[1/2] gdb: push target earlier in procfs_target::attach (PR 27435)

Message ID 20210221155630.3026942-1-simon.marchi@polymtl.ca
State New
Headers show
Series
  • [1/2] gdb: push target earlier in procfs_target::attach (PR 27435)
Related show

Commit Message

Kevin Buettner via Gdb-patches Feb. 21, 2021, 3:56 p.m.
Since this is a GDB 9 -> 10 regression, I would like to push it to
gdb-10-branch.

This is a follow-up to:

  https://sourceware.org/pipermail/gdb-patches/2021-February/176202.html

This patch fixes a segfault seen when attaching to a process on Solaris.
The steps leading to the segfault are:

 - procfs_target::attach calls do_attach, at this point the inferior's
   process slot in the target stack is empty.
 - do_attach adds a thread with `add_thread (&the_procfs_target, ptid)`
 - in add_thread_silent, the passed target (&the_procfs_target) is
   passed to find_inferior_ptid
 - find_inferior_ptid returns nullptr, as there is no inferior with this
   ptid that has &the_procfs_target as its process target
 - the nullptr `inf` is passed to find_thread_ptid, which dereferences
   it, causing a segfault
 - back in procfs_target::attach, after do_attach, we push the
   the_procfs_target on the inferior's target stack, although we never
   reach this because the segfault happens before.

To fix this, I think we need to do the same as is done in
inf_ptrace_target::attach: push the target early and unpush it in case
the attach fails (and keep it if the attach succeeds).

Implement it by moving target_unpush_up to target.h, so it can be
re-used here.  Make procfs_target::attach use it.  Note that just like
is mentioned in inf_ptrace_target::attach, we should push the target
before calling target_pid_to_str, so that calling target_pid_to_str ends
up in procfs_target::pid_to_str.

Tested by trying to attach on a process on gcc211 on the gcc compile
farm.

gdb/ChangeLog:

	PR gdb/27435
	* inf-ptrace.c (struct target_unpusher): Move to target.h.
	(target_unpush_up): Likewise.
	* procfs.c (procfs_target::attach): Push target early.  Use
	target_unpush_up to unpush target in case of error.
	* target.h (struct target_unpusher): Move here.
	(target_unpush_up): Likewise.

Change-Id: I88aff8b20204e1ca1d792e27ac6bc34fc1aa0d52
---
 gdb/inf-ptrace.c | 16 ----------------
 gdb/procfs.c     | 14 ++++++++++++--
 gdb/target.h     | 14 ++++++++++++++
 3 files changed, 26 insertions(+), 18 deletions(-)

-- 
2.30.1

Comments

Tom Tromey Feb. 21, 2021, 9:04 p.m. | #1
>>>>> "Simon" == Simon Marchi via Gdb-patches <gdb-patches@sourceware.org> writes:


Simon> Tested by trying to attach on a process on gcc211 on the gcc compile
Simon> farm.

Simon> gdb/ChangeLog:

Simon> 	PR gdb/27435
Simon> 	* inf-ptrace.c (struct target_unpusher): Move to target.h.
Simon> 	(target_unpush_up): Likewise.
Simon> 	* procfs.c (procfs_target::attach): Push target early.  Use
Simon> 	target_unpush_up to unpush target in case of error.
Simon> 	* target.h (struct target_unpusher): Move here.
Simon> 	(target_unpush_up): Likewise.

FWIW this looks reasonable to me.

Tom
Kevin Buettner via Gdb-patches Feb. 21, 2021, 9:18 p.m. | #2
On 2021-02-21 4:04 p.m., Tom Tromey wrote:
>>>>>> "Simon" == Simon Marchi via Gdb-patches <gdb-patches@sourceware.org> writes:

> 

> Simon> Tested by trying to attach on a process on gcc211 on the gcc compile

> Simon> farm.

> 

> Simon> gdb/ChangeLog:

> 

> Simon> 	PR gdb/27435

> Simon> 	* inf-ptrace.c (struct target_unpusher): Move to target.h.

> Simon> 	(target_unpush_up): Likewise.

> Simon> 	* procfs.c (procfs_target::attach): Push target early.  Use

> Simon> 	target_unpush_up to unpush target in case of error.

> Simon> 	* target.h (struct target_unpusher): Move here.

> Simon> 	(target_unpush_up): Likewise.

> 

> FWIW this looks reasonable to me.

> 

> Tom


Thanks, I'll wait for Libor's feedback before pushing.

Simon
Kevin Buettner via Gdb-patches Feb. 22, 2021, 11:03 a.m. | #3
Many thanks to Simon for the patch.

I confirm that it fixes the attach crash on Solaris x86/sparc.
The testsuite did not reveal any new regressions on Fedora Linux.
Regarding Solaris, the tests require much more time,
so, I will let you know when I have results.

AFAIK, the code changes look good.

Libor

On 2/21/21 10:04 PM, Tom Tromey wrote:
>>>>>> "Simon" == Simon Marchi via Gdb-patches <gdb-patches@sourceware.org> writes:

> Simon> Tested by trying to attach on a process on gcc211 on the gcc compile

> Simon> farm.

>

> Simon> gdb/ChangeLog:

>

> Simon> 	PR gdb/27435

> Simon> 	* inf-ptrace.c (struct target_unpusher): Move to target.h.

> Simon> 	(target_unpush_up): Likewise.

> Simon> 	* procfs.c (procfs_target::attach): Push target early.  Use

> Simon> 	target_unpush_up to unpush target in case of error.

> Simon> 	* target.h (struct target_unpusher): Move here.

> Simon> 	(target_unpush_up): Likewise.

>

> FWIW this looks reasonable to me.

>

> Tom
Kevin Buettner via Gdb-patches Feb. 22, 2021, 4:51 p.m. | #4
On 2021-02-22 6:03 a.m., Libor Bukata wrote:> Many thanks to Simon for the patch.
> 

> I confirm that it fixes the attach crash on Solaris x86/sparc.

> The testsuite did not reveal any new regressions on Fedora Linux.

> Regarding Solaris, the tests require much more time,

> so, I will let you know when I have results.

> 

> AFAIK, the code changes look good.


Thanks, I pushed the series to the master branch, and patch 1 to
gdb-10-branch.  Please let me know if you find unexpected regressions on
Solaris (although hopefully, you'll see some tests going from FAIL to
PASS).

Simon
Kevin Buettner via Gdb-patches Feb. 23, 2021, 6:35 a.m. | #5
Thank you for the integration. Unit tests did not reveal any new 
regressions on Solaris.

Libor

On 2/22/21 5:51 PM, Simon Marchi wrote:
> On 2021-02-22 6:03 a.m., Libor Bukata wrote:> Many thanks to Simon for the patch.

>> I confirm that it fixes the attach crash on Solaris x86/sparc.

>> The testsuite did not reveal any new regressions on Fedora Linux.

>> Regarding Solaris, the tests require much more time,

>> so, I will let you know when I have results.

>>

>> AFAIK, the code changes look good.

> Thanks, I pushed the series to the master branch, and patch 1 to

> gdb-10-branch.  Please let me know if you find unexpected regressions on

> Solaris (although hopefully, you'll see some tests going from FAIL to

> PASS).

>

> Simon

Patch

diff --git a/gdb/inf-ptrace.c b/gdb/inf-ptrace.c
index 0f2f56cc3660..7ca02dfd8764 100644
--- a/gdb/inf-ptrace.c
+++ b/gdb/inf-ptrace.c
@@ -49,22 +49,6 @@  gdb_ptrace (PTRACE_TYPE_ARG1 request, ptid_t ptid, PTRACE_TYPE_ARG3 addr,
 #endif
 }
 
-/* A unique_ptr helper to unpush a target.  */
-
-struct target_unpusher
-{
-  void operator() (struct target_ops *ops) const
-  {
-    unpush_target (ops);
-  }
-};
-
-/* A unique_ptr that unpushes a target on destruction.  */
-
-typedef std::unique_ptr<struct target_ops, target_unpusher> target_unpush_up;
-
-
-
 inf_ptrace_target::~inf_ptrace_target ()
 {}
 
diff --git a/gdb/procfs.c b/gdb/procfs.c
index e73faa8d41dc..cab29c3cbbcb 100644
--- a/gdb/procfs.c
+++ b/gdb/procfs.c
@@ -1767,6 +1767,14 @@  procfs_target::attach (const char *args, int from_tty)
   if (pid == getpid ())
     error (_("Attaching GDB to itself is not a good idea..."));
 
+  /* Push the target if needed, ensure it gets un-pushed it if attach fails.  */
+  target_unpush_up unpusher;
+  if (!target_is_pushed (this))
+    {
+      push_target (this);
+      unpusher.reset (this);
+    }
+
   if (from_tty)
     {
       const char *exec_file = get_exec_file (0);
@@ -1780,9 +1788,11 @@  procfs_target::attach (const char *args, int from_tty)
 
       fflush (stdout);
     }
+
   do_attach (ptid_t (pid));
-  if (!target_is_pushed (this))
-    push_target (this);
+
+  /* Everything went fine, keep the target pushed.  */
+  unpusher.release ();
 }
 
 void
diff --git a/gdb/target.h b/gdb/target.h
index 0de78075e9b5..52e23b05ffd8 100644
--- a/gdb/target.h
+++ b/gdb/target.h
@@ -2392,6 +2392,20 @@  extern void push_target (target_ops_up &&);
 
 extern int unpush_target (struct target_ops *);
 
+/* A unique_ptr helper to unpush a target.  */
+
+struct target_unpusher
+{
+  void operator() (struct target_ops *ops) const
+  {
+    unpush_target (ops);
+  }
+};
+
+/* A unique_ptr that unpushes a target on destruction.  */
+
+typedef std::unique_ptr<struct target_ops, target_unpusher> target_unpush_up;
+
 extern void target_pre_inferior (int);
 
 extern void target_preopen (int);