2017-03-20 05:22:31

by Stephen Rothwell

[permalink] [raw]
Subject: linux-next: build warnings after merge of the akpm-current tree

Hi Andrew,

After merging the akpm-current tree, today's linux-next build (x86_64
allmodconfig) produced these warnings:

drivers/crypto/cavium/zip/zip_main.c: In function 'zip_show_stats':
drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 4 has type 'long long int' [-Wformat=]
seq_printf(s, " ZIP Device %d Stats\n"
^
drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 5 has type 'long long int' [-Wformat=]
drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 6 has type 'long long int' [-Wformat=]
drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 7 has type 'long long int' [-Wformat=]
drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 10 has type 'long long int' [-Wformat=]
drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 11 has type 'long long int' [-Wformat=]
drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 12 has type 'long long int' [-Wformat=]
drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 13 has type 'long long int' [-Wformat=]
drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 14 has type 'long long int' [-Wformat=]
drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 15 has type 'long long int' [-Wformat=]

Introduced by commit

09ae5d37e093 ("crypto: zip - Add Compression/Decompression statistics")

from the crypto tree interacting with commit

3f4ca3d25e1a ("asm-generic, x86: wrap atomic operations")

from the akpm-current tree.

This latter commit changed atomic64read() from "long" to "long long"
on x86_64.

--
Cheers,
Stephen Rothwell


2017-03-20 09:38:47

by Dmitry Vyukov

[permalink] [raw]
Subject: Re: linux-next: build warnings after merge of the akpm-current tree

On Mon, Mar 20, 2017 at 6:22 AM, Stephen Rothwell <[email protected]> wrote:
> Hi Andrew,
>
> After merging the akpm-current tree, today's linux-next build (x86_64
> allmodconfig) produced these warnings:
>
> drivers/crypto/cavium/zip/zip_main.c: In function 'zip_show_stats':
> drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 4 has type 'long long int' [-Wformat=]
> seq_printf(s, " ZIP Device %d Stats\n"
> ^
> drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 5 has type 'long long int' [-Wformat=]
> drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 6 has type 'long long int' [-Wformat=]
> drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 7 has type 'long long int' [-Wformat=]
> drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 10 has type 'long long int' [-Wformat=]
> drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 11 has type 'long long int' [-Wformat=]
> drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 12 has type 'long long int' [-Wformat=]
> drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 13 has type 'long long int' [-Wformat=]
> drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 14 has type 'long long int' [-Wformat=]
> drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 15 has type 'long long int' [-Wformat=]
>
> Introduced by commit
>
> 09ae5d37e093 ("crypto: zip - Add Compression/Decompression statistics")
>
> from the crypto tree interacting with commit
>
> 3f4ca3d25e1a ("asm-generic, x86: wrap atomic operations")
>
> from the akpm-current tree.
>
> This latter commit changed atomic64read() from "long" to "long long"
> on x86_64.

Hi,

Previously atomic operations returned different types on 32/64 bits. I
think that was pretty unfortunate and could actually lead to lots of
such warnings (as there were no single format specifier that one could
use portably), and was difficult to wrap. So I used "unsigned long
long" for all operations.
I actually suspect that the new zip_main.c code would cause a similar
warning on 32 bits even without my change, because atomic ops already
returned unsigned long long there.
So I think we need to fix zip_main.c.

2017-03-20 12:31:28

by Jan Glauber

[permalink] [raw]
Subject: Re: linux-next: build warnings after merge of the akpm-current tree

On Mon, Mar 20, 2017 at 10:05:54AM +0100, Dmitry Vyukov wrote:
> On Mon, Mar 20, 2017 at 6:22 AM, Stephen Rothwell <[email protected]> wrote:
> > Hi Andrew,
> >
> > After merging the akpm-current tree, today's linux-next build (x86_64
> > allmodconfig) produced these warnings:
> >
> > drivers/crypto/cavium/zip/zip_main.c: In function 'zip_show_stats':
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 4 has type 'long long int' [-Wformat=]
> > seq_printf(s, " ZIP Device %d Stats\n"
> > ^
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 5 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 6 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 7 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 10 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 11 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 12 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 13 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 14 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 15 has type 'long long int' [-Wformat=]
> >
> > Introduced by commit
> >
> > 09ae5d37e093 ("crypto: zip - Add Compression/Decompression statistics")
> >
> > from the crypto tree interacting with commit
> >
> > 3f4ca3d25e1a ("asm-generic, x86: wrap atomic operations")
> >
> > from the akpm-current tree.
> >
> > This latter commit changed atomic64read() from "long" to "long long"
> > on x86_64.
>
> Hi,
>
> Previously atomic operations returned different types on 32/64 bits. I
> think that was pretty unfortunate and could actually lead to lots of
> such warnings (as there were no single format specifier that one could
> use portably), and was difficult to wrap. So I used "unsigned long
> long" for all operations.
> I actually suspect that the new zip_main.c code would cause a similar
> warning on 32 bits even without my change, because atomic ops already
> returned unsigned long long there.

The driver depends on 64BIT (HW is 64 bit only).

> So I think we need to fix zip_main.c.

I'm not entirely sure, all drivers that are available for x86_64 and
arm64 and use atomic64_read will show such warnings.

One possible solution would be to disable COMPILE_TEST for the zip
driver, but that would just hide the issue.

--Jan

2017-03-20 17:07:06

by Challa, Mahipal

[permalink] [raw]
Subject: Re: linux-next: build warnings after merge of the akpm-current tree



From: Jan Glauber
Sent: Monday, March 20, 2017 6:00 PM
To: Dmitry Vyukov
Cc: Stephen Rothwell; Andrew Morton; Herbert Xu; [email protected]; LKML; Challa, Mahipal
Subject: Re: linux-next: build warnings after merge of the akpm-current tree
?
On Mon, Mar 20, 2017 at 10:05:54AM +0100, Dmitry Vyukov wrote:
> On Mon, Mar 20, 2017 at 6:22 AM, Stephen Rothwell <[email protected]> wrote:
> > Hi Andrew,
> >
> > After merging the akpm-current tree, today's linux-next build (x86_64
> > allmodconfig) produced these warnings:
> >
> > drivers/crypto/cavium/zip/zip_main.c: In function 'zip_show_stats':
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 4 has type 'long long int' [-Wformat=]
> >???? seq_printf(s, "??????? ZIP Device %d Stats\n"
> >?????????????????? ^
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 5 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 6 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 7 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 10 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 11 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 12 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 13 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 14 has type 'long long int' [-Wformat=]
> > drivers/crypto/cavium/zip/zip_main.c:489:18: warning: format '%ld' expects argument of type 'long int', but argument 15 has type 'long long int' [-Wformat=]
> >
> > Introduced by commit
> >
> >?? 09ae5d37e093 ("crypto: zip - Add Compression/Decompression statistics")
> >
> > from the crypto tree interacting with commit
> >
> >?? 3f4ca3d25e1a ("asm-generic, x86: wrap atomic operations")
> >
> > from the akpm-current tree.
> >
> > This latter commit changed atomic64read() from "long" to "long long"
> > on x86_64.
>
> Hi,
>
> Previously atomic operations returned different types on 32/64 bits. I
> think that was pretty unfortunate and could actually lead to lots of
> such warnings (as there were no single format specifier that one could
> use portably), and was difficult to wrap. So I used "unsigned long
> long" for all operations.
> I actually suspect that the new zip_main.c code would cause a similar
> warning on 32 bits even without my change, because atomic ops already
> returned unsigned long long there.

>The driver depends on 64BIT (HW is 64 bit only).

>> So I think we need to fix zip_main.c.

>I'm not entirely sure, all drivers that are available for x86_64 and
>arm64 and use atomic64_read will show such warnings.

>One possible solution would be to disable COMPILE_TEST for the zip
>driver, but that would just hide the issue.

We found a decent solution to this issue and sent the patch to Herbert.
Solution:
Changing the format specifiers to %llu and adding a cast to u64 solves the
problem on all the architectures.

Thanks,
Mahipal