
>>>> not have
>>>> been initialised yet, and if CONFIG_EROFS_FS_ONDEMAND is enabled, it
>>>> will
>>>> be mistaken for fscache mode, and then attempt to free an anon_dev
>>>> that has
>>>> never been allocated, triggering the following warning:
>>>>
>>>> ============================================
>>>> ida_free called for id=0 which is not allocated.
>>>> WARNING: CPU: 14 PID: 926 at lib/idr.c:525 ida_free+0x134/0x140
>>>> Modules linked in:
>>>> CPU: 14 PID: 926 Comm: mount Not tainted 6.9.0-rc3-dirty #630
>>>> RIP: 0010:ida_free+0x134/0x140
>>>> Call Trace:
>>>>    <TASK>
>>>>    erofs_kill_sb+0x81/0x90
>>>>    deactivate_locked_super+0x35/0x80
>>>>    get_tree_bdev+0x136/0x1e0
>>>>    vfs_get_tree+0x2c/0xf0
>>>>    do_new_mount+0x190/0x2f0
>>>>    [...]
>>>> ============================================
>>>>
>>>> Instead of allocating the erofs_sb_info in fill_super() allocate it
>>>> during erofs_get_tree() and ensure that erofs can always have the info
>>>> available during erofs_kill_sb().
>>> I'm not sure if allocating erofs_sb_info in erofs_init_fs_context() will
>>> be better, as I see some filesystems (e.g. autofs) do this way.  Maybe
>>> another potential advantage of doing this way is that erofs_fs_context
>>> is not needed anymore and we can use sbi directly.
>> Yes, except for some extra memory usage when remounting,
>> this idea sounds great. Let me send a version of v3 to get rid
>> of erofs_fs_context.
> I'm not sure if Gao Xaing also prefers this.  I think it would be better
> to wait and listen for his thoughts before we sending v3.
  Okay, there's no rush on this.
>>>> Signed-off-by: Christian Brauner <brauner@...nel.org>
>>>> Signed-off-by: Baokun Li <libaokun1@...wei.com>
>>>> ---
>>>> Changes since v1:
>>>>     Allocate and initialise fc->s_fs_info in erofs_fc_get_tree()
>>>> instead of
>>>>     modifying fc->sb_flags.
>>>>
>>>> V1:
>>>> https://lore.kernel.org/r/20240415121746.1207242-1-libaokun1@huawei.com/
>>>>
>>>>    fs/erofs/super.c | 51 ++++++++++++++++++++++++++----------------------
>>>>    1 file changed, 28 insertions(+), 23 deletions(-)
>>>>
>>>> diff --git a/fs/erofs/super.c b/fs/erofs/super.c
>>>> index b21bd8f78dc1..4104280be2ea 100644
>>>> --- a/fs/erofs/super.c
>>>> +++ b/fs/erofs/super.c
>>>> @@ -581,8 +581,7 @@ static const struct export_operations
>>>> erofs_export_ops = {
>>>>    static int erofs_fc_fill_super(struct super_block *sb, struct
>>>> fs_context *fc)
>>>>    {
>>>>        struct inode *inode;
>>>> -    struct erofs_sb_info *sbi;
>>>> -    struct erofs_fs_context *ctx = fc->fs_private;
>>>> +    struct erofs_sb_info *sbi = EROFS_SB(sb);
>>>>        int err;
>>>>          sb->s_magic = EROFS_SUPER_MAGIC;
>>>> @@ -590,19 +589,6 @@ static int erofs_fc_fill_super(struct
>>>> super_block *sb, struct fs_context *fc)
>>>>        sb->s_maxbytes = MAX_LFS_FILESIZE;
>>>>        sb->s_op = &erofs_sops;
>>>>    -    sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
>>>> -    if (!sbi)
>>>> -        return -ENOMEM;
>>>> -
>>>> -    sb->s_fs_info = sbi;
>>>> -    sbi->opt = ctx->opt;
>>>> -    sbi->devs = ctx->devs;
>>>> -    ctx->devs = NULL;
>>>> -    sbi->fsid = ctx->fsid;
>>>> -    ctx->fsid = NULL;
>>>> -    sbi->domain_id = ctx->domain_id;
>>>> -    ctx->domain_id = NULL;
>>>> -
>>>>        sbi->blkszbits = PAGE_SHIFT;
>>>>        if (erofs_is_fscache_mode(sb)) {
>>>>            sb->s_blocksize = PAGE_SIZE;
>>>> @@ -704,11 +690,32 @@ static int erofs_fc_fill_super(struct
>>>> super_block *sb, struct fs_context *fc)
>>>>        return 0;
>>>>    }
>>>>    -static int erofs_fc_get_tree(struct fs_context *fc)
>>>> +static void erofs_ctx_to_info(struct fs_context *fc)
>>>>    {
>>>>        struct erofs_fs_context *ctx = fc->fs_private;
>>>> +    struct erofs_sb_info *sbi = fc->s_fs_info;
>>>> +
>>>> +    sbi->opt = ctx->opt;
>>>> +    sbi->devs = ctx->devs;
>>>> +    ctx->devs = NULL;
>>>> +    sbi->fsid = ctx->fsid;
>>>> +    ctx->fsid = NULL;
>>>> +    sbi->domain_id = ctx->domain_id;
>>>> +    ctx->domain_id = NULL;
>>>> +}
>>> I'm not sure if abstracting this logic into a seperate helper really
>>> helps understanding the code as the logic itself is quite simple and
>>> easy to be understood. Usually it's a hint of over-abstraction when a
>>> simple helper has only one caller.
>>>
>> Static functions that have only one caller are compiled inline, so we
>> don't have to worry about how that affects the code.
>>
>> The reason these codes are encapsulated in a separate function is so
>> that the code reader understands that these codes are integrated
>> as a whole, and that we shouldn't have to move one or two of these
>> lines individually.
>>
>> But after we get rid of erofs_fs_context, those won't be needed
>> anymore.
> Yeah, I understand. It's only coding style concerns.
>
>
>
Okay, thanks！

-- 
With Best Regards,
Baokun Li

Powered by blists - more mailing lists