Re: [PATCH] f2fs: support fault injection for flush submission error

[Matthew Wilcox]

On Thu, Nov 10, 2022 at 02:35:49AM +0800, Yangtao Li wrote:
> +++ b/fs/f2fs/segment.c
> @@ -486,7 +486,17 @@ void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi, bool from_bg)
>  static int __submit_flush_wait(struct f2fs_sb_info *sbi,
>  				struct block_device *bdev)
>  {
> -	int ret = blkdev_issue_flush(bdev);
> +	int ret;
> +
> +	if (time_to_inject(sbi, FAULT_FLUSH)) {
> +		f2fs_show_injection_info(sbi, FAULT_FLUSH);
> +		ret = -EIO;
> +		goto submit;

submit?  You don't submit anything at the 'submit' label.  it should
be called 'skip' or something.  But I think this is just badly written
and you don't need a goto at all.

What I don't understand about this patch is why you want it at all.
What failure are you simulating?  It appears to me that you're
simulating a bug in the filesystem, or maybe a bug in the device
firmware.  But why do you want to do that?

> +	}
> +
> +	ret = blkdev_issue_flush(bdev);
> +
> +submit:
>  
>  	trace_f2fs_issue_flush(bdev, test_opt(sbi, NOBARRIER),
>  				test_opt(sbi, FLUSH_MERGE), ret);
> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> index a43d8a46a6e5..3d3d22ac527b 100644
> --- a/fs/f2fs/super.c
> +++ b/fs/f2fs/super.c
> @@ -62,6 +62,7 @@ const char *f2fs_fault_name[FAULT_MAX] = {
>  	[FAULT_DQUOT_INIT]	= "dquot initialize",
>  	[FAULT_LOCK_OP]		= "lock_op",
>  	[FAULT_BLKADDR]		= "invalid blkaddr",
> +	[FAULT_FLUSH]		= "flush error",
>  };
>  
>  void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate,
> -- 
> 2.25.1
>