[Bug runtime/29076] New: user_string_quoted() doesn't properly generate strings in .rodata on x86_64 machines

[Bug runtime/29076] New: user_string_quoted() doesn't properly generate strings in .rodata on x86_64 machines

[wcohen at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29076

            Bug ID: 29076
           Summary: user_string_quoted() doesn't properly generate strings
                    in .rodata on x86_64 machines
           Product: systemtap
           Version: unspecified
            Status: NEW
          Severity: normal
          Priority: P2
         Component: runtime
          Assignee: systemtap at sourceware dot org
          Reporter: wcohen at redhat dot com
  Target Milestone: ---

Created attachment 14073
  --> https://sourceware.org/bugzilla/attachment.cgi?id=14073&action=edit
User-space and systemtap script showing issue with .rodata strings

When triaging the testsuite systemtap.syscall/syscall.exp failures I found that
strings that were stored in the .rodata section were not properly printed on
x86_64 fedora 35 machine.  In their place was the pointer value passed in for
the string location.

Have a little example demonstrating the issue in x86_ro_string.tar.gz:
$ uname -a
Linux fedora 5.16.20-200.fc35.x86_64 #1 SMP PREEMPT Wed Apr 13 22:09:20 UTC
2022 x86_64 x86_64 x86_64 GNU/Linux
$ gcc -o acct2 -O2 acct2.c
$ sudo stap -c ./acct2 sys_acct.stp 
[sudo] password for wcohen: 
kprobe.function("__x64_sys_acct")?(0x402010)
kprobe.function("__x64_sys_acct")?("foobar")
kprobe.function("__x64_sys_acct")?(0xffffffffffffffff)

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug runtime/29076] user_string_quoted() doesn't properly generate strings in .rodata on x86_64 machines

[wcohen at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29076

--- Comment #1 from William Cohen <wcohen at redhat dot com> ---
Tried the example on non-x86 platforms such as aarch64 and it gave the expected
result:

[wcohen@fedora ~]$ uname -a
Linux fedora 5.16.20-200.fc35.aarch64 #1 SMP Wed Apr 13 21:49:21 UTC 2022
aarch64 aarch64 aarch64 GNU/Linux
[wcohen@fedora ~]$ gcc -o acct2 -O2 acct2.c 
[wcohen@fedora ~]$ sudo stap -c ./acct2 sys_acct.stp
warning: the compiler differs from the one used to build the kernel
  The kernel was built by: gcc (GCC) 11.2.1 20220127 (Red Hat 11.2.1-9)
  You are using:           gcc (GCC) 11.2.1 20220401 (Red Hat 11.2.1-10)
kprobe.function("__arm64_sys_acct")?("foobar")
kprobe.function("__arm64_sys_acct")?("foobar")
kprobe.function("__arm64_sys_acct")?(0xffffffffffffffff)

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug runtime/29076] user_string_quoted() doesn't properly generate strings in .rodata on x86_64 machines

[wcohen at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29076

William Cohen <wcohen at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mcermak at redhat dot com

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug runtime/29076] user_string_quoted() doesn't properly generate strings in .rodata on x86_64 machines

[fche at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29076

--- Comment #2 from Frank Ch. Eigler <fche at redhat dot com> ---
Could this simply be a paging issue?  Or does the test case cause the rodata
bits to be properly paged in, before the test syscalls are fired?  Maybe it
should call mlockall() at startup.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug runtime/29076] user_string_quoted() doesn't properly generate strings in .rodata on x86_64 machines

[wcohen at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29076

--- Comment #3 from William Cohen <wcohen at redhat dot com> ---
That appears to be the case on x86_64.  The question is how to get the syscall
tests to work better.  Have two examples that only differ with where the strcpy
is done.  One result has a pointer and the other prints the string:

[wcohen@fedora x86_ro_string]$ cat acct4.c
#include <unistd.h>
#include <string.h>

char *a="foobar";
char buffer[32];

int main()
{
  acct(a);
  strcpy (&buffer[0], a);
  acct(&buffer[0]);
  acct((char *)-1);
  return 0;
}
[wcohen@fedora x86_ro_string]$ gcc -o acct5 -O2 acct5.c
[wcohen@fedora x86_ro_string]$ sudo stap -c ./acct4 sys_acct.stp 
[sudo] password for wcohen: 
kprobe.function("__x64_sys_acct")?(0x402010)
kprobe.function("__x64_sys_acct")?("foobar")
kprobe.function("__x64_sys_acct")?(0xffffffffffffffff)
[wcohen@fedora x86_ro_string]$ cat acct5.c
#include <unistd.h>
#include <string.h>

char *a="foobar";
char buffer[32];

int main()
{
  strcpy (&buffer[0], a);
  acct(a);
  acct(&buffer[0]);
  acct((char *)-1);
  return 0;
}
[wcohen@fedora x86_ro_string]$ gcc -o acct5 -O2 acct5.c
[wcohen@fedora x86_ro_string]$ sudo stap -c ./acct5 sys_acct.stp 
kprobe.function("__x64_sys_acct")?("foobar")
kprobe.function("__x64_sys_acct")?("foobar")
kprobe.function("__x64_sys_acct")?(0xffffffffffffffff)
[wcohen@fedora x86_ro_string]$

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug runtime/29076] user_string_quoted() doesn't properly generate strings in .rodata on x86_64 machines

[fche at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29076

Frank Ch. Eigler <fche at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fche at redhat dot com

--- Comment #4 from Frank Ch. Eigler <fche at redhat dot com> ---
OK, that seems to confirm it's just demand-paging artifact.
Try adding an  mlockall(MCL_CURRENT)  syscall at the top of main().

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug runtime/29076] user_string_quoted() doesn't properly generate strings in .rodata on x86_64 machines

[wcohen at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29076

--- Comment #5 from William Cohen <wcohen at redhat dot com> ---
The mlockall allows the access to the string in the readonly section.

[wcohen@fedora x86_ro_string]$ cat acct6.c
#include <unistd.h>
#include <string.h>
#include <sys/mman.h>

char *a="foobar";
char buffer[32];

int main()
{
  mlockall(MCL_CURRENT);
  acct(a);
  strcpy (&buffer[0], a);
  acct(&buffer[0]);
  acct((char *)-1);
  return 0;
}
[wcohen@fedora x86_ro_string]$ gcc -o acct6 -O2 acct6.c
[wcohen@fedora x86_ro_string]$ sudo stap -c ./acct6 sys_acct.stp 
kprobe.function("__x64_sys_acct")?("foobar")
kprobe.function("__x64_sys_acct")?("foobar")
kprobe.function("__x64_sys_acct")?(0xffffffffffffffff)

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug runtime/29076] user_string_quoted() doesn't properly generate strings in .rodata on x86_64 machines

[mcermak at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29076

--- Comment #6 from Martin Cermak <mcermak at redhat dot com> ---
Created attachment 14075
  --> https://sourceware.org/bugzilla/attachment.cgi?id=14075&action=edit
proposed patch

Thank you for looking into this.  The attached patch based on your findings
already significantly improves the situation for the affected testcases.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug runtime/29076] user_string_quoted() doesn't properly generate strings in .rodata on x86_64 machines

[mcermak at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=29076

Martin Cermak <mcermak at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |FIXED

--- Comment #7 from Martin Cermak <mcermak at redhat dot com> ---
Fixed in commit 6577f2237368a39a5fb289f7deef0cf73b7d290b

-- 
You are receiving this mail because:
You are the assignee for the bug.