[PATCH] ubi: account the fastmap data blocks when checking found_pebs

From: richard@nod.at (Richard Weinberger)
To: linux-mtd@lists.infradead.org
Subject: [PATCH] ubi: account the fastmap data blocks when checking found_pebs
Date: Wed, 16 Jan 2019 09:00:26 +0100	[thread overview]
Message-ID: <1729890.X3pJ5IsDyS@blindfold> (raw)
In-Reply-To: <20190116005935.141529-1-houtao1@huawei.com>

Am Mittwoch, 16. Januar 2019, 01:59:35 CET schrieb Hou Tao:
> Now the data blocks used by current fastmap are not tracked by
> wear-leveling subsystem. So taking these blocks into account
> when checking found_pebs in ubi_wl_init(), else there will
> be ubi assertion as shown below when fastmap is enabled on
> a large NAND Flash (e.g. 2GB):
> 
> [  335.877389] UBI assert failed in ubi_wl_init at 1705 (pid 2349)
> [  335.878315] CPU: 7 PID: 2349 Comm: ubiattach Not tainted 4.20.0-10912-gd5d655675196-dirty #1
> [  335.879436] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
> [  335.881012] Call Trace:
> [  335.881402]  dump_stack+0x5b/0x8b
> [  335.881869]  ubi_wl_init+0xf8b/0x1500
> [  335.882960]  ubi_attach+0x63b/0xaa0
> [  335.883889]  ubi_attach_mtd_dev+0x107f/0x2e50
> [  335.887300]  ctrl_cdev_ioctl+0x13e/0x1d0
> [  335.888910]  do_vfs_ioctl+0x197/0xe60
> [  335.893337]  ksys_ioctl+0x66/0x70
> [  335.893788]  __x64_sys_ioctl+0x6f/0xb0
> [  335.894309]  do_syscall_64+0x91/0x270
> [  335.895337]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
> 
> Signed-off-by: Hou Tao <houtao1@huawei.com>
> ---
>  drivers/mtd/ubi/ubi.h | 13 +++++++++++++
>  drivers/mtd/ubi/wl.c  |  3 ++-
>  2 files changed, 15 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/mtd/ubi/ubi.h b/drivers/mtd/ubi/ubi.h
> index d47b9e436e67..90c28522aa07 100644
> --- a/drivers/mtd/ubi/ubi.h
> +++ b/drivers/mtd/ubi/ubi.h
> @@ -1235,4 +1235,17 @@ static inline struct ubi_wl_entry *ubi_find_fm_block(const struct ubi_device *ub
>  	return NULL;
>  }
>  
> +/**
> + * ubi_fm_data_blocks - return the number of data blocks used by fastmap
> + * @ubi: UBI device description object
> + *
> + * This function returns zero if fastmap is not available.
> + */
> +static inline int ubi_fm_data_blocks(struct ubi_device *ubi)
> +{
> +	/* the first block is the super block of fastmap */
> +	if (ubi->fm)
> +		return ubi->fm->used_blocks - 1;
> +	return 0;
> +}
>  #endif /* !__UBI_UBI_H__ */
> diff --git a/drivers/mtd/ubi/wl.c b/drivers/mtd/ubi/wl.c
> index 6f2ac865ff05..1b8f37ed8063 100644
> --- a/drivers/mtd/ubi/wl.c
> +++ b/drivers/mtd/ubi/wl.c
> @@ -1702,7 +1702,8 @@ int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
>  
>  	dbg_wl("found %i PEBs", found_pebs);
>  
> -	ubi_assert(ubi->good_peb_count == found_pebs);
> +	ubi_assert(ubi->good_peb_count ==
> +		   found_pebs + ubi_fm_data_blocks(ubi));
>  
>  	reserved_pebs = WL_RESERVED_PEBS;
>  	ubi_fastmap_init(ubi, &reserved_pebs);

This should get accounted in the loop above.
         list_for_each_entry(aeb, &ai->fastmap, u.list) {

I agree that this logic needs a redesign.
Please see this patch series I sent last year but had no time to further work
on it:
https://lkml.org/lkml/2018/6/13/755

I'd highly appreciate if you could revive it. :-)

Thanks,
//richard