On 8/5/2025 2:28 AM, Catalin Marinas wrote:
On Mon, Aug 04, 2025 at 05:34:10AM -0700, Breno Leitao wrote:
On Mon, Aug 04, 2025 at 01:08:15PM +0100, Catalin Marinas wrote:
weird enought, lockdep never picked this issue, and I have a few set of hosts running kmemleak and lockdep for a while.
This time was different because I have decided to invstiage the code, and found the deadlock. Still, no lockdep complain at all.
I guess it's because kmemleak is quiet in general, unless problems are found, and lockdep never registered this combination - printk() called with the kmemleak_lock held.
As you have guessed, this issue occurred right after the kmemleak warning.
Thanks for the reminder about printk_deferred_* and the suggestion regarding code comments..
I will send a new version after verification.
Best regards, Guber