linux-kernel - Re: [interval_tree_test_init] BUG: soft lockup

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 7 Nov 2017 08:58:14 -0800
From:   Davidlohr Bueso <dave@...olabs.net>
To:     Fengguang Wu <fengguang.wu@...el.com>
Cc:     Chris Wilson <chris@...is-wilson.co.uk>,
        linux-kernel@...r.kernel.org,
        Linus Torvalds <torvalds@...ux-foundation.org>
Subject: Re: [interval_tree_test_init] BUG: soft lockup - CPU#0 stuck for
 23s! [swapper/0:1]

On Tue, 07 Nov 2017, Fengguang Wu wrote:

>[  265.102312] xz_dec_test: module loaded
>[  265.111774] xz_dec_test: Create a device node with 'mknod xz_dec_test c 246 0' and write .xz files to it.
>[  265.160320] atomic64_test: passed for x86-64 platform with CX8 and with SSE
>[  265.181966] glob: 64 self-tests passed, 0 failed
>[  265.196996] interval tree insert/remove
>[  292.348094] watchdog: BUG: soft lockup - CPU#0 stuck for 23s! [swapper/0:1]
>[  292.348094] irq event stamp: 6190340
>[  292.396418] hardirqs last  enabled at (6190339): [<ffffffff81fb006b>] restore_regs_and_iret+0x0/0x1d
>[  292.396418] hardirqs last disabled at (6190340): [<ffffffff81fb0cb8>] apic_timer_interrupt+0x98/0xb0
>[  292.396418] softirqs last  enabled at (6182242): [<ffffffff81fb2942>] __do_softirq+0x382/0x3f7
>[  292.396418] softirqs last disabled at (6182235): [<ffffffff810b6696>] irq_exit+0x51/0x7b
>[  292.396418] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.14.0-rc8 #40
>[  292.396418] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
>[  292.396418] task: ffff880072bf4040 task.stack: ffffc9000031c000
>[  292.396418] RIP: 0010:interval_tree_insert+0x4c/0xd4

That would be:

     for (i = 0; i < perf_loops; i++) {
	 for (j = 0; j < nnodes; j++)
		interval_tree_insert(nodes + j, &root);
	 for (j = 0; j < nnodes; j++)
		interval_tree_remove(nodes + j, &root);
     }

Which if you are using the default params would end up doing insert() 10 million times.
How about decrementing perf_loops, it's pretty high.

diff --git a/lib/interval_tree_test.c b/lib/interval_tree_test.c
index 0e343fd29570..1dbdf3704247 100644
--- a/lib/interval_tree_test.c
+++ b/lib/interval_tree_test.c
@@ -11,7 +11,7 @@
        MODULE_PARM_DESC(name, msg);
 
 __param(int, nnodes, 100, "Number of nodes in the interval tree");
-__param(int, perf_loops, 100000, "Number of iterations modifying the tree");
+__param(int, perf_loops, 1000, "Number of iterations modifying the tree");
 
 __param(int, nsearches, 100, "Number of searches to the interval tree");
 __param(int, search_loops, 10000, "Number of iterations searching the tree");

If that works I guess we'd have to limit the max values for such parameters
to avoid soft lockups. The same applies to your rbtree test report.

Thanks,
Davidlohr