Below error can be seen when doing Yocto build of the toolstack:
| io.c: In function 'p9_error':
| io.c:684:5: error: ignoring return value of 'strerror_r' declared
with attribute 'warn_unused_result' [-Werror=unused-result]
| 684 | strerror_r(err, ring->buffer, ring->ring_size);
| | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| cc1: all warnings being treated as errors
Using strerror_r() without special casing different build environments
is impossible due to the different return types (int vs char *)
depending on the environment. As p9_error() is not on a performance
critical path, using strerror() with a mutex ought to be fine. So,
fix the build by using strerror() to replace strerror_r(). The steps
would then become: Acquire the mutex first, invoke strerror(), copy
the string from strerror() to the designated buffer and then drop the
mutex.
Fixes: f4900d6d69b5 ("9pfsd: allow building with old glibc")
Signed-off-by: Henry Wang <xin.wang2@amd.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
---
v3:
- Add in-code comment and commit msg to explain why use strerror_r().
- Cosmetic and style changes about the logic.
- Add Jan's Reviewed-by tag.
- Add a "Fixes" tag.
- Use a macro-defined value for the maximum error string length.
---
tools/9pfsd/io.c | 19 ++++++++++++++++++-
1 file changed, 18 insertions(+), 1 deletion(-)
diff --git a/tools/9pfsd/io.c b/tools/9pfsd/io.c
index adb887c7d9..df1be3df7d 100644
--- a/tools/9pfsd/io.c
+++ b/tools/9pfsd/io.c
@@ -677,11 +677,28 @@ static bool name_ok(const char *str)
return true;
}
+/* Including the '\0' */
+#define MAX_ERRSTR_LEN 80
static void p9_error(struct ring *ring, uint16_t tag, uint32_t err)
{
unsigned int erroff;
+ static pthread_mutex_t mutex = PTHREAD_MUTEX_INITIALIZER;
+ char *str;
+ size_t len = 0;
+
+ /*
+ * While strerror_r() exists, it comes in a POSIX and a GNU flavor.
+ * Let's try to avoid trying to be clever with determining which
+ * one it is that the underlying C library offers, when really we
+ * don't expect this function to be called very often.
+ */
+ pthread_mutex_lock(&mutex);
+ str = strerror(err);
+ len = min(strlen(str), (size_t)(MAX_ERRSTR_LEN - 1));
+ memcpy(ring->buffer, str, len);
+ ((char *)ring->buffer)[len] = '\0';
+ pthread_mutex_unlock(&mutex);
- strerror_r(err, ring->buffer, ring->ring_size);
erroff = add_string(ring, ring->buffer, strlen(ring->buffer));
fill_buffer(ring, P9_CMD_ERROR, tag, "SU",
erroff != ~0 ? ring->str + erroff : "cannot allocate memory",
--
2.34.1
On 07.03.24 14:56, Henry Wang wrote: > Below error can be seen when doing Yocto build of the toolstack: > > | io.c: In function 'p9_error': > | io.c:684:5: error: ignoring return value of 'strerror_r' declared > with attribute 'warn_unused_result' [-Werror=unused-result] > | 684 | strerror_r(err, ring->buffer, ring->ring_size); > | | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > | cc1: all warnings being treated as errors > > Using strerror_r() without special casing different build environments > is impossible due to the different return types (int vs char *) > depending on the environment. As p9_error() is not on a performance > critical path, using strerror() with a mutex ought to be fine. So, > fix the build by using strerror() to replace strerror_r(). The steps > would then become: Acquire the mutex first, invoke strerror(), copy > the string from strerror() to the designated buffer and then drop the > mutex. > > Fixes: f4900d6d69b5 ("9pfsd: allow building with old glibc") > Signed-off-by: Henry Wang <xin.wang2@amd.com> > Reviewed-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Juergen Gross <jgross@suse.com> Juergen
On 07.03.24 14:56, Henry Wang wrote: > Below error can be seen when doing Yocto build of the toolstack: > > | io.c: In function 'p9_error': > | io.c:684:5: error: ignoring return value of 'strerror_r' declared > with attribute 'warn_unused_result' [-Werror=unused-result] > | 684 | strerror_r(err, ring->buffer, ring->ring_size); > | | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > | cc1: all warnings being treated as errors > > Using strerror_r() without special casing different build environments > is impossible due to the different return types (int vs char *) > depending on the environment. As p9_error() is not on a performance > critical path, using strerror() with a mutex ought to be fine. So, > fix the build by using strerror() to replace strerror_r(). The steps > would then become: Acquire the mutex first, invoke strerror(), copy > the string from strerror() to the designated buffer and then drop the > mutex. > > Fixes: f4900d6d69b5 ("9pfsd: allow building with old glibc") > Signed-off-by: Henry Wang <xin.wang2@amd.com> > Reviewed-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Juergen Gross <jgross@suse.com> Juergen
On Thu, Mar 07, 2024 at 09:56:16PM +0800, Henry Wang wrote: > Below error can be seen when doing Yocto build of the toolstack: > > | io.c: In function 'p9_error': > | io.c:684:5: error: ignoring return value of 'strerror_r' declared > with attribute 'warn_unused_result' [-Werror=unused-result] > | 684 | strerror_r(err, ring->buffer, ring->ring_size); > | | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > | cc1: all warnings being treated as errors > > Using strerror_r() without special casing different build environments > is impossible due to the different return types (int vs char *) > depending on the environment. As p9_error() is not on a performance > critical path, using strerror() with a mutex ought to be fine. So, > fix the build by using strerror() to replace strerror_r(). The steps > would then become: Acquire the mutex first, invoke strerror(), copy > the string from strerror() to the designated buffer and then drop the > mutex. > > Fixes: f4900d6d69b5 ("9pfsd: allow building with old glibc") > Signed-off-by: Henry Wang <xin.wang2@amd.com> > Reviewed-by: Jan Beulich <jbeulich@suse.com> Acked-by: Anthony PERARD <anthony.perard@citrix.com> -- Anthony PERARD
© 2016 - 2024 Red Hat, Inc.