Skip to contentSkip to conversation

Document kernel bugs in delivery of signals from CPU exceptions

PATCH In review Zack Weinberg 2 participants 2 messages Dec 23, 2019 – Apr 9, 2020 Man-pages: manual pages for Linux -- Sections 2, 3, 4, 5, and 7

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Document kernel bugs in delivery of signals from CPU exceptions

signal.7: Which signal is delivered in response to a CPU exception is under-documented and does not always make sense. See
<https://bugzilla.kernel.org/show_bug.cgi?id=205831> for an example where it doesn’t make sense; per the discussion there, this cannot be changed because of backward compatibility concerns, so let’s instead document the problem.

sigaction.2: For related reasons, the kernel doesn’t always fill in all of the fields of the siginfo_t when delivering signals from CPU exceptions. Document this as well. I imagine this one _could_ be fixed, but the problem would still be relevant to anyone using an older kernel.

---
 man2/sigaction.2 |  8 ++++++++
 man7/signal.7    | 40 ++++++++++++++++++++++++++++++----------
 2 files changed, 38 insertions(+), 10 deletions(-)
man2/sigaction.2 +8 -0
index 8ee878672..10d1c4882 100644
@@ -1020,6 +1020,14 @@ handler.
 See the relevant Linux kernel sources for details.
 This use is obsolete now.
 .SH BUGS
+When delivering a signal with a
+.B SA_SIGINFO
+handler,
+the kernel does not always provide meaningful values
+for all of the fields of the
+.I siginfo_t
+that are relevant for that signal.
+.PP
 In kernels up to and including 2.6.13, specifying
 .B SA_NODEFER
 in
man7/signal.7 +30 -10
index d34e536f1..a9fe076fd 100644
@@ -796,16 +796,36 @@ Linux 2.4 and earlier:
 .BR nanosleep (2).
 .SH CONFORMING TO
 POSIX.1, except as noted.
-.\" It must be a *very* long time since this was true:
-.\" .SH BUGS
-.\" .B SIGIO
-.\" and
-.\" .B SIGLOST
-.\" have the same value.
-.\" The latter is commented out in the kernel source, but
-.\" the build process of some software still thinks that
-.\" signal 29 is
-.\" .BR SIGLOST .
+.SH BUGS
+There are six signals that can be delivered
+as a consequence of a hardware exception:
+.BR SIGBUS ,
+.BR SIGEMT ,
+.BR SIGFPE ,
+.BR SIGILL ,
+.BR SIGSEGV ,
+and
+.BR SIGTRAP .
+Which of these signals is delivered,
+for any given hardware exception,
+is not documented and does not always make sense.
+.PP
+For example, an invalid memory access that causes delivery of
+.B SIGSEGV
+on one CPU architecture may cause delivery of
+.B SIGBUS
+on another architecture, or vice versa.
+.PP
+For another example, using the x86
+.I int
+instruction with a forbidden argument
+(any number other than 3 or 128)
+causes delivery of
+.BR SIGSEGV ,
+even though
+.B SIGILL
+would make more sense,
+because of how the CPU reports the forbidden operation to the kernel.
 .SH NOTES
 For a discussion of async-signal-safe functions, see
 .BR signal-safety (7).
Signature · 1 line
2.24.1
#1 lore raw

Helolo Zack,

Quoted 7 lines · Zack Weinberg

signal.7: Which signal is delivered in response to a CPU exception is under-documented and does not always make sense. See
<https://bugzilla.kernel.org/show_bug.cgi?id=205831> for an example where it doesn’t make sense; per the discussion there, this cannot be changed because of backward compatibility concerns, so let’s instead document the problem.

Quoted 5 lines

sigaction.2: For related reasons, the kernel doesn’t always fill in all of the fields of the siginfo_t when delivering signals from CPU exceptions. Document this as well. I imagine this one _could_ be fixed, but the problem would still be relevant to anyone using an older kernel.

Thanks. I've applied this patch.

Cheers,

Michael

Quoted 76 lines
---
 man2/sigaction.2 |  8 ++++++++
 man7/signal.7    | 40 ++++++++++++++++++++++++++++++----------
 2 files changed, 38 insertions(+), 10 deletions(-)
man2/sigaction.2 +8 -0
index 8ee878672..10d1c4882 100644
@@ -1020,6 +1020,14 @@ handler.
 See the relevant Linux kernel sources for details.
 This use is obsolete now.
 .SH BUGS
+When delivering a signal with a
+.B SA_SIGINFO
+handler,
+the kernel does not always provide meaningful values
+for all of the fields of the
+.I siginfo_t
+that are relevant for that signal.
+.PP
 In kernels up to and including 2.6.13, specifying
 .B SA_NODEFER
 in
man7/signal.7 +30 -10
index d34e536f1..a9fe076fd 100644
@@ -796,16 +796,36 @@ Linux 2.4 and earlier:
 .BR nanosleep (2).
 .SH CONFORMING TO
 POSIX.1, except as noted.
-.\" It must be a *very* long time since this was true:
-.\" .SH BUGS
-.\" .B SIGIO
-.\" and
-.\" .B SIGLOST
-.\" have the same value.
-.\" The latter is commented out in the kernel source, but
-.\" the build process of some software still thinks that
-.\" signal 29 is
-.\" .BR SIGLOST .
+.SH BUGS
+There are six signals that can be delivered
+as a consequence of a hardware exception:
+.BR SIGBUS ,
+.BR SIGEMT ,
+.BR SIGFPE ,
+.BR SIGILL ,
+.BR SIGSEGV ,
+and
+.BR SIGTRAP .
+Which of these signals is delivered,
+for any given hardware exception,
+is not documented and does not always make sense.
+.PP
+For example, an invalid memory access that causes delivery of
+.B SIGSEGV
+on one CPU architecture may cause delivery of
+.B SIGBUS
+on another architecture, or vice versa.
+.PP
+For another example, using the x86
+.I int
+instruction with a forbidden argument
+(any number other than 3 or 128)
+causes delivery of
+.BR SIGSEGV ,
+even though
+.B SIGILL
+would make more sense,
+because of how the CPU reports the forbidden operation to the kernel.
 .SH NOTES
 For a discussion of async-signal-safe functions, see
 .BR signal-safety (7).
Signature · 3 lines
Michael Kerrisk
Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
Linux/UNIX System Programming Training: http://man7.org/training/