Skip to contentSkip to conversation

Potential security problem in patch: Fix reading /proc/<pid>/mem when parent dies.

DISCUSSION Alan Cox 2 participants 3 messages Nov 19, 2004 1 message not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

Alan Cox

author
#1 lore raw
Quoted 6 lines · Linux Kernel Mailing List

ChangeSet 1.2155, 2004/11/18 08:01:00-08:00, [email protected]

Fix reading /proc/<pid>/mem when parent dies.

We should not touch "self_exec_id" here. The parent changed, not we.

The original point of this was that if our parent changed then our new parent is not aware of our special status. As a result we can send random signals to init and since it does not see SIGCLD we can get zombies or worse when we exit.

The original code was correct here for protecting init. The side effect does need fixing but not this way. Perhaps it would be simpler just to protect init as it is already "special" for signal handling.

Alan

#2 lore raw
Quoted 5 lines · Alan Cox

The original point of this was that if our parent changed then our new parent is not aware of our special status. As a result we can send random signals to init and since it does not see SIGCLD we can get zombies or worse when we exit.

How could we send random signals? That's what the "exit_signal" thing is for, and the code does

	if (p->exit_signal != -1)
		p->exit_signal = SIGCHLD;

for that.

Is there any other way to set exit_signal afterwards? If so, I think we should have a security check at _that_ point.

Linus

author
#3 lore raw
Quoted 10 lines · Linus Torvalds

How could we send random signals? That's what the "exit_signal" thing is for, and the code does

	if (p->exit_signal != -1)
		p->exit_signal = SIGCHLD;

for that.

Is there any other way to set exit_signal afterwards? If so, I think we should have a security check at _that_ point.

Ok that makes sense now I look harder at it. While it was added to protect agains that case the code you quote already covers all the cases I can see. We can clone new threads but they too will get reparented or will simply kill us.