linux-kernel - Re: ubifs_recover_master_node: failed to recover master node

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <87ses4ibo5.fsf@bootlin.com>
Date: Wed, 06 Nov 2024 17:12:26 +0100
From: Miquel Raynal <miquel.raynal@...tlin.com>
To: Chris Packham <chris.packham@...iedtelesis.co.nz>
Cc: "linux-mtd@...ts.infradead.org" <linux-mtd@...ts.infradead.org>,
  "linux-spi@...r.kernel.org" <linux-spi@...r.kernel.org>,
  "broonie@...nel.org" <broonie@...nel.org>,
  "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: ubifs_recover_master_node: failed to recover master node


Hi Chris,

On 30/10/2024 at 10:13:45 +13, Chris Packham <chris.packham@...iedtelesis.co.nz> wrote:

> On 29/10/24 13:38, Chris Packham wrote:
>> (resend as plaintext)
>>
>> Hi,
>>
>> I recently added support for the SPI-NAND controller on the RTL9302C
>> SoC[1]. I did most of the work against Linux 6.11 and it's working
>> fine there. I recently rebased against the tip of Linus's tree
>> (6.12-rc5) and found I was getting ubifs errors when mounting:
>>
>> [    1.255191] spi-nand spi1.0: Macronix SPI NAND was found.
>> [    1.261283] spi-nand spi1.0: 256 MiB, block size: 128 KiB, page
>> size: 2048, OOB size: 64
>> [    1.271134] 2 fixed-partitions partitions found on MTD device spi1.0
>> [    1.278247] Creating 2 MTD partitions on "spi1.0":
>> [    1.283631] 0x000000000000-0x00000f000000 : "user"
>> [   20.481108] 0x00000f000000-0x000010000000 : "Reserved"
>> [   72.240347] ubi0: scanning is finished
>> [   72.270577] ubi0: attached mtd3 (name "user", size 240 MiB)
>> [   72.276815] ubi0: PEB size: 131072 bytes (128 KiB), LEB size:
>> 126976 bytes
>> [   72.284537] ubi0: min./max. I/O unit sizes: 2048/2048, sub-page
>> size 2048
>> [   72.292132] ubi0: VID header offset: 2048 (aligned 2048), data
>> offset: 4096
>> [   72.299885] ubi0: good PEBs: 1920, bad PEBs: 0, corrupted PEBs: 0
>> [   72.306689] ubi0: user volume: 1, internal volumes: 1, max. volumes
>> count: 128
>> [   72.314747] ubi0: max/mean erase counter: 1/0, WL threshold: 4096,
>> image sequence number: 252642230
>> [   72.324850] ubi0: available PEBs: 0, total reserved PEBs: 1920,
>> PEBs reserved for bad PEB handling: 40
>> [   72.370123] ubi0: background thread "ubi_bgt0d" started, PID 141
>> [   72.470740] UBIFS (ubi0:0): Mounting in unauthenticated mode
>> [   72.490246] UBIFS (ubi0:0): background thread "ubifs_bgt0_0"
>> started, PID 144
>> [   72.528272] UBIFS error (ubi0:0 pid 143):
>> ubifs_recover_master_node: failed to recover master node
>> [   72.550122] UBIFS (ubi0:0): background thread "ubifs_bgt0_0" stops
>> [   72.710720] UBIFS (ubi0:0): Mounting in unauthenticated mode
>> [   72.717447] UBIFS (ubi0:0): background thread "ubifs_bgt0_0"
>> started, PID 149
>> [   72.777602] UBIFS error (ubi0:0 pid 148):
>> ubifs_recover_master_node: failed to recover master node
>> [   72.787792] UBIFS (ubi0:0): background thread "ubifs_bgt0_0" stops
>>
>> Full dmesg output is at[2]
>>
>> git bisect lead me to commit 11813857864f ("mtd: spi-nand: macronix:
>> Continuous read support"). Reverting the blamed commit from 6.12-rc5
>> seems to avoid the problem. The flash chip on my board is a
>> MX30LF2G28AD-TI. I'm not sure if there is a problem with 11813857864f
>> or with my spi-mem driver that is exposed after support for continuous
>> read is enabled.
>>
> A bit of an update. The ubifs failure is from the is_empty() check in
> get_master_node(). It looks like portions of the LEB are 0 instead of
> 0xff. I've also found if I avoid use the non-DMA path in my driver I
> don't have such a problem. I think there is at least one problem in my
> driver because I don't handle DMAing more than 0xffff bytes.

I am going through my mails in a chronological order :-)

Glad to see you found a lead. I was already a bit suspicious about the
DMA path, glad to see we might narrow down the problem.

Is the 0xffff limitation a hard constraint or is it just a pure software
constraint? If we reach a hard constraint, maybe you should check that
when you decide which path you take.

Miquèl