-Wignored-attributes in drivers/net/ethernet/intel/iavf/iavf_osdep.h #111

tpimh · 2018-09-18T13:46:38Z

drivers/net/ethernet/intel/i40evf/i40e_osdep.h:49:18: warning: 'format' attribute argument not supported: gnu_printf [-Wignored-attributes]
        __attribute__ ((format(gnu_printf, 3, 4)));
                        ^
1 warning generated.

The text was updated successfully, but these errors were encountered:

nickdesaulniers · 2018-10-03T08:13:13Z

Probably want to use @ojeda 's proper attribute detection patches, once they land in -next.

ojeda · 2018-10-08T09:46:17Z

They have now landed in linux-next.

ojeda · 2018-11-03T13:04:18Z

The compiler attributes series were merged into mainline yesterday, so we should get them for v4.20-rc1.

nathanchance · 2018-11-03T14:52:16Z

This is still an issue, we just can fix it now that we have the attributes tree merged.

nickdesaulniers · 2018-11-03T20:01:43Z

Is this in -next?

nathanchance · 2018-11-03T20:02:52Z

The warning?

nickdesaulniers · 2018-11-03T20:13:44Z

drivers/net/ethernet/intel/i40evf/i40e_osdep.h doesn't exist in torvalds/linux.

nathanchance · 2018-11-03T20:18:53Z

Got shuffled to drivers/net/ethernet/intel/iavf/i40e_osdep.h in commit 8062b22 ("intel-ethernet: rename i40evf to iavf") then to drivers/net/ethernet/intel/iavf/iavf_osdep.h in commit 66bc8e0 ("iavf: finish renaming files to iavf").

nathanchance · 2019-01-08T02:06:05Z

@ojeda how would you suggesting adding this to the compiler attributes file? Clang obviously supports __attribute__((__format__(...))), it just doesn't support __attribute__((__format__(gnu_printf, ...)))

I can only think of adding it based on CONFIG_CC_IS_GCC like so:

diff --git a/drivers/net/ethernet/intel/iavf/iavf_osdep.h b/drivers/net/ethernet/intel/iavf/iavf_osdep.h
index e6e0b0328706..a1cfabdd452c 100644
--- a/drivers/net/ethernet/intel/iavf/iavf_osdep.h
+++ b/drivers/net/ethernet/intel/iavf/iavf_osdep.h
@@ -46,7 +46,7 @@ struct iavf_virt_mem {

 #define iavf_debug(h, m, s, ...)  iavf_debug_d(h, m, s, ##__VA_ARGS__)
 extern void iavf_debug_d(void *hw, u32 mask, char *fmt_str, ...)
-       __attribute__ ((format(gnu_printf, 3, 4)));
+       __gnu_printf(3, 4);

 typedef enum iavf_status_code iavf_status;
 #endif /* _IAVF_OSDEP_H_ */
diff --git a/include/linux/compiler_attributes.h b/include/linux/compiler_attributes.h
index 19f32b0c29af..683c67a6c100 100644
--- a/include/linux/compiler_attributes.h
+++ b/include/linux/compiler_attributes.h
@@ -142,6 +142,11 @@
  */
 #define __printf(a, b)                  __attribute__((__format__(printf, a, b)))
 #define __scanf(a, b)                   __attribute__((__format__(scanf, a, b)))
+#ifdef CONFIG_CC_IS_GCC
+# define __gnu_printf(a, b)             __attribute__((__format__(gnu_printf, a, b)))
+#else
+# define __gnu_printf(a, b)
+#endif

 /*
  *   gcc: https://gcc.gnu.org/onlinedocs/gcc/Common-Function-Attributes.html#index-gnu_005finline-function-attribute

Any guidance would be appreciated :)

ojeda · 2019-01-09T09:56:04Z

One of the original goals of compiler_attributes.h was to clean up the #ifdefentanglements we had in compiler*.h files, so I wrote in the comments on top of the attributes header that we should avoid re-introducing them. So if this was really needed, we could simply leave it in compiler-gcc.h or make an exception (arguing that there is no way to test for support of the format options).

Having said that, there is a single usage of gnu_printf in the entire kernel, so we could say "just leave the warning in so that clang implements the option". That would be likely and desirable, since (from a quick glance at clang's code) it already implements GNU's extensions to printf.

Now, tracking down the actual usages through the iavf_debug macro, there are only around a dozen of them, and they all seem to be standard printf formatting strings. So, actually, we should simply change that to __printf. :)

nathanchance · 2019-01-10T02:50:07Z

As it turns out, GCC has had the printf attribute aliased to gnu_printf since at least 2010 (source):

/* Attributes such as "printf" are equivalent to those such as
   "gnu_printf" unless this is overridden by a target.  */
static const target_ovr_attr gnu_target_overrides_format_attributes[] =
{
  { "gnu_printf",   "printf" },
  { "gnu_scanf",    "scanf" },
  { "gnu_strftime", "strftime" },
  { "gnu_strfmon",  "strfmon" },
  { NULL,           NULL }
};

The formats only get overridden on Windows (so who cares amirite) so it should be totally fine to convert that to __printf. I'll draft up the commit now, thank you for looking into that @ojeda :)

nathanchance · 2019-01-10T04:24:12Z

Patch submitted: https://lore.kernel.org/lkml/20190110042157.8445-1-natechancellor@gmail.com/

Hopefully the wording is okay.

Clang warns: In file included from drivers/net/ethernet/intel/iavf/iavf_main.c:4: In file included from drivers/net/ethernet/intel/iavf/iavf.h:37: In file included from drivers/net/ethernet/intel/iavf/iavf_type.h:8: drivers/net/ethernet/intel/iavf/iavf_osdep.h:49:18: warning: 'format' attribute argument not supported: gnu_printf [-Wignored-attributes] __attribute__ ((format(gnu_printf, 3, 4))); ^ 1 warning generated. We can convert from gnu_printf to printf without any side effects for two reasons: 1. All iavf_debug instances use standard printf formats, as pointed out by Miguel Ojeda at the below link, meaning gnu_printf is not strictly required. 2. However, GCC has aliased printf to gnu_printf on Linux since at least 2010 based on git history. From gcc/c-family/c-format.c: /* Attributes such as "printf" are equivalent to those such as "gnu_printf" unless this is overridden by a target. */ static const target_ovr_attr gnu_target_overrides_format_attributes[] = { { "gnu_printf", "printf" }, { "gnu_scanf", "scanf" }, { "gnu_strftime", "strftime" }, { "gnu_strfmon", "strfmon" }, { NULL, NULL } }; The mentioned override only happens on Windows (mingw32). Changing from gnu_printf to printf is a no-op for GCC and stops Clang from warning. Link: ClangBuiltLinux#111 Suggested-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com> Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>

nathanchance · 2019-01-15T17:37:47Z

Patch accepted: https://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue.git/commit/?=id=80f42d0bad4de147e27c43aec9f4e009c1e0a28e

Clang warns: In file included from drivers/net/ethernet/intel/iavf/iavf_main.c:4: In file included from drivers/net/ethernet/intel/iavf/iavf.h:37: In file included from drivers/net/ethernet/intel/iavf/iavf_type.h:8: drivers/net/ethernet/intel/iavf/iavf_osdep.h:49:18: warning: 'format' attribute argument not supported: gnu_printf [-Wignored-attributes] __attribute__ ((format(gnu_printf, 3, 4))); ^ 1 warning generated. We can convert from gnu_printf to printf without any side effects for two reasons: 1. All iavf_debug instances use standard printf formats, as pointed out by Miguel Ojeda at the below link, meaning gnu_printf is not strictly required. 2. However, GCC has aliased printf to gnu_printf on Linux since at least 2010 based on git history. From gcc/c-family/c-format.c: /* Attributes such as "printf" are equivalent to those such as "gnu_printf" unless this is overridden by a target. */ static const target_ovr_attr gnu_target_overrides_format_attributes[] = { { "gnu_printf", "printf" }, { "gnu_scanf", "scanf" }, { "gnu_strftime", "strftime" }, { "gnu_strfmon", "strfmon" }, { NULL, NULL } }; The mentioned override only happens on Windows (mingw32). Changing from gnu_printf to printf is a no-op for GCC and stops Clang from warning. Link: ClangBuiltLinux#111 Suggested-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com> Signed-off-by: Nathan Chancellor <natechancellor@gmail.com> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>

nickdesaulniers · 2019-03-19T18:07:39Z

@nathanchance what happened to that patch? Broken link

nathanchance · 2019-03-19T18:12:16Z

The dev-queue branch gets constantly rebased until it gets sent to net-next it seems.

Current hash: https://git.kernel.org/jkirsher/next-queue/c/9e5f261679e81a027bb33efbbb2720aac80e2fe7

Clang warns: In file included from drivers/net/ethernet/intel/iavf/iavf_main.c:4: In file included from drivers/net/ethernet/intel/iavf/iavf.h:37: In file included from drivers/net/ethernet/intel/iavf/iavf_type.h:8: drivers/net/ethernet/intel/iavf/iavf_osdep.h:49:18: warning: 'format' attribute argument not supported: gnu_printf [-Wignored-attributes] __attribute__ ((format(gnu_printf, 3, 4))); ^ 1 warning generated. We can convert from gnu_printf to printf without any side effects for two reasons: 1. All iavf_debug instances use standard printf formats, as pointed out by Miguel Ojeda at the below link, meaning gnu_printf is not strictly required. 2. However, GCC has aliased printf to gnu_printf on Linux since at least 2010 based on git history. From gcc/c-family/c-format.c: /* Attributes such as "printf" are equivalent to those such as "gnu_printf" unless this is overridden by a target. */ static const target_ovr_attr gnu_target_overrides_format_attributes[] = { { "gnu_printf", "printf" }, { "gnu_scanf", "scanf" }, { "gnu_strftime", "strftime" }, { "gnu_strfmon", "strfmon" }, { NULL, NULL } }; The mentioned override only happens on Windows (mingw32). Changing from gnu_printf to printf is a no-op for GCC and stops Clang from warning. Link: ClangBuiltLinux#111 Suggested-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com> Signed-off-by: Nathan Chancellor <natechancellor@gmail.com> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>

nickdesaulniers · 2019-05-13T21:53:54Z

link to the tree instead: https://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue.git/log/?h=dev-queue

Clang warns: In file included from drivers/net/ethernet/intel/iavf/iavf_main.c:4: In file included from drivers/net/ethernet/intel/iavf/iavf.h:37: In file included from drivers/net/ethernet/intel/iavf/iavf_type.h:8: drivers/net/ethernet/intel/iavf/iavf_osdep.h:49:18: warning: 'format' attribute argument not supported: gnu_printf [-Wignored-attributes] __attribute__ ((format(gnu_printf, 3, 4))); ^ 1 warning generated. We can convert from gnu_printf to printf without any side effects for two reasons: 1. All iavf_debug instances use standard printf formats, as pointed out by Miguel Ojeda at the below link, meaning gnu_printf is not strictly required. 2. However, GCC has aliased printf to gnu_printf on Linux since at least 2010 based on git history. From gcc/c-family/c-format.c: /* Attributes such as "printf" are equivalent to those such as "gnu_printf" unless this is overridden by a target. */ static const target_ovr_attr gnu_target_overrides_format_attributes[] = { { "gnu_printf", "printf" }, { "gnu_scanf", "scanf" }, { "gnu_strftime", "strftime" }, { "gnu_strfmon", "strfmon" }, { NULL, NULL } }; The mentioned override only happens on Windows (mingw32). Changing from gnu_printf to printf is a no-op for GCC and stops Clang from warning. Link: ClangBuiltLinux#111 Suggested-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com> Signed-off-by: Nathan Chancellor <natechancellor@gmail.com> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>

Clang warns: In file included from drivers/net/ethernet/intel/iavf/iavf_main.c:4: In file included from drivers/net/ethernet/intel/iavf/iavf.h:37: In file included from drivers/net/ethernet/intel/iavf/iavf_type.h:8: drivers/net/ethernet/intel/iavf/iavf_osdep.h:49:18: warning: 'format' attribute argument not supported: gnu_printf [-Wignored-attributes] __attribute__ ((format(gnu_printf, 3, 4))); ^ 1 warning generated. We can convert from gnu_printf to printf without any side effects for two reasons: 1. All iavf_debug instances use standard printf formats, as pointed out by Miguel Ojeda at the below link, meaning gnu_printf is not strictly required. 2. However, GCC has aliased printf to gnu_printf on Linux since at least 2010 based on git history. From gcc/c-family/c-format.c: /* Attributes such as "printf" are equivalent to those such as "gnu_printf" unless this is overridden by a target. */ static const target_ovr_attr gnu_target_overrides_format_attributes[] = { { "gnu_printf", "printf" }, { "gnu_scanf", "scanf" }, { "gnu_strftime", "strftime" }, { "gnu_strfmon", "strfmon" }, { NULL, NULL } }; The mentioned override only happens on Windows (mingw32). Changing from gnu_printf to printf is a no-op for GCC and stops Clang from warning. Link: ClangBuiltLinux#111 Suggested-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com> Signed-off-by: Nathan Chancellor <natechancellor@gmail.com> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

Clang warns: In file included from drivers/net/ethernet/intel/iavf/iavf_main.c:4: In file included from drivers/net/ethernet/intel/iavf/iavf.h:37: In file included from drivers/net/ethernet/intel/iavf/iavf_type.h:8: drivers/net/ethernet/intel/iavf/iavf_osdep.h:49:18: warning: 'format' attribute argument not supported: gnu_printf [-Wignored-attributes] __attribute__ ((format(gnu_printf, 3, 4))); ^ 1 warning generated. We can convert from gnu_printf to printf without any side effects for two reasons: 1. All iavf_debug instances use standard printf formats, as pointed out by Miguel Ojeda at the below link, meaning gnu_printf is not strictly required. 2. However, GCC has aliased printf to gnu_printf on Linux since at least 2010 based on git history. From gcc/c-family/c-format.c: /* Attributes such as "printf" are equivalent to those such as "gnu_printf" unless this is overridden by a target. */ static const target_ovr_attr gnu_target_overrides_format_attributes[] = { { "gnu_printf", "printf" }, { "gnu_scanf", "scanf" }, { "gnu_strftime", "strftime" }, { "gnu_strfmon", "strfmon" }, { NULL, NULL } }; The mentioned override only happens on Windows (mingw32). Changing from gnu_printf to printf is a no-op for GCC and stops Clang from warning. Link: ClangBuiltLinux/linux#111 Suggested-by: Miguel Ojeda <miguel.ojeda.sandonis@gmail.com> Signed-off-by: Nathan Chancellor <natechancellor@gmail.com> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Andrew Bowers <andrewx.bowers@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

nathanchance · 2019-06-01T06:27:09Z

https://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next.git/commit/?id=3dc215416673f9ba8f136aafbb0069ed9ebff376

Finally this has been picked up into net-next for 5.3!

nathanchance · 2019-07-11T18:54:07Z

Merged into mainline: https://git.kernel.org/torvalds/c/3dc215416673f9ba8f136aafbb0069ed9ebff376

tpimh added help wanted good first issue Good for newcomers [BUG] linux A bug that should be fixed in the mainline kernel. low priority This bug is not critical and not a priority labels Sep 18, 2018

nickdesaulniers added -Wignored-attributes and removed help wanted labels Sep 24, 2018

tpimh closed this as completed Nov 3, 2018

nathanchance removed the [FIXED][LINUX] 4.20 This bug was fixed in Linux 4.20 label Nov 3, 2018

nathanchance reopened this Nov 3, 2018

nickdesaulniers added the good first issue Good for newcomers label Nov 3, 2018

nickdesaulniers removed the good first issue Good for newcomers label Jan 8, 2019

tpimh changed the title ~~-Wignored-attributes in drivers/net/ethernet/intel/i40evf/i40e_osdep.h~~ -Wignored-attributes in drivers/net/ethernet/intel/iavf/iavf_osdep.h Jan 8, 2019

nathanchance added the [PATCH] Submitted A patch has been submitted for review label Jan 10, 2019

nathanchance added [PATCH] Accepted A submitted patch has been accepted upstream and removed [PATCH] Submitted A patch has been submitted for review labels Jan 15, 2019

nathanchance closed this as completed Jul 11, 2019

nathanchance added [FIXED][LINUX] 5.3 This bug was fixed in Linux 5.3 and removed [PATCH] Accepted A submitted patch has been accepted upstream labels Jul 11, 2019

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

-Wignored-attributes in drivers/net/ethernet/intel/iavf/iavf_osdep.h #111

-Wignored-attributes in drivers/net/ethernet/intel/iavf/iavf_osdep.h #111

tpimh commented Sep 18, 2018

nickdesaulniers commented Oct 3, 2018

ojeda commented Oct 8, 2018

ojeda commented Nov 3, 2018

nathanchance commented Nov 3, 2018

nickdesaulniers commented Nov 3, 2018

nathanchance commented Nov 3, 2018

nickdesaulniers commented Nov 3, 2018

nathanchance commented Nov 3, 2018

nathanchance commented Jan 8, 2019

ojeda commented Jan 9, 2019 •

edited

Loading

nathanchance commented Jan 10, 2019 •

edited

Loading

nathanchance commented Jan 10, 2019

nathanchance commented Jan 15, 2019

nickdesaulniers commented Mar 19, 2019

nathanchance commented Mar 19, 2019

nickdesaulniers commented May 13, 2019

nathanchance commented Jun 1, 2019

nathanchance commented Jul 11, 2019

-Wignored-attributes in drivers/net/ethernet/intel/iavf/iavf_osdep.h #111

-Wignored-attributes in drivers/net/ethernet/intel/iavf/iavf_osdep.h #111

Comments

tpimh commented Sep 18, 2018

nickdesaulniers commented Oct 3, 2018

ojeda commented Oct 8, 2018

ojeda commented Nov 3, 2018

nathanchance commented Nov 3, 2018

nickdesaulniers commented Nov 3, 2018

nathanchance commented Nov 3, 2018

nickdesaulniers commented Nov 3, 2018

nathanchance commented Nov 3, 2018

nathanchance commented Jan 8, 2019

ojeda commented Jan 9, 2019 • edited Loading

nathanchance commented Jan 10, 2019 • edited Loading

nathanchance commented Jan 10, 2019

nathanchance commented Jan 15, 2019

nickdesaulniers commented Mar 19, 2019

nathanchance commented Mar 19, 2019

nickdesaulniers commented May 13, 2019

nathanchance commented Jun 1, 2019

nathanchance commented Jul 11, 2019

ojeda commented Jan 9, 2019 •

edited

Loading

nathanchance commented Jan 10, 2019 •

edited

Loading