Message-ID: <29d6503d-9e34-4d64-aacb-32f36152ce9c@iopsys.eu>
Date: Fri, 22 Aug 2025 23:00:57 +0300
From: Mikhail Kshevetskiy <mikhail.kshevetskiy@...sys.eu>
To: Lorenzo Bianconi <lorenzo@...nel.org>
Cc: Ray Liu <ray.liu@...oha.com>, Mark Brown <broonie@...nel.org>,
 linux-arm-kernel@...ts.infradead.org, linux-spi@...r.kernel.org,
 linux-kernel@...r.kernel.org, Frieder Schrempf <frieder.schrempf@...tron.de>
Subject: Re: [PATCH v3 00/14] spi: airoha: driver fixes & improvements

patch 3 (spi: airoha: add support of dual/quad wires spi modes) is
necessary for non-dma reading. This can happen if dirmap creation fails
by some reason. In this case the kernel uses exec_op() handler to
execute dirmap requests.

According to dtb the device supports dualwire operations, so dualwire
commands issued for cache reading. Unfortunately  exec_op() handler does
not supports dual/quad operations, so flash reading fails. Here is the
corresponding boot log
-----------------------------------------------------
[    2.980719] spi-nand spi0.0: Micron SPI NAND was found.
[    2.986040] spi-nand spi0.0: 256 MiB, block size: 128 KiB, page size:
2048, OOB size: 128
[    2.994709] 2 fixed-partitions partitions found on MTD device spi0.0
[    3.001075] Creating 2 MTD partitions on "spi0.0":
[    3.005862] 0x000000000000-0x000000020000 : "bl2"
[    3.011272] 0x000000020000-0x000010000000 : "ubi"
...
[    6.195594] ubi0: attaching mtd1
[   13.338398] ubi0: scanning is finished
[   13.342188] ubi0 error: ubi_read_volume_table: the layout volume was
not found
[   13.349784] ubi0 error: ubi_attach_mtd_dev: failed to attach mtd1,
error -22
[   13.356897] UBI error: cannot attach mtd1
-------------------------------------------------------

If dma=off, but patch-3 is applied -- everything works (but slowly).

Here is a patch to disable DMA operations:

 --- a/drivers/spi/spi-airoha-snfi.c
+++ b/drivers/spi/spi-airoha-snfi.c
@@ -618,6 +618,7 @@ static int airoha_snand_dirmap_create(struct
spi_mem_dirmap_desc *desc)
        if (desc->info.offset + desc->info.length > U32_MAX)
                return -EINVAL;
 
+       return -E2BIG;

        if (!airoha_snand_supports_op(desc->mem, &desc->info.op_tmpl))
                return -EOPNOTSUPP;



On 22.08.2025 16:14, Lorenzo Bianconi wrote:
>> On 22.08.2025 09:46, Lorenzo Bianconi wrote:
>>>> On 21.08.2025 14:34, Lorenzo Bianconi wrote:
>>>>>> On 21.08.2025 09:58, Lorenzo Bianconi wrote:
>>>>>>>> This patch series greatly improve airoha snfi driver and fix a
>>>>>>>> number of serious bug.
>>>>>>>>
>>>>>>>> Fixed bugs:
>>>>>>>>  * Fix reading/writing of flashes with more than one plane per lun
>>>>>>>>  * Fix inability to read/write oob area
>>>>>>>>  * Fill the buffer with 0xff before writing
>>>>>>>>  * Fix reading of flashes supporting continuous reading mode
>>>>>>>>  * Fix error paths
>>>>>>>>
>>>>>>>> Improvements:
>>>>>>>>  * Add support of dual/quad wires spi modes in exec_op().
>>>>>>>>  * Support of dualio/quadio flash reading commands
>>>>>>>>  * Remove dirty hack that reads flash page settings from SNFI registers
>>>>>>>>    during driver startup
>>>>>>>>
>>>>>>>> Unfortunately I am unable to test the driver with linux at the moment,
>>>>>>>> so only the following testing was done:
>>>>>>> It seems to me this is quite an important rework of the driver. I would prefer
>>>>>>> to have some test results for this series. Are you able to run mtd_test kernel
>>>>>>> module for testing?
>>>>>> I'll try to build latest openwrt with this patches  and mtd_test kernel
>>>>>> module and try it on one of our boards.
>>>>> what board are you using for testing? If it is based on Airoha-7581 you could
>>>>> use the following repo for testing.
>>>>>
>>>>> https://github.com/Ansuel/openwrt/tree/openwrt-24.10-airoha-an7581-stable
>>>>>
>>>>> Regards,
>>>>> Lorenzo
>>>> What tests do you suggest to run?
>>> IIRC I run all of them. Can you please report even if there are some
>>> improvements (or penalties) in read/write speed
>> Do you mean to run it twice? with patches and without?
> I mean I would suggest to check if there are any difference in read/write
> speed (so yes, you need a baseline).
>
> Regards,
> Lorenzo
>
>>>> I have a single flash I boot from. It have only 2 mtd partitions:
>>>>
>>>> [    2.980849] spi-nand spi0.0: Micron SPI NAND was found.
>>>> [    2.986102] spi-nand spi0.0: 256 MiB, block size: 128 KiB, page size:
>>>> 2048, OOB size: 128
>>>> [    2.994493] 2 fixed-partitions partitions found on MTD device spi0.0
>>>> [    3.000856] Creating 2 MTD partitions on "spi0.0":
>>>> [    3.005651] 0x000000000000-0x000000020000 : "bl2"
>>>> [    3.011247] 0x000000020000-0x000010000000 : "ubi"
>>>>
>>>> Most of tests are destructive. So If I use "bl2" or "ubi" partition for
>>>> test, next time I will be unable to boot :-(
>>> yes, I flashed the device after carrying out the test.
>>>
>>> Regards,
>>> Lorenzo
>>>
>>>> Do you suggest to patch u-boot & linux to have more mtd partitions?
>>>>
>>>> This is the results of the only read-only test I found.
>>>>
>>>> root@...nWrt:/lib/modules/6.6.79# insmod mtd_test.ko
>>>> root@...nWrt:/lib/modules/6.6.79# insmod mtd_readtest.ko dev=1
>>>> [  159.121706]
>>>> [  159.123220] =================================================
>>>> [  159.129053] mtd_readtest: MTD device: 1
>>>> [  159.132898] mtd_readtest: MTD device size 268304384, eraseblock size
>>>> 131072, page size 2048, count of eraseblocks 2047, pages per eraseblock
>>>> 64, OOB size 128
>>>> [  159.147008] mtd_test: scanning for bad eraseblocks
>>>> [  159.152141] mtd_test: scanned 2047 eraseblocks, 0 are bad
>>>> [  159.157549] mtd_readtest: testing page read
>>>>
>>>> Mikhail
>>>>
>>>>>> Actually patches can be divided on to parts:
>>>>>> * fixes of current driver (patches 1-10)
>>>>>> * change of behavior to avoid reading flash page settings from SNFI
>>>>>> registers during driver startup (patches 11-14)
>>>>>>
>>>>>> The changes are based on the code we are using for more than 3 years. I
>>>>>> adapt it to latest linux/u-boot code.
>>>>>>
>>>>>> Up to now the only known issue appears on en7523 chips only. Here a
>>>>>> corresponding patch description (not added to this series)
>>>>>> ====================================================== spi: airoha:
>>>>>> en7523: workaround flash damaging if UART_TXD was short to GND We found
>>>>>> that some serial console may pull TX line to GROUND during board boot
>>>>>> time. Airoha uses TX line as one of it's BOOT pins. This will lead to
>>>>>> booting in RESERVED boot mode. It was found that some flashes operates
>>>>>> incorrectly in RESERVED mode. Micron and Skyhigh flashes are definitely
>>>>>> affected by the issue, Winbond flashes are NOT affected. Details:
>>>>>> -------- DMA reading of odd pages on affected flashes operates
>>>>>> incorrectly. Page reading offset (start of the page) on hardware level
>>>>>> is replaced by 0x10. Thus results in incorrect data reading. Usage of
>>>>>> UBI make things even worse. Any attempt to access UBI leads to ubi
>>>>>> damaging. As result OS loading becomes impossible. Non-DMA reading is
>>>>>> OK. =======================================================
>>>>>>
>>>>>> Regards,
>>>>>> Mikhail
>>>>>>
>>>>>>
>>>>>>> Regards,
>>>>>>> Lorenzo
>>>>>>>
>>>>>>>>  * Driver compiles without error.
>>>>>>>>  * All changes were tested with corresponding u-boot driver. U-Boot
>>>>>>>>    SpiNAND driver was modified as well to match linux-6.17-rc2 with
>>>>>>>>    additional fixes for continuous mode.
>>>>>>>>
>>>>>>>> Changes v2:
>>>>>>>>  * minor fix
>>>>>>>>  * add comments to code
>>>>>>>>
>>>>>>>> Changes v3:
>>>>>>>>  * add patch to prevent continuous reading
>>>>>>>>
>>>>>>>> Mikhail Kshevetskiy (14):
>>>>>>>>   spi: airoha: return an error for continuous mode dirmap creation cases
>>>>>>>>   spi: airoha: remove unnecessary restriction length
>>>>>>>>   spi: airoha: add support of dual/quad wires spi modes
>>>>>>>>   spi: airoha: remove unnecessary switch to non-dma mode
>>>>>>>>   spi: airoha: unify dirmap read/write code
>>>>>>>>   spi: airoha: switch back to non-dma mode in the case of error
>>>>>>>>   spi: airoha: fix reading/writing of flashes with more than one plane
>>>>>>>>     per lun
>>>>>>>>   spi: airoha: support of dualio/quadio flash reading commands
>>>>>>>>   spi: airoha: allow reading/writing of oob area
>>>>>>>>   spi: airoha: buffer must be 0xff-ed before writing
>>>>>>>>   spi: airoha: avoid setting of page/oob sizes in REG_SPI_NFI_PAGEFMT
>>>>>>>>   spi: airoha: reduce the number of modification of REG_SPI_NFI_CNFG and
>>>>>>>>     REG_SPI_NFI_SECCUS_SIZE registers
>>>>>>>>   spi: airoha: set custom sector size equal to flash page size
>>>>>>>>   spi: airoha: avoid reading flash page settings from SNFI registers
>>>>>>>>     during driver startup
>>>>>>>>
>>>>>>>>  drivers/spi/spi-airoha-snfi.c | 508 +++++++++++++++++-----------------
>>>>>>>>  1 file changed, 260 insertions(+), 248 deletions(-)
>>>>>>>>
>>>>>>>> -- 
>>>>>>>> 2.50.1
>>>>>>>>

Powered by blists - more mailing lists