handle xlat_styles when decoding sa_mask

Dmitry V. Levin ldv at altlinux.org
Sun Mar 17 22:42:11 UTC 2019


Hi,

On Sun, Feb 17, 2019 at 12:43:59PM -0600, shankarapailoor wrote:
> Hi,
> 
> > I hope you understand that the output produced by your parser
> > is different between big-endian and little-endian architectures.
> 
> Yes I understood. How do you think we should decode the sa_mask field under
> XLAT_RAW?

There seems to be no obvious way of printing sa_mask under XLAT_RAW,
so the question is how to print sa_mask under XLAT_RAW
in the least confusing way.

Since sa_mask is implemented as an array of target long integers,
what if we printed sa_mask as an array of hexadecimal values with explicit
length, e.g. tprintf("%#0*" PRI_klx, (int) elem_size * 2 + 2, elem_value)?

> On Fri, Feb 8, 2019 at 1:46 PM Dmitry V. Levin <ldv at altlinux.org> wrote:
> 
> > On Fri, Feb 08, 2019 at 10:32:52PM +0300, Dmitry V. Levin wrote:
> > > Hi,
> > >
> > > On Sat, Jan 05, 2019 at 12:10:38PM -0800, shankarapailoor wrote:
> > > > Hi,
> > > >
> > > > Strace doesn't handle xlat styles properly when decoding the sa_mask
> > > > field of the sigaction struct,  Attached is a patch which decodes the
> > > > field as an array of bytes. Let me know what you think!
> > > >
> > > > Regards,
> > > > Shankara Pailoor
> > >
> > > > From 41d7cac5cf000f4e8b6ee675d231038cb786d2e1 Mon Sep 17 00:00:00 2001
> > > > From: Shankara Pailoor <shankarapailoor at gmail.com>
> > > > Date: Sat, 5 Jan 2019 11:49:28 -0800
> > > > Subject: [PATCH v1] properly handle xlat_styles when decoding
> > sigaction mask
> > > >
> > > > * signal.c (sprintsigmask_n): Decode sigaction mask as array of bytes
> > under XLAT_RAW, VERBOSE.
> > > > * tests/printsamask.c: New file.
> > > > * tests/printsamask-Xabbrev.c: Likewise.
> > > > * tests/printsamask-Xraw.c: Likewise.
> > > > * tests/printsamask-Xverbose.c: Likewise.
> > > > * tests/pure_executables.list: Add printsamask-Xabbrev,
> > printsamask-Xraw, printsamask-Xverbose.
> > > > * tests/gen_tests.in: Likewise.
> > > > ---
> > > >  signal.c                     | 32 +++++++++++++-
> > > >  tests/.gitignore             |  3 ++
> > > >  tests/gen_tests.in           |  3 ++
> > > >  tests/printsamask-Xabbrev.c  |  1 +
> > > >  tests/printsamask-Xraw.c     |  2 +
> > > >  tests/printsamask-Xverbose.c |  2 +
> > > >  tests/printsamask.c          | 81 ++++++++++++++++++++++++++++++++++++
> > > >  tests/pure_executables.list  |  3 ++
> > > >  8 files changed, 126 insertions(+), 1 deletion(-)
> > > >  create mode 100644 tests/printsamask-Xabbrev.c
> > > >  create mode 100644 tests/printsamask-Xraw.c
> > > >  create mode 100644 tests/printsamask-Xverbose.c
> > > >  create mode 100644 tests/printsamask.c
> > > >
> > > > diff --git a/signal.c b/signal.c
> > > > index 418905f0..e3c5d8c3 100644
> > > > --- a/signal.c
> > > > +++ b/signal.c
> > > > @@ -166,8 +166,11 @@ sprintsigmask_n(const char *prefix, const void
> > *sig_mask, unsigned int bytes)
> > > >      * Most of signal names have length 7,
> > > >      * average length of signal names is less than 7.
> > > >      * The length of prefix string does not exceed 16.
> > > > +    * Under xlat_verbose we may have at most NSIG_BYTES in
> > > > +    * the raw sa_mask array. We can safely allocate 8 bytes per entry
> > > >      */
> > > > -   static char outstr[128 + 8 * (NSIG_BYTES * 8 * 2 / 3)];
> > > > +   static char outstr[128 + 8 * (NSIG_BYTES * 8 * 2 / 3)
> > > > +                           + 8 * NSIG_BYTES];
> > > >
> > > >     char *s;
> > > >     const uint32_t *mask;
> > > > @@ -178,6 +181,29 @@ sprintsigmask_n(const char *prefix, const void
> > *sig_mask, unsigned int bytes)
> > > >
> > > >     s = stpcpy(outstr, prefix);
> > > >
> > > > +   if (xlat_verbose(xlat_verbosity) != XLAT_STYLE_ABBREV) {
> > > > +           const char *mask = (char *) sig_mask;
> > > > +           const char *sep = "[";
> > > > +           unsigned int i, size;
> > > > +
> > > > +           size = (bytes >= NSIG_BYTES) ? NSIG_BYTES : bytes;
> > > > +           for (i = 0; i < size; i++) {
> > > > +                   s = xappendstr(outstr, s, "%s0x%02x", sep,
> > mask[i]);
> > > > +                   sep = ", ";
> > > > +           }
> > >
> > > I hope you understand that the output produced by your parser
> > > is different between big-endian and little-endian architectures.
> > >
> > > [...]
> > > > +int
> > > > +main(void)
> > > > +{
> > > > +   unsigned int set_size = NSIG / 8;
> > > > +   void *const k_set = tail_alloc(set_size);
> > > > +   void *const old_set = tail_alloc(set_size);
> > > > +   TAIL_ALLOC_OBJECT_CONST_PTR(sigset_t, libc_set);
> > > > +
> > > > +   memset(k_set, 0, set_size);
> > > > +   sigemptyset(libc_set);
> > > > +   sigaddset(libc_set, SIGHUP);
> > > > +   memcpy(k_set, libc_set, set_size);
> > > > +
> > > > +   int rc = k_sigprocmask(SIG_SETMASK, k_set, old_set, set_size);
> > > > +#if XLAT_RAW
> > > > +   tprintf("rt_sigprocmask(0x2"
> > > > +           ", [0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00]"
> > > > +           ", [0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00]"
> > > > +           ", %u) = %s\n", set_size, sprintrc(rc));
> > > > +#elif XLAT_VERBOSE
> > > > +   tprintf("rt_sigprocmask(0x2 /* SIG_SETMASK */"
> > > > +           ", [0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00] /*
> > [HUP] */"
> > > > +           ", [0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00] /* []
> > */, %u) = %s\n",
> > > > +                   set_size, sprintrc(rc));
> > > > +#else /* XLAT_ABBREV */
> > > > +   tprintf("rt_sigprocmask(SIG_SETMASK, [HUP], [], %u) = %s\n",
> > > > +                   set_size, sprintrc(rc));
> > > > +#endif
> > >
> > > Your test assumes that the output is produced on a big-endian
> >
> > s/big-endian/little-endian/
> >
> > > architecture, so it fails on big-endian architectures.


-- 
ldv
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: not available
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20190318/b3962701/attachment.bin>


More information about the Strace-devel mailing list