[PATCH] kexec/crash: no crash update when kexec in progress

[Baoquan He]

On 08/19/24 at 09:45am, Sourabh Jain wrote:
> Hello Michael and Boaquan
> 
> On 01/08/24 12:21, Sourabh Jain wrote:
> > Hello Michael,
> > 
> > On 01/08/24 08:04, Michael Ellerman wrote:
> > > Sourabh Jain <sourabhjain at linux.ibm.com> writes:
> > > > The following errors are observed when kexec is done with SMT=off on
> > > > powerpc.
> > > > 
> > > > [  358.458385] Removing IBM Power 842 compression device
> > > > [  374.795734] kexec_core: Starting new kernel
> > > > [  374.795748] kexec: Waking offline cpu 1.
> > > > [  374.875695] crash hp: kexec_trylock() failed, elfcorehdr may
> > > > be inaccurate
> > > > [  374.935833] kexec: Waking offline cpu 2.
> > > > [  375.015664] crash hp: kexec_trylock() failed, elfcorehdr may
> > > > be inaccurate
> > > > snip..
> > > > [  375.515823] kexec: Waking offline cpu 6.
> > > > [  375.635667] crash hp: kexec_trylock() failed, elfcorehdr may
> > > > be inaccurate
> > > > [  375.695836] kexec: Waking offline cpu 7.
> > > Are they actually errors though? Do they block the actual kexec from
> > > happening? Or are they just warnings in dmesg?
> > 
> > The kexec kernel boots fine.
> > 
> > This warning appears regardless of whether the kdump kernel is loaded.
> > 
> > However, when the kdump kernel is loaded, we will not be able to update
> > the kdump image (FDT).
> > I think this should be fine given that kexec is in progress.
> > 
> > Please let me know your opinion.
> > 
> > > Because the fix looks like it could be racy.
> > 
> > It seems like it is racy, but given that kexec takes the lock first and
> > then
> > brings the CPU up, which triggers the kdump image, which always fails to
> > update the kdump image because it could not take the same lock.
> > 
> > Note: the kexec lock is not released unless kexec boot fails.
> 
> Any comments or suggestions on this fix?

Is this a little better?

diff --git a/kernel/crash_core.c b/kernel/crash_core.c
index 63cf89393c6e..0355ffb712f4 100644
--- a/kernel/crash_core.c
+++ b/kernel/crash_core.c
@@ -504,7 +504,7 @@ int crash_check_hotplug_support(void)
 
 	crash_hotplug_lock();
 	/* Obtain lock while reading crash information */
-	if (!kexec_trylock()) {
+	if (!kexec_trylock() && kexec_in_progress) {
 		pr_info("kexec_trylock() failed, elfcorehdr may be inaccurate\n");
 		crash_hotplug_unlock();
 		return 0;
@@ -539,7 +539,7 @@ static void crash_handle_hotplug_event(unsigned int hp_action, unsigned int cpu,
 
 	crash_hotplug_lock();
 	/* Obtain lock while changing crash information */
-	if (!kexec_trylock()) {
+	if (!kexec_trylock() && kexec_in_progress) {
 		pr_info("kexec_trylock() failed, elfcorehdr may be inaccurate\n");
 		crash_hotplug_unlock();
 		return;