[LEDE-DEV] Advice needed for fixing a regression causing ar71xx mtd chips > 16MB to fail reading on kernel 4.9

[Christian Beier]

Hi list,

I am currently working on a fix for a regression that causes reads on 32 MB
flash to fail on ar71xx Linux 4.9 and need some help with this.

The bug is already reported in all detail at [1], but in shorts, it's:

* on the way to Linux 4.9, a mmap'ed read request was added to m25p80.c that
  gets used when a spi_flash_read_supported() callback returns true ([2]).
* the respective callback was wired to a function with [3], but this only took
  care of multi-chip setups.
* I suspect that this [4] check in ath79-spi.c causes mtd read to fail on chips
  > 16GB. I tried removing it, read _seems_ to work at first, but then the
  device reproducibly crashes. So it seems this is not a proper solution.

I reckon the right way to handle this is to add another test in
ath79_spi_flash_read_supported() [5].

My question: how to access the underlying flash device's size info from a
struct spi_device* ?

Thanks a lot,

   Christian

[1] https://bugs.openwrt.org/index.php?do=details&task_id=1376
[2] https://github.com/torvalds/linux/commit/08922f644878c9163ada8df3ef9def89be1d5e90
[3] https://git.openwrt.org/?p=openwrt/openwrt.git;a=commit;h=e0d58157d8e2ec3da0a1fe62a65aa756c5514871
[4] https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/ar71xx/patches-4.9/461-spi-ath79-add-fast-flash-read.patch;h=5c3583f758bb51cefacb31dce9f6a4c73f3a6384;hb=HEAD#l30
[5] https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/ar71xx/patches-4.9/461-spi-ath79-add-fast-flash-read.patch;h=5c3583f758bb51cefacb31dce9f6a4c73f3a6384;hb=HEAD#l17

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 195 bytes
Desc: Digitale Signatur von OpenPGP
URL: <http://lists.infradead.org/pipermail/lede-dev/attachments/20180220/4bc280e1/attachment.sig>