linux-kernel - Re: [mtd:nand/next 11/31] drivers/mtd/nand/raw/cadence-nand-controller.c:1893:4: error: implicit declaration of function 'ioread64

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-next>] [day] [month] [year] [list]

Message-ID: <20220921164720.6bbc56d5@xps-13>
Date:   Wed, 21 Sep 2022 16:47:19 +0200
From:   Miquel Raynal <miquel.raynal@...tlin.com>
To:     Valentin Korenblit <vkorenblit@...uans.com>
Cc:     kernel test robot <lkp@...el.com>, llvm@...ts.linux.dev,
        kbuild-all@...ts.01.org, linux-kernel@...r.kernel.org,
        Arnd Bergmann <arnd@...db.de>
Subject: Re: [mtd:nand/next 11/31]
 drivers/mtd/nand/raw/cadence-nand-controller.c:1893:4: error: implicit
 declaration of function 'ioread64_rep' is invalid in C99

Hi Valentin,

+ Arnd

The (erased) context of this thread:
https://lore.kernel.org/llvm/202209210641.MziHAbW7-lkp@intel.com/

vkorenblit@...uans.com wrote on Wed, 21 Sep 2022 14:27:47 +0200:

> Hi Miquel,
> 
> I see that x86_64 doesn't support readsq/writesq because of
> CONFIG_GENERIC_IOMAP (I was building for arm64). However,
> __raw_readq/writeq are available and there are a few drivers
> using them.

I would suggest to rather try using [read|write]sq() to get rid
of the CONFIG_GENERIC_IOMAP dependency. But it looks like those
functions are not defined on 32-bit architectures anyway. So if we want
the driver to compile on both arm and aarch64, it will not be enough.

In practice, I guess we should never have the 64-bit accessor executed
when running on a 32-bit platform thanks to the following conditions:

  1885			u8 data_dma_width = cdns_ctrl->caps2.data_dma_width;
  1886	
  1887			int len_in_words = (data_dma_width == 4) ? len >> 2 : len >> 3;
  1888	
  1889			/* read alingment data */
  1890			if (data_dma_width == 4)
  1891				ioread32_rep(cdns_ctrl->io.virt, buf, len_in_words);
  1892			else
> 1893				ioread64_rep(cdns_ctrl->io.virt, buf, len_in_words);

So maybe we could have something awful yet simple, like the following
within the Cadence driver:

#if !CONFIG_64BIT
readsq(...) { BUG()? }
#endif

Arnd, I've seen a couple of similar issues on the mailing lists in the
past 5 years but I could not find any working solution, I don't know
how all these threads have settled in the end. I thought maybe you
would have a better idea than the above hack :)

> Do you want me to re-implement readsq and writesq in Cadence
> driver privately or do you suggest a different (cleaner) approach?

I would rather prefer to avoid this solution, as anyway I believe there
is no "generic" implementation working in all cases, there were a
couple of attempts IIRC to bring generic helpers like the above for all
architectures, but none of them landed in Linus' tree, probably because
it just cannot be made...

Thanks,
Miquèl