On 03/02/23 02:31, Eric DeVolder wrote:
On 2/2/23 09:37, Eric DeVolder wrote:
On 2/1/23 00:38, Sourabh Jain wrote:
On architectures like PowerPC the crash notes are available for all
possible CPUs. So let's populate the elfcorehdr for all possible
CPUs having crash notes to avoid updating elfcorehdr during in-kernel
crash update on CPU hotplug events.
The similar technique is used in kexec-tool for kexec_load case.
Signed-off-by: Sourabh Jain <sourabhjain@xxxxxxxxxxxxx>
---
kernel/crash_core.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/kernel/crash_core.c b/kernel/crash_core.c
index 37c594858fd51..898d8d2fe2e2e 100644
--- a/kernel/crash_core.c
+++ b/kernel/crash_core.c
@@ -364,8 +364,8 @@ int crash_prepare_elf64_headers(struct kimage
*image, struct crash_mem *mem,
ehdr->e_ehsize = sizeof(Elf64_Ehdr);
ehdr->e_phentsize = sizeof(Elf64_Phdr);
- /* Prepare one phdr of type PT_NOTE for each present CPU */
- for_each_present_cpu(cpu) {
+ /* Prepare one phdr of type PT_NOTE for possible CPU with crash
note. */
+ for_each_possible_cpu(cpu) {
Sourabh,
Thomas Gleixner is suggesting moving away from for_each_present_cpu()
to for_each_online_cpu(). Using for_each_online_cpu() is going to the
minimum number of needed, whereas your approach of
for_each_possible_cpu() would be to the maximum number needed.
What would be the ramifications to ppc for moving towards
for_each_online_cpu()?
In my next patch series, I have finally figured out how to use cpuhp
framework to where it is possible to use for_each_online_cpu() here,
but that is at odds with your changes here.
Thanks,
eric
Without knowing the ramifications of changing to
for_each_online_cpu(), I currently am
using the following:
diff --git a/kernel/crash_core.c b/kernel/crash_core.c
index e1a3430f06f4..a019b691d974 100644
--- a/kernel/crash_core.c
+++ b/kernel/crash_core.c
@@ -366,6 +366,9 @@ int crash_prepare_elf64_headers(struct crash_mem
*mem, int n
/* Prepare one phdr of type PT_NOTE for each present CPU */
for_each_present_cpu(cpu) {
+ if (IS_ENABLED(CONFIG_CRASH_HOTPLUG)) {
+ if (!cpu_online(cpu)) continue;
+ }
How about let the arch decide the list of CPUs they want to pack in?
Because on
PowerPC the crash notes are created for possible CPUs and we can utilize
this
to avoid re-generating elfcorehdr for every hotplug operation.
phdr->p_type = PT_NOTE;
notes_addr = per_cpu_ptr_to_phys(per_cpu_ptr(crash_notes, cpu));
phdr->p_offset = phdr->p_paddr = notes_addr;
Thomas points out that the above can be simply the
for_each_online_cpu(), but again
I'm not sure how that impacts ppc,
which appears to layout all possible cpus rather
than just online ones. How are present but not online cpus handled by
crash analysis
utility?
As per my testing all worked fine if we replace for_each_present_cpu
with for_each_online_cpu
but again I don't know the reason why it worked. I will investigate it
and let you know.
How packing PT_LOAD for present CPU is impacting x86? Because on PowerPC
when system is on crash path it only populates the crash notes for
online CPUs, and skip
all other CPUs.
- Sourabh Jain
_______________________________________________
kexec mailing list
kexec@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/kexec
