| Message ID | 20250109033136.2845676-1-guoweikang.kernel@gmail.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | arch/s390: save_area_alloc default failure behavior changed to panic | expand |
On Thu, Jan 09, 2025 at 11:31:36AM +0800, Guo Weikang wrote: Hi Guo, > Now with the memblock_alloc_or_panic interface, save_area_alloc no longer > needs to handle panic itself. > > Signed-off-by: Guo Weikang <guoweikang.kernel@gmail.com> > --- > arch/s390/kernel/crash_dump.c | 4 +--- > arch/s390/kernel/numa.c | 3 +-- > arch/s390/kernel/smp.c | 4 ---- > 3 files changed, 2 insertions(+), 9 deletions(-) This patch is a follow-up to v7, but instead it needs to be part of v8. I guess Andrew would refresh mm-everything (or whatever he finds appropriate) with the new version. @Andrew, please correct me if I am wroing. Thanks!
Hi, Alexander > > Alexander Gordeev <agordeev@linux.ibm.com> wrote on Thursday, 9 January 2025 16:08 > > On Thu, Jan 09, 2025 at 11:31:36AM +0800, Guo Weikang wrote: > > Hi Guo, > > > Now with the memblock_alloc_or_panic interface, save_area_alloc no longer > > needs to handle panic itself. > > > > Signed-off-by: Guo Weikang <guoweikang.kernel@gmail.com> > > --- > > arch/s390/kernel/crash_dump.c | 4 +--- > > arch/s390/kernel/numa.c | 3 +-- > > arch/s390/kernel/smp.c | 4 ---- > > 3 files changed, 2 insertions(+), 9 deletions(-) > > This patch is a follow-up to v7, but instead it needs to be part of v8. > I guess Andrew would refresh mm-everything (or whatever he finds appropriate) > with the new version. Sorry to confuse you, `memblock_alloc_or_panic` is already merged into mm/mm-everything by Andrew, so this is an additional patch specifically to fix the problem you mentioned. This patch is based on the latest mm/mm-everything branch, and it has also been merged by Andrew，You should have received the email. @Alexander I hope I cleared your confusion. ^ ^ > > @Andrew, please correct me if I am wroing. > > Thanks! Best regards. --- Guo
On Thu, Jan 09, 2025 at 04:28:06PM +0800, Weikang Guo wrote: > > This patch is a follow-up to v7, but instead it needs to be part of v8. > > I guess Andrew would refresh mm-everything (or whatever he finds appropriate) > > with the new version. > > Sorry to confuse you, `memblock_alloc_or_panic` is already merged > into mm/mm-everything > by Andrew, so this is an additional patch specifically to fix the > problem you mentioned. > > This patch is based on the latest mm/mm-everything branch, and it has > also been merged > by Andrew，You should have received the email. > > @Alexander I hope I cleared your confusion. ^ ^ I do not think I am confused. The merged patch is not good enough for s390. This patch addresses my comments, but it needs to be merged into v7 or re- posted as v8. @Andrew, is that still possible? > > @Andrew, please correct me if I am wroing. > > > > Thanks! > > > Best regards. > --- > Guo Thanks!
Alexander Gordeev <agordeev@linux.ibm.com> wrote on Thursday, 9 January 2025 21:56: > > On Thu, Jan 09, 2025 at 04:28:06PM +0800, Weikang Guo wrote: > > > This patch is a follow-up to v7, but instead it needs to be part of v8. > > > I guess Andrew would refresh mm-everything (or whatever he finds appropriate) > > > with the new version. > > > > Sorry to confuse you, `memblock_alloc_or_panic` is already merged > > into mm/mm-everything > > by Andrew, so this is an additional patch specifically to fix the > > problem you mentioned. > > > > This patch is based on the latest mm/mm-everything branch, and it has > > also been merged > > by Andrew，You should have received the email. > > > > @Alexander I hope I cleared your confusion. ^ ^ > > I do not think I am confused. The merged patch is not good enough for s390. > This patch addresses my comments, but it needs to be merged into v7 or re- > posted as v8. > I understand your point, you want to modify directly based on the original patch. If we repost it,it may not be v8 but v9. Last v8 patch: https://lkml.kernel.org/r/20250102072528.650926-1-guoweikang.kernel@gmail.com I reduced the recipients for V8 because I remembered that the number of CCs might have exceeded the limit at the time, and there hadn't been any comments on V7 for a long time > @Andrew, is that still possible? @Andrew, I can cooperate if necessary, I could merge them into one patch and repost. > > > @Andrew, please correct me if I am wrong. > > > > > > Thanks! > > > > > > Best regards. > > --- > > Guo > > Thanks! --- Guo
diff --git a/arch/s390/kernel/crash_dump.c b/arch/s390/kernel/crash_dump.c index f699df2a2b11..276cb4c1e11b 100644 --- a/arch/s390/kernel/crash_dump.c +++ b/arch/s390/kernel/crash_dump.c @@ -63,9 +63,7 @@ struct save_area * __init save_area_alloc(bool is_boot_cpu) { struct save_area *sa; - sa = memblock_alloc(sizeof(*sa), 8); - if (!sa) - return NULL; + sa = memblock_alloc_or_panic(sizeof(*sa), 8); if (is_boot_cpu) list_add(&sa->list, &dump_save_areas); diff --git a/arch/s390/kernel/numa.c b/arch/s390/kernel/numa.c index a33e20f73330..2fc40f97c0ad 100644 --- a/arch/s390/kernel/numa.c +++ b/arch/s390/kernel/numa.c @@ -21,9 +21,8 @@ void __init numa_setup(void) nodes_clear(node_possible_map); node_set(0, node_possible_map); node_set_online(0); - for (nid = 0; nid < MAX_NUMNODES; nid++) { + for (nid = 0; nid < MAX_NUMNODES; nid++) NODE_DATA(nid) = memblock_alloc_or_panic(sizeof(pg_data_t), 8); - } NODE_DATA(0)->node_spanned_pages = memblock_end_of_DRAM() >> PAGE_SHIFT; NODE_DATA(0)->node_id = 0; } diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c index d77aaefb59bd..7b08399b0846 100644 --- a/arch/s390/kernel/smp.c +++ b/arch/s390/kernel/smp.c @@ -611,8 +611,6 @@ void __init smp_save_dump_ipl_cpu(void) if (!dump_available()) return; sa = save_area_alloc(true); - if (!sa) - panic("could not allocate memory for boot CPU save area\n"); regs = memblock_alloc_or_panic(512, 8); copy_oldmem_kernel(regs, __LC_FPREGS_SAVE_AREA, 512); save_area_add_regs(sa, regs); @@ -646,8 +644,6 @@ void __init smp_save_dump_secondary_cpus(void) SIGP_CC_NOT_OPERATIONAL) continue; sa = save_area_alloc(false); - if (!sa) - panic("could not allocate memory for save area\n"); __pcpu_sigp_relax(addr, SIGP_STORE_STATUS_AT_ADDRESS, __pa(page)); save_area_add_regs(sa, page); if (cpu_has_vx()) {
Now with the memblock_alloc_or_panic interface, save_area_alloc no longer needs to handle panic itself. Signed-off-by: Guo Weikang <guoweikang.kernel@gmail.com> --- arch/s390/kernel/crash_dump.c | 4 +--- arch/s390/kernel/numa.c | 3 +-- arch/s390/kernel/smp.c | 4 ---- 3 files changed, 2 insertions(+), 9 deletions(-)