Messages in this thread | | | Date | Thu, 3 Aug 2023 09:48:14 +0200 | From | Michal Hocko <> | Subject | Re: [PATCH v2] watchdog/hardlockup: Avoid large stack frames in watchdog_hardlockup_check() |
| |
On Wed 02-08-23 12:26:00, Douglas Anderson wrote: > After commit 77c12fc95980 ("watchdog/hardlockup: add a "cpu" param to > watchdog_hardlockup_check()") we started storing a `struct cpumask` on > the stack in watchdog_hardlockup_check(). On systems with > CONFIG_NR_CPUS set to 8192 this takes up 1K on the stack. That > triggers warnings with `CONFIG_FRAME_WARN` set to 1024. > > Instead of putting this `struct cpumask` on the stack, we'll allocate > it on the heap whenever userspace tells us that they want to backtrace > all CPUs upon a hardlockup. > > NOTE: the reason that this mask is even needed is to make sure that we > can print the hung CPU first, which makes the logs much easier to > understand. > > Fixes: 77c12fc95980 ("watchdog/hardlockup: add a "cpu" param to watchdog_hardlockup_check()") > Reported-by: kernel test robot <lkp@intel.com> > Closes: https://lore.kernel.org/r/202307310955.pLZDhpnl-lkp@intel.com > Signed-off-by: Douglas Anderson <dianders@chromium.org> > --- > > Changes in v2: > - Allocate space when userspace requests all cpus be backtraced. > > kernel/watchdog.c | 44 ++++++++++++++++++++++++++++++++++---------- > 1 file changed, 34 insertions(+), 10 deletions(-) > > diff --git a/kernel/watchdog.c b/kernel/watchdog.c > index be38276a365f..25d5627a6580 100644 > --- a/kernel/watchdog.c > +++ b/kernel/watchdog.c > @@ -93,6 +93,8 @@ static DEFINE_PER_CPU(bool, watchdog_hardlockup_warned); > static DEFINE_PER_CPU(bool, watchdog_hardlockup_touched); > static unsigned long watchdog_hardlockup_all_cpu_dumped; > > +static struct cpumask *hardlockup_backtrace_mask; > + > notrace void arch_touch_nmi_watchdog(void) > { > /* > @@ -106,6 +108,29 @@ notrace void arch_touch_nmi_watchdog(void) > } > EXPORT_SYMBOL(arch_touch_nmi_watchdog); > > +static int hardlockup_all_cpu_backtrace_proc_handler(struct ctl_table *table, int write, > + void *buffer, size_t *lenp, loff_t *ppos) > +{ > + int ret; > + > + ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); > + > + /* > + * Only allocate memory for the backtrace mask if userspace actually > + * wants to trace all CPUs since this can take up 1K of space on a > + * system with CONFIG_NR_CPUS=8192. > + */ > + if (sysctl_hardlockup_all_cpu_backtrace && !hardlockup_backtrace_mask) { > + hardlockup_backtrace_mask = > + kzalloc(sizeof(*hardlockup_backtrace_mask), GFP_KERNEL); > + } else if (!sysctl_hardlockup_all_cpu_backtrace && hardlockup_backtrace_mask) { > + kfree(hardlockup_backtrace_mask); > + hardlockup_backtrace_mask = NULL; > + }
While unlikely, this can race with the consumer and cause either use-after-free or NULL ptr deref. -- Michal Hocko SUSE Labs
| |