fs/proc: report eip/esp in /prod/PID/stat for coredumping

Commit 0a1eb2d474ed ("fs/proc: Stop reporting eip and esp in /proc/PID/stat") stopped reporting eip/esp because it is racey and dangerous for executing tasks. The comment adds:

As far as I know, there are no use programs that make any material use of these fields, so just get rid of them.

However, existing userspace core-dump-handler applications (for example, minicoredumper) are using these fields since they provide an excellent cross-platform interface to these valuable pointers. So that commit introduced a user space visible regression.

Partially revert the change and make the readout possible for tasks with the proper permissions and only if the target task has the PF_DUMPCORE flag set.

---
 fs/proc/array.c |    8 ++++++++
 1 file changed, 8 insertions(+)

index 88c3555..696cc68 100644
@@ -421,7 +421,15 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 		 * esp and eip are intentionally zeroed out.  There is no
 		 * non-racy way to read them without freezing the task.
 		 * Programs that need reliable values can use ptrace(2).
+		 *
+		 * The only exception is if the task is core dumping because
+		 * a program is not able to use ptrace(2) in that case. It is
+		 * safe because the task has stopped executing permanently.
 		 */
+		if (permitted && (task->flags & PF_DUMPCORE)) {
+			eip = KSTK_EIP(task);
+			esp = KSTK_ESP(task);
+		}
 	}
 
 	get_task_comm(tcomm, task);

1.7.10.4

Looks okay to me.

--Andy

Duh, I completely forgot about mini core dumper when I acked that commit.

Ingo, I'm assuming I'll get this from the -tip tree, which is where I got the original commit that this fixes.

Thanks,
Linus

...

Yeah, will do!

Thanks,

Ingo

Commit-ID:  451eb3f2053ea4eeb40f94947c542cfbd7636186
Gitweb:     http://git.kernel.org/tip/451eb3f2053ea4eeb40f94947c542cfbd7636186
Author:     John Ogness <[email protected]>
AuthorDate: Thu, 14 Sep 2017 11:42:17 +0200
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Fri, 15 Sep 2017 23:04:59 +0200

fs/proc: Report eip/esp in /prod/PID/stat for coredumping

Commit 0a1eb2d474ed ("fs/proc: Stop reporting eip and esp in /proc/PID/stat") stopped reporting eip/esp because it is racey and dangerous for executing tasks. The comment adds:

As far as I know, there are no use programs that make any material use of these fields, so just get rid of them.

However, existing userspace core-dump-handler applications (for example, minicoredumper) are using these fields since they provide an excellent cross-platform interface to these valuable pointers. So that commit introduced a user space visible regression.

Partially revert the change and make the readout possible for tasks with the proper permissions and only if the target task has the PF_DUMPCORE flag set.

---
 fs/proc/array.c | 8 ++++++++
 1 file changed, 8 insertions(+)

index 88c3555..696cc68 100644
@@ -421,7 +421,15 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 		 * esp and eip are intentionally zeroed out.  There is no
 		 * non-racy way to read them without freezing the task.
 		 * Programs that need reliable values can use ptrace(2).
+		 *
+		 * The only exception is if the task is core dumping because
+		 * a program is not able to use ptrace(2) in that case. It is
+		 * safe because the task has stopped executing permanently.
 		 */
+		if (permitted && (task->flags & PF_DUMPCORE)) {
+			eip = KSTK_EIP(task);
+			esp = KSTK_ESP(task);
+		}
 	}
 
 	get_task_comm(tcomm, task);

Commit-ID:  fd7d56270b526ca3ed0c224362e3c64a0f86687a
Gitweb:     http://git.kernel.org/tip/fd7d56270b526ca3ed0c224362e3c64a0f86687a
Author:     John Ogness <[email protected]>
AuthorDate: Thu, 14 Sep 2017 11:42:17 +0200
Committer:  Thomas Gleixner <[email protected]>
CommitDate: Fri, 15 Sep 2017 23:31:16 +0200

fs/proc: Report eip/esp in /prod/PID/stat for coredumping

Commit 0a1eb2d474ed ("fs/proc: Stop reporting eip and esp in /proc/PID/stat") stopped reporting eip/esp because it is racy and dangerous for executing tasks. The comment adds:

As far as I know, there are no use programs that make any material use of these fields, so just get rid of them.

However, existing userspace core-dump-handler applications (for example, minicoredumper) are using these fields since they provide an excellent cross-platform interface to these valuable pointers. So that commit introduced a user space visible regression.

Partially revert the change and make the readout possible for tasks with the proper permissions and only if the target task has the PF_DUMPCORE flag set.

---
 fs/proc/array.c | 9 +++++++++
 1 file changed, 9 insertions(+)

index 88c3555..525157c 100644
@@ -62,6 +62,7 @@
 #include <linux/mman.h>
 #include <linux/sched/mm.h>
 #include <linux/sched/numa_balancing.h>
+#include <linux/sched/task_stack.h>
 #include <linux/sched/task.h>
 #include <linux/sched/cputime.h>
 #include <linux/proc_fs.h>

@@ -421,7 +422,15 @@ static int do_task_stat(struct seq_file *m, struct pid_namespace *ns,
 		 * esp and eip are intentionally zeroed out.  There is no
 		 * non-racy way to read them without freezing the task.
 		 * Programs that need reliable values can use ptrace(2).
+		 *
+		 * The only exception is if the task is core dumping because
+		 * a program is not able to use ptrace(2) in that case. It is
+		 * safe because the task has stopped executing permanently.
 		 */
+		if (permitted && (task->flags & PF_DUMPCORE)) {
+			eip = KSTK_EIP(task);
+			esp = KSTK_ESP(task);
+		}
 	}
 
 	get_task_comm(tcomm, task);