Hello Mimi,
On Tue, Nov 19, 2024 at 01:10:10PM -0500, Mimi Zohar wrote:
> Hi Breno,
>
> On Mon, 2024-11-04 at 02:47 -0800, Breno Leitao wrote:
> > Fix a potential RCU issue where ima_measurements list is traversed using
> > list_for_each_entry_rcu() without proper RCU read lock protection. This
> > caused warnings when CONFIG_PROVE_RCU was enabled:
> >
> > security/integrity/ima/ima_kexec.c:40 RCU-list traversed in non-reader
> > section!!
> >
> > Add rcu_read_lock() before iterating over ima_measurements list to ensure
> > proper RCU synchronization, consistent with other RCU list traversals in
> > the codebase.
>
> The synchronization is to prevent freeing of data while walking the RCU list.
> In
> this case, new measurements are only appended to the IMA measurement list. So
> there shouldn't be an issue.
>
> The IMA measurement list is being copied during kexec "load", while other
> processes are still running. Depending on the IMA policy, the kexec "load",
> itself, and these other processes may result in additional measurements, which
> should be copied across kexec. Adding the rcu_read_{lock, unlock} would
> unnecessarily prevent them from being copied.
Thank you for the detailed explanation. Since rcu_read_lock() operations are
lightweight, I believe keeping them wouldn't impact performance significantly.
However, if you prefer the lockless approach, I would suggest adding an
argument to list_for_each_entry_rcu() to keep the warning out. What are
your thoughts on this?
Author: Breno Leitao <[email protected]>
Date: Mon Nov 4 02:26:45 2024 -0800
ima: kexec: silence RCU list traversal warning
The ima_measurements list is append-only and doesn't require rcu_read_lock()
protection. However, lockdep issues a warning when traversing RCU lists
without the read lock:
security/integrity/ima/ima_kexec.c:40 RCU-list traversed in non-reader
section!!
Fix this by using the lockless variant of list_for_each_entry_rcu() with
the last argument set to true. This tells the RCU subsystem that
traversing this append-only list without the read lock is intentional
and safe.
This change silences the lockdep warning while maintaining the correct
semantics for the append-only list traversal.
Signed-off-by: Breno Leitao <[email protected]>
diff --git a/security/integrity/ima/ima_kexec.c
b/security/integrity/ima/ima_kexec.c
index 52e00332defed..9d45f4d26f731 100644
--- a/security/integrity/ima/ima_kexec.c
+++ b/security/integrity/ima/ima_kexec.c
@@ -37,7 +37,8 @@ static int ima_dump_measurement_list(unsigned long
*buffer_size, void **buffer,
memset(&khdr, 0, sizeof(khdr));
khdr.version = 1;
- list_for_each_entry_rcu(qe, &ima_measurements, later) {
+ /* This is an append-only list, no need to hold the RCU read lock */
+ list_for_each_entry_rcu(qe, &ima_measurements, later, true) {
if (file.count < file.size) {
khdr.count++;
ima_measurements_show(&file, qe);
