On 15. 03. 2023. 21:59, Eric Dumazet wrote:
On Wed, Mar 15, 2023 at 1:57 PM Kuniyuki Iwashima kuniyu@amazon.com wrote:
However, we don't assume the delay and also the failure in tun_set_real_num_queues().
In this case, we have to re-initialise the queues without touching kobjects.
Eric, Are you working on this? If not, let me try fixing this :)
I am not working on this, please go ahead, thanks !
Hi,
It's me again. I just have new findings.
[root@pc-mtodorov linux_torvalds]# grep -E '(KOBJECT|TRACKER)' /boot/config-6.3.0-rc2-00006-gfc89d7fb499b CONFIG_REF_TRACKER=y CONFIG_NET_DEV_REFCNT_TRACKER=y CONFIG_NET_NS_REFCNT_TRACKER=y CONFIG_DEBUG_KOBJECT=y # CONFIG_DEBUG_KOBJECT_RELEASE is not set # CONFIG_SAMPLE_KOBJECT is not set # CONFIG_TEST_REF_TRACKER is not set [root@pc-mtodorov linux_torvalds]# uname -rms Linux 6.3.0-rc2-00006-gfc89d7fb499b x86_64 [root@pc-mtodorov linux_torvalds]# grep -E '(KOBJECT|TRACKER)' /boot/config-6.3.0-rc2-00006-gfc89d7fb499b CONFIG_REF_TRACKER=y CONFIG_NET_DEV_REFCNT_TRACKER=y CONFIG_NET_NS_REFCNT_TRACKER=y CONFIG_DEBUG_KOBJECT=y # CONFIG_DEBUG_KOBJECT_RELEASE is not set # CONFIG_SAMPLE_KOBJECT is not set # CONFIG_TEST_REF_TRACKER is not set [root@pc-mtodorov linux_torvalds]# tools/testing/selftests/net/tun TAP version 13 1..5 # Starting 5 tests from 1 test cases. # RUN tun.delete_detach_close ... # OK tun.delete_detach_close ok 1 tun.delete_detach_close # RUN tun.detach_delete_close ... # OK tun.detach_delete_close ok 2 tun.detach_delete_close # RUN tun.detach_close_delete ... # OK tun.detach_close_delete ok 3 tun.detach_close_delete # RUN tun.reattach_delete_close ... # OK tun.reattach_delete_close ok 4 tun.reattach_delete_close # RUN tun.reattach_close_delete ... # OK tun.reattach_close_delete ok 5 tun.reattach_close_delete # PASSED: 5 / 5 tests passed. # Totals: pass:5 fail:0 xfail:0 xpass:0 skip:0 error:0 [root@pc-mtodorov linux_torvalds]#
My interpretation if you allow it is that the bug search can be narrowed to the code that depends on CONFIG_DEBUG_KOBJECT_RELEASE=y.
Best regards, Mirsad
CONFIG_DEBUG_KOBJECT=y alone doesn't seem to be sufficient to trigger the reference leak.
Hope this helps narrow down the search.