From: Steven Rostedt <rostedt@goodmis.org>
To: Linux Trace Devel <linux-trace-devel@vger.kernel.org>
Cc: Ross Zwisler <zwisler@google.com>,
Stevie Alvarez <stevie.6strings@gmail.com>
Subject: [PATCH] libtraceeval: Use trick to force static array usage where needed
Date: Wed, 4 Oct 2023 18:52:25 -0400 [thread overview]
Message-ID: <20231004185225.366c1739@gandalf.local.home> (raw)
From: "Steven Rostedt (Google)" <rostedt@goodmis.org>
Some functions use the TRACEEVAL_ARRAY_SIZE() macro to determine the size
of the array passed to it. But this will not work if the developer uses a
pointer to the array. gcc may give a warning, but it will still happily
compile it leaving the developer wondering why their code does not work.
Use a little trick that checks and tests if the array is static, and will
fail the build if it is not.
#define TRACEEVAL_ARRAY_SIZE(data) \
((sizeof(data) / sizeof((data)[0])) + \
(int)(sizeof(struct { \
int:(-!!(__builtin_types_compatible_p(typeof(data), \
typeof(&((data)[0]))))); \
})))
Going backwards to explain the above, we have:
__builtin_types_compatible_p(typeof(data), typeof(&((data)[0]))
Which is a gcc/clang compiler directive that returns true if the two
pointers are compatible [ (a, b) where a = b is valid ]. For a static
array, we would have
struct traceeval_data data[5]
Where typeof(data) is "struct traceeval_data []" and the type of
&data[0] is a pointer to "struct traceeval_data", and the above
would return false (zero) [ data = &data[0] is invalid ].
For pointers:
struct traceeval_data *data;
Then type of data is the same as &data[0] and it would return true (1).
[ data = &data[0] is valid ]
Now we have a structure defined:
struct {
int: (-!!(<result>));
}
Which if <result> of the __builtin_types_compatible_p() returned false
(zero), then it would be:
struct {
int: 0; // structure with int size of 0 bits
}
Which is perfectly valid. But if <result> returned true (as it would if it
was a pointer and not a static array), then it would be:
struct {
int: -1; // structure with int size of -1 bits!
}
Which is not valid to compile, and the build will fail.
But in order to make sure this is in the code that uses
TRACEEVAL_ARRAY_SIZE(), it needs to be part of the computation. To do
that, as "struct { int:0; }" is of size zero, we can simply add a sizeof()
around it, and attach the above with an addition "+".
... + sizeof((int)(sizeof(struct { int:0;}))) is the same as:
... + 0
Now with this logic, if a pointer is passed to something like
traceeval_init(), it will fail to build, and not cause hours of scratching
head debugging for the developer at runtime.
Of course, the developer will likely now scratch their head on why it
doesn't build, but that's because they didn't RTFM!
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
---
include/traceeval.h | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/include/traceeval.h b/include/traceeval.h
index 4cc5eb6ef3de..6c6e09c53129 100644
--- a/include/traceeval.h
+++ b/include/traceeval.h
@@ -17,7 +17,12 @@
/* Field name/descriptor for number of hits */
#define TRACEEVAL_VAL_HITS ((const char *)(-1UL))
-#define TRACEEVAL_ARRAY_SIZE(data) (sizeof(data) / sizeof((data)[0]))
+#define TRACEEVAL_ARRAY_SIZE(data) \
+ ((sizeof(data) / sizeof((data)[0])) + \
+ (int)(sizeof(struct { \
+ int:(-!!(__builtin_types_compatible_p(typeof(data), \
+ typeof(&((data)[0]))))); \
+ })))
/* Data type distinguishers */
enum traceeval_data_type {
--
2.40.1
next reply other threads:[~2023-10-04 22:51 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-10-04 22:52 Steven Rostedt [this message]
2023-10-05 21:25 ` [PATCH] libtraceeval: Use trick to force static array usage where needed Ross Zwisler
2023-10-05 21:59 ` Steven Rostedt
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20231004185225.366c1739@gandalf.local.home \
--to=rostedt@goodmis.org \
--cc=linux-trace-devel@vger.kernel.org \
--cc=stevie.6strings@gmail.com \
--cc=zwisler@google.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).