diff mbox series

padata: Disable BH when taking works lock on MT path

Message ID Zg0jEu5OsZaUFzn0@gondor.apana.org.au (mailing list archive)
State Accepted
Delegated to: Herbert Xu
Headers show
Series padata: Disable BH when taking works lock on MT path | expand

Commit Message

Herbert Xu April 3, 2024, 9:36 a.m. UTC
On Mon, Apr 01, 2024 at 07:08:28AM -0700, syzbot wrote:
> 
> syzbot found the following issue on:
> 
> HEAD commit:    18737353cca0 Merge tag 'edac_urgent_for_v6.9_rc2' of git:/..
> git tree:       upstream
> console output: https://syzkaller.appspot.com/x/log.txt?x=15d605e5180000
> kernel config:  https://syzkaller.appspot.com/x/.config?x=f64ec427e98bccd7
> dashboard link: https://syzkaller.appspot.com/bug?extid=0cb5bb0f4bf9e79db3b3
> compiler:       gcc (Debian 12.2.0-14) 12.2.0, GNU ld (GNU Binutils for Debian) 2.40

Strictly speaking this can't happen because for the time being
padata_do_multithreaded cannot run at the same time as the old
padata which occurs in BH context.

But the simplest fix is to just disable BH:

---8<---
As the old padata code can execute in softirq context, disable
softirqs for the new padata_do_mutithreaded code too as otherwise
lockdep will get antsy.

Reported-by: syzbot+0cb5bb0f4bf9e79db3b3@syzkaller.appspotmail.com
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Comments

Daniel Jordan April 3, 2024, 9:25 p.m. UTC | #1
On Wed, Apr 03, 2024 at 05:36:18PM +0800, Herbert Xu wrote:
> On Mon, Apr 01, 2024 at 07:08:28AM -0700, syzbot wrote:
> > 
> > syzbot found the following issue on:
> > 
> > HEAD commit:    18737353cca0 Merge tag 'edac_urgent_for_v6.9_rc2' of git:/..
> > git tree:       upstream
> > console output: https://syzkaller.appspot.com/x/log.txt?x=15d605e5180000
> > kernel config:  https://syzkaller.appspot.com/x/.config?x=f64ec427e98bccd7
> > dashboard link: https://syzkaller.appspot.com/bug?extid=0cb5bb0f4bf9e79db3b3
> > compiler:       gcc (Debian 12.2.0-14) 12.2.0, GNU ld (GNU Binutils for Debian) 2.40
> 
> Strictly speaking this can't happen because for the time being
> padata_do_multithreaded cannot run at the same time as the old
> padata which occurs in BH context.
> 
> But the simplest fix is to just disable BH:
> 
> ---8<---
> As the old padata code can execute in softirq context, disable
> softirqs for the new padata_do_mutithreaded code too as otherwise
> lockdep will get antsy.
> 
> Reported-by: syzbot+0cb5bb0f4bf9e79db3b3@syzkaller.appspotmail.com
> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Acked-by: Daniel Jordan <daniel.m.jordan@oracle.com>
diff mbox series

Patch

diff --git a/kernel/padata.c b/kernel/padata.c
index e3f639ff1670..53f4bc912712 100644
--- a/kernel/padata.c
+++ b/kernel/padata.c
@@ -106,7 +106,7 @@  static int __init padata_work_alloc_mt(int nworks, void *data,
 {
 	int i;
 
-	spin_lock(&padata_works_lock);
+	spin_lock_bh(&padata_works_lock);
 	/* Start at 1 because the current task participates in the job. */
 	for (i = 1; i < nworks; ++i) {
 		struct padata_work *pw = padata_work_alloc();
@@ -116,7 +116,7 @@  static int __init padata_work_alloc_mt(int nworks, void *data,
 		padata_work_init(pw, padata_mt_helper, data, 0);
 		list_add(&pw->pw_list, head);
 	}
-	spin_unlock(&padata_works_lock);
+	spin_unlock_bh(&padata_works_lock);
 
 	return i;
 }
@@ -134,12 +134,12 @@  static void __init padata_works_free(struct list_head *works)
 	if (list_empty(works))
 		return;
 
-	spin_lock(&padata_works_lock);
+	spin_lock_bh(&padata_works_lock);
 	list_for_each_entry_safe(cur, next, works, pw_list) {
 		list_del(&cur->pw_list);
 		padata_work_free(cur);
 	}
-	spin_unlock(&padata_works_lock);
+	spin_unlock_bh(&padata_works_lock);
 }
 
 static void padata_parallel_worker(struct work_struct *parallel_work)