| Message ID | 20240214095146.1527369-1-mahesh@linux.ibm.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | [v4] powerpc: Avoid nmi_enter/nmi_exit in real mode interrupt. | expand |
| Context | Check | Description |
|---|---|---|
| snowpatch_ozlabs/github-powerpc_ppctests | success | Successfully ran 8 jobs. |
| snowpatch_ozlabs/github-powerpc_selftests | success | Successfully ran 8 jobs. |
| snowpatch_ozlabs/github-powerpc_sparse | success | Successfully ran 4 jobs. |
| snowpatch_ozlabs/github-powerpc_clang | success | Successfully ran 6 jobs. |
| snowpatch_ozlabs/github-powerpc_kernel_qemu | success | Successfully ran 23 jobs. |
Le 14/02/2024 à 10:51, Mahesh Salgaonkar a écrit : > nmi_enter()/nmi_exit() touches per cpu variables which can lead to kernel > crash when invoked during real mode interrupt handling (e.g. early HMI/MCE > interrupt handler) if percpu allocation comes from vmalloc area. > > Early HMI/MCE handlers are called through DEFINE_INTERRUPT_HANDLER_NMI() > wrapper which invokes nmi_enter/nmi_exit calls. We don't see any issue when > percpu allocation is from the embedded first chunk. However with > CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there are chances where percpu > allocation can come from the vmalloc area. > > With kernel command line "percpu_alloc=page" we can force percpu allocation > to come from vmalloc area and can see kernel crash in machine_check_early: > > [ 1.215714] NIP [c000000000e49eb4] rcu_nmi_enter+0x24/0x110 > [ 1.215717] LR [c0000000000461a0] machine_check_early+0xf0/0x2c0 > [ 1.215719] --- interrupt: 200 > [ 1.215720] [c000000fffd73180] [0000000000000000] 0x0 (unreliable) > [ 1.215722] [c000000fffd731b0] [0000000000000000] 0x0 > [ 1.215724] [c000000fffd73210] [c000000000008364] machine_check_early_common+0x134/0x1f8 > > Fix this by avoiding use of nmi_enter()/nmi_exit() in real mode if percpu > first chunk is not embedded. > > Signed-off-by: Mahesh Salgaonkar <mahesh@linux.ibm.com> Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu> > --- > Changes in v4: > - Fix coding style issues. > > Changes in v3: > - Address comments from Christophe Leroy to avoid using #ifdefs in the > code > - v2 at https://lore.kernel.org/linuxppc-dev/20240205053647.1763446-1-mahesh@linux.ibm.com/ > > Changes in v2: > - Rebase to upstream master > - Use jump_labels, if CONFIG_JUMP_LABEL is enabled, to avoid redoing the > embed first chunk test at each interrupt entry. > - v1 is at https://lore.kernel.org/linuxppc-dev/164578465828.74956.6065296024817333750.stgit@jupiter/ > --- > arch/powerpc/include/asm/interrupt.h | 10 ++++++++++ > arch/powerpc/include/asm/percpu.h | 10 ++++++++++ > arch/powerpc/kernel/setup_64.c | 3 +++ > 3 files changed, 23 insertions(+) > > diff --git a/arch/powerpc/include/asm/interrupt.h b/arch/powerpc/include/asm/interrupt.h > index a4196ab1d0167..0b96464ff0339 100644 > --- a/arch/powerpc/include/asm/interrupt.h > +++ b/arch/powerpc/include/asm/interrupt.h > @@ -336,6 +336,14 @@ static inline void interrupt_nmi_enter_prepare(struct pt_regs *regs, struct inte > if (IS_ENABLED(CONFIG_KASAN)) > return; > > + /* > + * Likewise, do not use it in real mode if percpu first chunk is not > + * embedded. With CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there > + * are chances where percpu allocation can come from vmalloc area. > + */ > + if (IS_ENABLED(CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK) && !is_embed_first_chunk) > + return; > + > /* Otherwise, it should be safe to call it */ > nmi_enter(); > } > @@ -351,6 +359,8 @@ static inline void interrupt_nmi_exit_prepare(struct pt_regs *regs, struct inter > // no nmi_exit for a pseries hash guest taking a real mode exception > } else if (IS_ENABLED(CONFIG_KASAN)) { > // no nmi_exit for KASAN in real mode > + } else if (IS_ENABLED(CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK) && !is_embed_first_chunk) { > + // no nmi_exit if percpu first chunk is not embedded > } else { > nmi_exit(); > } > diff --git a/arch/powerpc/include/asm/percpu.h b/arch/powerpc/include/asm/percpu.h > index 8e5b7d0b851c6..e24063eb0b33b 100644 > --- a/arch/powerpc/include/asm/percpu.h > +++ b/arch/powerpc/include/asm/percpu.h > @@ -15,6 +15,16 @@ > #endif /* CONFIG_SMP */ > #endif /* __powerpc64__ */ > > +#ifdef CONFIG_PPC64 > +#include <linux/jump_label.h> > +DECLARE_STATIC_KEY_FALSE(__percpu_embed_first_chunk); > + > +#define is_embed_first_chunk \ > + (static_key_enabled(&__percpu_embed_first_chunk.key)) > +#else > +#define is_embed_first_chunk true > +#endif /* CONFIG_PPC64 */ > + > #include <asm-generic/percpu.h> > > #include <asm/paca.h> > diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c > index 2f19d5e944852..e04f0ff69d4b6 100644 > --- a/arch/powerpc/kernel/setup_64.c > +++ b/arch/powerpc/kernel/setup_64.c > @@ -834,6 +834,7 @@ static __init int pcpu_cpu_to_node(int cpu) > > unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; > EXPORT_SYMBOL(__per_cpu_offset); > +DEFINE_STATIC_KEY_FALSE(__percpu_embed_first_chunk); > > void __init setup_per_cpu_areas(void) > { > @@ -869,6 +870,8 @@ void __init setup_per_cpu_areas(void) > pr_warn("PERCPU: %s allocator failed (%d), " > "falling back to page size\n", > pcpu_fc_names[pcpu_chosen_fc], rc); > + else > + static_key_enable(&__percpu_embed_first_chunk.key); > } > > if (rc < 0)
On Wed, 2024-02-14 at 15:21 +0530, Mahesh Salgaonkar wrote: > nmi_enter()/nmi_exit() touches per cpu variables which can lead to > kernel > crash when invoked during real mode interrupt handling (e.g. early > HMI/MCE > interrupt handler) if percpu allocation comes from vmalloc area. > > Early HMI/MCE handlers are called through > DEFINE_INTERRUPT_HANDLER_NMI() > wrapper which invokes nmi_enter/nmi_exit calls. We don't see any > issue when > percpu allocation is from the embedded first chunk. However with > CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there are chances where > percpu > allocation can come from the vmalloc area. > > With kernel command line "percpu_alloc=page" we can force percpu > allocation > to come from vmalloc area and can see kernel crash in > machine_check_early: > > [ 1.215714] NIP [c000000000e49eb4] rcu_nmi_enter+0x24/0x110 > [ 1.215717] LR [c0000000000461a0] machine_check_early+0xf0/0x2c0 > [ 1.215719] --- interrupt: 200 > [ 1.215720] [c000000fffd73180] [0000000000000000] 0x0 (unreliable) > [ 1.215722] [c000000fffd731b0] [0000000000000000] 0x0 > [ 1.215724] [c000000fffd73210] [c000000000008364] > machine_check_early_common+0x134/0x1f8 > > Fix this by avoiding use of nmi_enter()/nmi_exit() in real mode if > percpu > first chunk is not embedded. > > Signed-off-by: Mahesh Salgaonkar <mahesh@linux.ibm.com> Thanks for the Patch. I have tested the patch and the fix works fine. selftests/powerpc/mce/inject-ra-err testcase is working as expected after enabling percpu_alloc=page with the patch applied. Output with Patch: # ./inject-ra-err test: inject-ra-err tags: git_version:unknown success: inject-ra-err # Tested-by: Shirisha Ganta <shirisha@linux.ibm.com> > --- > Changes in v4: > - Fix coding style issues. > > Changes in v3: > - Address comments from Christophe Leroy to avoid using #ifdefs in > the > code > - v2 at > https://lore.kernel.org/linuxppc-dev/20240205053647.1763446-1-mahesh@linux.ibm.com/ > > Changes in v2: > - Rebase to upstream master > - Use jump_labels, if CONFIG_JUMP_LABEL is enabled, to avoid redoing > the > embed first chunk test at each interrupt entry. > - v1 is at > https://lore.kernel.org/linuxppc-dev/164578465828.74956.6065296024817333750.stgit@jupiter/ > --- > arch/powerpc/include/asm/interrupt.h | 10 ++++++++++ > arch/powerpc/include/asm/percpu.h | 10 ++++++++++ > arch/powerpc/kernel/setup_64.c | 3 +++ > 3 files changed, 23 insertions(+) > > diff --git a/arch/powerpc/include/asm/interrupt.h > b/arch/powerpc/include/asm/interrupt.h > index a4196ab1d0167..0b96464ff0339 100644 > --- a/arch/powerpc/include/asm/interrupt.h > +++ b/arch/powerpc/include/asm/interrupt.h > @@ -336,6 +336,14 @@ static inline void > interrupt_nmi_enter_prepare(struct pt_regs *regs, struct inte > if (IS_ENABLED(CONFIG_KASAN)) > return; > > + /* > + * Likewise, do not use it in real mode if percpu first chunk > is not > + * embedded. With CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled > there > + * are chances where percpu allocation can come from vmalloc > area. > + */ > + if (IS_ENABLED(CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK) && > !is_embed_first_chunk) > + return; > + > /* Otherwise, it should be safe to call it */ > nmi_enter(); > } > @@ -351,6 +359,8 @@ static inline void > interrupt_nmi_exit_prepare(struct pt_regs *regs, struct inter > // no nmi_exit for a pseries hash guest taking a real > mode exception > } else if (IS_ENABLED(CONFIG_KASAN)) { > // no nmi_exit for KASAN in real mode > + } else if (IS_ENABLED(CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK) && > !is_embed_first_chunk) { > + // no nmi_exit if percpu first chunk is not embedded > } else { > nmi_exit(); > } > diff --git a/arch/powerpc/include/asm/percpu.h > b/arch/powerpc/include/asm/percpu.h > index 8e5b7d0b851c6..e24063eb0b33b 100644 > --- a/arch/powerpc/include/asm/percpu.h > +++ b/arch/powerpc/include/asm/percpu.h > @@ -15,6 +15,16 @@ > #endif /* CONFIG_SMP */ > #endif /* __powerpc64__ */ > > +#ifdef CONFIG_PPC64 > +#include <linux/jump_label.h> > +DECLARE_STATIC_KEY_FALSE(__percpu_embed_first_chunk); > + > +#define is_embed_first_chunk \ > + (static_key_enabled(&__percpu_embed_first_chunk.key)) > +#else > +#define is_embed_first_chunk true > +#endif /* CONFIG_PPC64 */ > + > #include <asm-generic/percpu.h> > > #include <asm/paca.h> > diff --git a/arch/powerpc/kernel/setup_64.c > b/arch/powerpc/kernel/setup_64.c > index 2f19d5e944852..e04f0ff69d4b6 100644 > --- a/arch/powerpc/kernel/setup_64.c > +++ b/arch/powerpc/kernel/setup_64.c > @@ -834,6 +834,7 @@ static __init int pcpu_cpu_to_node(int cpu) > > unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; > EXPORT_SYMBOL(__per_cpu_offset); > +DEFINE_STATIC_KEY_FALSE(__percpu_embed_first_chunk); > > void __init setup_per_cpu_areas(void) > { > @@ -869,6 +870,8 @@ void __init setup_per_cpu_areas(void) > pr_warn("PERCPU: %s allocator failed (%d), " > "falling back to page size\n", > pcpu_fc_names[pcpu_chosen_fc], rc); > + else > + static_key_enable(&__percpu_embed_first_chunk.k > ey); > } > > if (rc < 0)
Hi Mahesh, Mahesh Salgaonkar <mahesh@linux.ibm.com> writes: > nmi_enter()/nmi_exit() touches per cpu variables which can lead to kernel > crash when invoked during real mode interrupt handling (e.g. early HMI/MCE > interrupt handler) if percpu allocation comes from vmalloc area. > > Early HMI/MCE handlers are called through DEFINE_INTERRUPT_HANDLER_NMI() > wrapper which invokes nmi_enter/nmi_exit calls. We don't see any issue when > percpu allocation is from the embedded first chunk. However with > CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there are chances where percpu > allocation can come from the vmalloc area. > > With kernel command line "percpu_alloc=page" we can force percpu allocation > to come from vmalloc area and can see kernel crash in machine_check_early: > > [ 1.215714] NIP [c000000000e49eb4] rcu_nmi_enter+0x24/0x110 > [ 1.215717] LR [c0000000000461a0] machine_check_early+0xf0/0x2c0 > [ 1.215719] --- interrupt: 200 > [ 1.215720] [c000000fffd73180] [0000000000000000] 0x0 (unreliable) > [ 1.215722] [c000000fffd731b0] [0000000000000000] 0x0 > [ 1.215724] [c000000fffd73210] [c000000000008364] machine_check_early_common+0x134/0x1f8 > > Fix this by avoiding use of nmi_enter()/nmi_exit() in real mode if percpu > first chunk is not embedded. My system (powernv) doesn't even boot with percpu_alloc=page. AFAIK the only reason we added support for it was to handle 4K kernels with HPT. See commit eb553f16973a ("powerpc/64/mm: implement page mapping percpu first chunk allocator"). So I wonder if we should change the Kconfig to only offer it as an option in that case, and change the logic in setup_per_cpu_areas() to only use it as a last resort. I guess we probably still need this commit though, even if just for 4K HPT. > Signed-off-by: Mahesh Salgaonkar <mahesh@linux.ibm.com> > --- > Changes in v4: > - Fix coding style issues. > > Changes in v3: > - Address comments from Christophe Leroy to avoid using #ifdefs in the > code > - v2 at https://lore.kernel.org/linuxppc-dev/20240205053647.1763446-1-mahesh@linux.ibm.com/ > > Changes in v2: > - Rebase to upstream master > - Use jump_labels, if CONFIG_JUMP_LABEL is enabled, to avoid redoing the > embed first chunk test at each interrupt entry. > - v1 is at https://lore.kernel.org/linuxppc-dev/164578465828.74956.6065296024817333750.stgit@jupiter/ > --- > arch/powerpc/include/asm/interrupt.h | 10 ++++++++++ > arch/powerpc/include/asm/percpu.h | 10 ++++++++++ > arch/powerpc/kernel/setup_64.c | 3 +++ > 3 files changed, 23 insertions(+) > > diff --git a/arch/powerpc/include/asm/interrupt.h b/arch/powerpc/include/asm/interrupt.h > index a4196ab1d0167..0b96464ff0339 100644 > --- a/arch/powerpc/include/asm/interrupt.h > +++ b/arch/powerpc/include/asm/interrupt.h > @@ -336,6 +336,14 @@ static inline void interrupt_nmi_enter_prepare(struct pt_regs *regs, struct inte > if (IS_ENABLED(CONFIG_KASAN)) > return; > > + /* > + * Likewise, do not use it in real mode if percpu first chunk is not > + * embedded. With CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there > + * are chances where percpu allocation can come from vmalloc area. > + */ > + if (IS_ENABLED(CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK) && !is_embed_first_chunk) I think this would be clearer if it was inverted, eg: if (percpu_first_chunk_is_paged) return; That way you shouldn't need to check CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK here. Instead it can be part of the ifdef in the header. > @@ -351,6 +359,8 @@ static inline void interrupt_nmi_exit_prepare(struct pt_regs *regs, struct inter > // no nmi_exit for a pseries hash guest taking a real mode exception > } else if (IS_ENABLED(CONFIG_KASAN)) { > // no nmi_exit for KASAN in real mode > + } else if (IS_ENABLED(CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK) && !is_embed_first_chunk) { > + // no nmi_exit if percpu first chunk is not embedded > } else { > nmi_exit(); > } > diff --git a/arch/powerpc/include/asm/percpu.h b/arch/powerpc/include/asm/percpu.h > index 8e5b7d0b851c6..e24063eb0b33b 100644 > --- a/arch/powerpc/include/asm/percpu.h > +++ b/arch/powerpc/include/asm/percpu.h > @@ -15,6 +15,16 @@ > #endif /* CONFIG_SMP */ > #endif /* __powerpc64__ */ > > +#ifdef CONFIG_PPC64 > +#include <linux/jump_label.h> > +DECLARE_STATIC_KEY_FALSE(__percpu_embed_first_chunk); > + > +#define is_embed_first_chunk \ > + (static_key_enabled(&__percpu_embed_first_chunk.key)) > +#else > +#define is_embed_first_chunk true > +#endif /* CONFIG_PPC64 */ > + Something like: #ifdef CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK #include <linux/jump_label.h> DECLARE_STATIC_KEY_FALSE(__percpu_first_chunk_is_paged); #define percpu_first_chunk_is_paged \ (static_key_enabled(&__percpu_first_chunk_is_paged.key)) #else #define percpu_first_chunk_is_paged false #endif /* CONFIG_PPC64 */ > diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c > index 2f19d5e944852..e04f0ff69d4b6 100644 > --- a/arch/powerpc/kernel/setup_64.c > +++ b/arch/powerpc/kernel/setup_64.c > @@ -834,6 +834,7 @@ static __init int pcpu_cpu_to_node(int cpu) > > unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; > EXPORT_SYMBOL(__per_cpu_offset); > +DEFINE_STATIC_KEY_FALSE(__percpu_embed_first_chunk); > > void __init setup_per_cpu_areas(void) > { > @@ -869,6 +870,8 @@ void __init setup_per_cpu_areas(void) > pr_warn("PERCPU: %s allocator failed (%d), " > "falling back to page size\n", > pcpu_fc_names[pcpu_chosen_fc], rc); > + else > + static_key_enable(&__percpu_embed_first_chunk.key); > } > > if (rc < 0) Finally, the current patch breaks the microwatt build: $ make microwatt_defconfig ; make -s -j (nproc) make[1]: Entering directory '/home/michael/linux/.build' GEN Makefile # # configuration written to .config # make[1]: Leaving directory '/home/michael/linux/.build' ld: arch/powerpc/kernel/traps.o:(.toc+0x0): undefined reference to `__percpu_embed_first_chunk' ld: arch/powerpc/kernel/mce.o:(.toc+0x0): undefined reference to `__percpu_embed_first_chunk' make[3]: *** [../scripts/Makefile.vmlinux:37: vmlinux] Error 1 I guess because it has CONFIG_JUMP_LABEL=n? cheers
On 3/7/24 5:13 PM, Michael Ellerman wrote: > Hi Mahesh, > > Mahesh Salgaonkar <mahesh@linux.ibm.com> writes: >> nmi_enter()/nmi_exit() touches per cpu variables which can lead to kernel >> crash when invoked during real mode interrupt handling (e.g. early HMI/MCE >> interrupt handler) if percpu allocation comes from vmalloc area. >> >> Early HMI/MCE handlers are called through DEFINE_INTERRUPT_HANDLER_NMI() >> wrapper which invokes nmi_enter/nmi_exit calls. We don't see any issue when >> percpu allocation is from the embedded first chunk. However with >> CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there are chances where percpu >> allocation can come from the vmalloc area. >> >> With kernel command line "percpu_alloc=page" we can force percpu allocation >> to come from vmalloc area and can see kernel crash in machine_check_early: >> >> [ 1.215714] NIP [c000000000e49eb4] rcu_nmi_enter+0x24/0x110 >> [ 1.215717] LR [c0000000000461a0] machine_check_early+0xf0/0x2c0 >> [ 1.215719] --- interrupt: 200 >> [ 1.215720] [c000000fffd73180] [0000000000000000] 0x0 (unreliable) >> [ 1.215722] [c000000fffd731b0] [0000000000000000] 0x0 >> [ 1.215724] [c000000fffd73210] [c000000000008364] machine_check_early_common+0x134/0x1f8 >> >> Fix this by avoiding use of nmi_enter()/nmi_exit() in real mode if percpu >> first chunk is not embedded. > > My system (powernv) doesn't even boot with percpu_alloc=page. > Can you share the crash details? > AFAIK the only reason we added support for it was to handle 4K kernels > with HPT. See commit eb553f16973a ("powerpc/64/mm: implement page > mapping percpu first chunk allocator"). > > So I wonder if we should change the Kconfig to only offer it as an > option in that case, and change the logic in setup_per_cpu_areas() to > only use it as a last resort. > > I guess we probably still need this commit though, even if just for 4K HPT. > > We have also observed some error when we have large gap between the start memory of NUMA nodes. That made the percpu offset really large causing boot failures even on 64K. -aneesh
On 2024-03-07 22:43:07 Thu, Michael Ellerman wrote: > > diff --git a/arch/powerpc/include/asm/interrupt.h b/arch/powerpc/include/asm/interrupt.h > > index a4196ab1d0167..0b96464ff0339 100644 > > --- a/arch/powerpc/include/asm/interrupt.h > > +++ b/arch/powerpc/include/asm/interrupt.h > > @@ -336,6 +336,14 @@ static inline void interrupt_nmi_enter_prepare(struct pt_regs *regs, struct inte > > if (IS_ENABLED(CONFIG_KASAN)) > > return; > > > > + /* > > + * Likewise, do not use it in real mode if percpu first chunk is not > > + * embedded. With CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there > > + * are chances where percpu allocation can come from vmalloc area. > > + */ > > + if (IS_ENABLED(CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK) && !is_embed_first_chunk) > > I think this would be clearer if it was inverted, eg: > > if (percpu_first_chunk_is_paged) > return; Agree. > > That way you shouldn't need to check CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK here. > Instead it can be part of the ifdef in the header. > > > @@ -351,6 +359,8 @@ static inline void interrupt_nmi_exit_prepare(struct pt_regs *regs, struct inter > > // no nmi_exit for a pseries hash guest taking a real mode exception > > } else if (IS_ENABLED(CONFIG_KASAN)) { > > // no nmi_exit for KASAN in real mode > > + } else if (IS_ENABLED(CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK) && !is_embed_first_chunk) { > > + // no nmi_exit if percpu first chunk is not embedded > > } else { > > nmi_exit(); > > } > > diff --git a/arch/powerpc/include/asm/percpu.h b/arch/powerpc/include/asm/percpu.h > > index 8e5b7d0b851c6..e24063eb0b33b 100644 > > --- a/arch/powerpc/include/asm/percpu.h > > +++ b/arch/powerpc/include/asm/percpu.h > > @@ -15,6 +15,16 @@ > > #endif /* CONFIG_SMP */ > > #endif /* __powerpc64__ */ > > > > +#ifdef CONFIG_PPC64 > > +#include <linux/jump_label.h> > > +DECLARE_STATIC_KEY_FALSE(__percpu_embed_first_chunk); > > + > > +#define is_embed_first_chunk \ > > + (static_key_enabled(&__percpu_embed_first_chunk.key)) > > +#else > > +#define is_embed_first_chunk true > > +#endif /* CONFIG_PPC64 */ > > + > > Something like: > > #ifdef CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK > #include <linux/jump_label.h> > DECLARE_STATIC_KEY_FALSE(__percpu_first_chunk_is_paged); > > #define percpu_first_chunk_is_paged \ > (static_key_enabled(&__percpu_first_chunk_is_paged.key)) > #else > #define percpu_first_chunk_is_paged false > #endif /* CONFIG_PPC64 */ Sure, will fix it. > > > diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c > > index 2f19d5e944852..e04f0ff69d4b6 100644 > > --- a/arch/powerpc/kernel/setup_64.c > > +++ b/arch/powerpc/kernel/setup_64.c > > @@ -834,6 +834,7 @@ static __init int pcpu_cpu_to_node(int cpu) > > > > unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; > > EXPORT_SYMBOL(__per_cpu_offset); > > +DEFINE_STATIC_KEY_FALSE(__percpu_embed_first_chunk); > > > > void __init setup_per_cpu_areas(void) > > { > > @@ -869,6 +870,8 @@ void __init setup_per_cpu_areas(void) > > pr_warn("PERCPU: %s allocator failed (%d), " > > "falling back to page size\n", > > pcpu_fc_names[pcpu_chosen_fc], rc); > > + else > > + static_key_enable(&__percpu_embed_first_chunk.key); > > } > > > > if (rc < 0) > > Finally, the current patch breaks the microwatt build: > > $ make microwatt_defconfig ; make -s -j (nproc) > make[1]: Entering directory '/home/michael/linux/.build' > GEN Makefile > # > # configuration written to .config > # > make[1]: Leaving directory '/home/michael/linux/.build' > ld: arch/powerpc/kernel/traps.o:(.toc+0x0): undefined reference to `__percpu_embed_first_chunk' > ld: arch/powerpc/kernel/mce.o:(.toc+0x0): undefined reference to `__percpu_embed_first_chunk' > make[3]: *** [../scripts/Makefile.vmlinux:37: vmlinux] Error 1 > > I guess because it has CONFIG_JUMP_LABEL=n? Even with CONFIG_JUMP_LABEL=n it should still work. Let me take look and fix this for microwatt build. Thanks for your review. -Mahesh. > > cheers
Aneesh Kumar K V <aneesh.kumar@linux.ibm.com> writes: > On 3/7/24 5:13 PM, Michael Ellerman wrote: >> Mahesh Salgaonkar <mahesh@linux.ibm.com> writes: >>> nmi_enter()/nmi_exit() touches per cpu variables which can lead to kernel >>> crash when invoked during real mode interrupt handling (e.g. early HMI/MCE >>> interrupt handler) if percpu allocation comes from vmalloc area. >>> >>> Early HMI/MCE handlers are called through DEFINE_INTERRUPT_HANDLER_NMI() >>> wrapper which invokes nmi_enter/nmi_exit calls. We don't see any issue when >>> percpu allocation is from the embedded first chunk. However with >>> CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there are chances where percpu >>> allocation can come from the vmalloc area. >>> >>> With kernel command line "percpu_alloc=page" we can force percpu allocation >>> to come from vmalloc area and can see kernel crash in machine_check_early: >>> >>> [ 1.215714] NIP [c000000000e49eb4] rcu_nmi_enter+0x24/0x110 >>> [ 1.215717] LR [c0000000000461a0] machine_check_early+0xf0/0x2c0 >>> [ 1.215719] --- interrupt: 200 >>> [ 1.215720] [c000000fffd73180] [0000000000000000] 0x0 (unreliable) >>> [ 1.215722] [c000000fffd731b0] [0000000000000000] 0x0 >>> [ 1.215724] [c000000fffd73210] [c000000000008364] machine_check_early_common+0x134/0x1f8 >>> >>> Fix this by avoiding use of nmi_enter()/nmi_exit() in real mode if percpu >>> first chunk is not embedded. >> >> My system (powernv) doesn't even boot with percpu_alloc=page. > > > Can you share the crash details? Yes but it's not pretty :) [ 1.725257][ T714] systemd-journald[714]: Collecting audit messages is disabled. [ 1.729401][ T1] systemd[1]: Finished systemd-tmpfiles-setup-dev.service - Create Static Device Nodes in /dev. [^[[0;32m OK ^[[0m] Finished ^[[0;1;39msystemd-tmpfiles-…reate Static Device Nodes in /dev. [ 1.773902][ C22] Disabling lock debugging due to kernel taint [ 1.773905][ C23] Oops: Machine check, sig: 7 [#1] [ 1.773911][ C23] LE PAGE_SIZE=64K MMU=Radix SMP NR_CPUS=2048 NUMA PowerNV [ 1.773916][ C23] Modules linked in: [ 1.773920][ C23] CPU: 23 PID: 0 Comm: swapper/23 Tainted: G M 6.8.0-rc7-02500-g23515c370cbb #1 [ 1.773924][ C23] Hardware name: 8335-GTH POWER9 0x4e1202 opal:skiboot-v6.5.3-35-g1851b2a06 PowerNV [ 1.773926][ C23] NIP: 0000000000000000 LR: 0000000000000000 CTR: 0000000000000000 [ 1.773929][ C23] REGS: c000000fffa6ef50 TRAP: 0000 Tainted: G M (6.8.0-rc7-02500-g23515c370cbb) [ 1.773932][ C23] MSR: 0000000000000000 <> CR: 00000000 XER: 00000000 [ 1.773937][ C23] CFAR: 0000000000000000 IRQMASK: 3 [ 1.773937][ C23] GPR00: 0000000000000000 c000000fffa6efe0 c000000fffa6efb0 0000000000000000 [ 1.773937][ C23] GPR04: c00000000003d8c0 c000000001f5f000 0000000000000000 0000000000000103 [ 1.773937][ C23] GPR08: 0000000000000003 653a0d962a590300 0000000000000000 0000000000000000 [ 1.773937][ C23] GPR12: c000000fffa6f280 0000000000000000 c0000000000084a4 0000000000000000 [ 1.773937][ C23] GPR16: 0000000053474552 0000000000000000 c00000000003d8c0 c000000fffa6f280 [ 1.773937][ C23] GPR20: c000000001f5f000 c000000fffa6f340 c000000fffa6f2e8 0000000000000000 [ 1.773937][ C23] GPR24: 0007fffffecf0000 c0000000065bbb80 0000000000550102 c000000002172b20 [ 1.773937][ C23] GPR28: 0000000000000000 0000000053474552 0000000000000000 c000000ffffc6d80 [ 1.773982][ C23] NIP [0000000000000000] 0x0 [ 1.773988][ C23] LR [0000000000000000] 0x0 [ 1.773990][ C23] Call Trace: [ 1.773991][ C23] [c000000fffa6efe0] [c000000001f5f000] .TOC.+0x0/0xa1000 (unreliable) [ 1.773999][ C23] Code: XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX [ 1.774021][ C23] ---[ end trace 0000000000000000 ]--- Something has gone badly wrong. That was a test kernel with some other commits, but nothing that should cause that. Removing percpu_alloc=page fix it. It's based on fddff98e83b4b4d54470902ea0d520c4d423ca3b. >> AFAIK the only reason we added support for it was to handle 4K kernels >> with HPT. See commit eb553f16973a ("powerpc/64/mm: implement page >> mapping percpu first chunk allocator"). >> >> So I wonder if we should change the Kconfig to only offer it as an >> option in that case, and change the logic in setup_per_cpu_areas() to >> only use it as a last resort. >> >> I guess we probably still need this commit though, even if just for 4K HPT. >> >> > We have also observed some error when we have large gap between the start memory of > NUMA nodes. That made the percpu offset really large causing boot failures even on 64K. Yeah, I have vague memories of that :) cheers
On 2024-03-08 19:08:50 Fri, Michael Ellerman wrote: > Aneesh Kumar K V <aneesh.kumar@linux.ibm.com> writes: > > On 3/7/24 5:13 PM, Michael Ellerman wrote: > >> Mahesh Salgaonkar <mahesh@linux.ibm.com> writes: > >>> nmi_enter()/nmi_exit() touches per cpu variables which can lead to kernel > >>> crash when invoked during real mode interrupt handling (e.g. early HMI/MCE > >>> interrupt handler) if percpu allocation comes from vmalloc area. > >>> > >>> Early HMI/MCE handlers are called through DEFINE_INTERRUPT_HANDLER_NMI() > >>> wrapper which invokes nmi_enter/nmi_exit calls. We don't see any issue when > >>> percpu allocation is from the embedded first chunk. However with > >>> CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there are chances where percpu > >>> allocation can come from the vmalloc area. > >>> > >>> With kernel command line "percpu_alloc=page" we can force percpu allocation > >>> to come from vmalloc area and can see kernel crash in machine_check_early: > >>> > >>> [ 1.215714] NIP [c000000000e49eb4] rcu_nmi_enter+0x24/0x110 > >>> [ 1.215717] LR [c0000000000461a0] machine_check_early+0xf0/0x2c0 > >>> [ 1.215719] --- interrupt: 200 > >>> [ 1.215720] [c000000fffd73180] [0000000000000000] 0x0 (unreliable) > >>> [ 1.215722] [c000000fffd731b0] [0000000000000000] 0x0 > >>> [ 1.215724] [c000000fffd73210] [c000000000008364] machine_check_early_common+0x134/0x1f8 > >>> > >>> Fix this by avoiding use of nmi_enter()/nmi_exit() in real mode if percpu > >>> first chunk is not embedded. > >> > >> My system (powernv) doesn't even boot with percpu_alloc=page. > > > > > > Can you share the crash details? > > Yes but it's not pretty :) > > [ 1.725257][ T714] systemd-journald[714]: Collecting audit messages is disabled. > [ 1.729401][ T1] systemd[1]: Finished systemd-tmpfiles-setup-dev.service - Create Static Device Nodes in /dev. > [^[[0;32m OK ^[[0m] Finished ^[[0;1;39msystemd-tmpfiles-…reate Static Device Nodes in /dev. > [ 1.773902][ C22] Disabling lock debugging due to kernel taint > [ 1.773905][ C23] Oops: Machine check, sig: 7 [#1] > [ 1.773911][ C23] LE PAGE_SIZE=64K MMU=Radix SMP NR_CPUS=2048 NUMA PowerNV > [ 1.773916][ C23] Modules linked in: > [ 1.773920][ C23] CPU: 23 PID: 0 Comm: swapper/23 Tainted: G M 6.8.0-rc7-02500-g23515c370cbb #1 > [ 1.773924][ C23] Hardware name: 8335-GTH POWER9 0x4e1202 opal:skiboot-v6.5.3-35-g1851b2a06 PowerNV > [ 1.773926][ C23] NIP: 0000000000000000 LR: 0000000000000000 CTR: 0000000000000000 > [ 1.773929][ C23] REGS: c000000fffa6ef50 TRAP: 0000 Tainted: G M (6.8.0-rc7-02500-g23515c370cbb) > [ 1.773932][ C23] MSR: 0000000000000000 <> CR: 00000000 XER: 00000000 > [ 1.773937][ C23] CFAR: 0000000000000000 IRQMASK: 3 > [ 1.773937][ C23] GPR00: 0000000000000000 c000000fffa6efe0 c000000fffa6efb0 0000000000000000 > [ 1.773937][ C23] GPR04: c00000000003d8c0 c000000001f5f000 0000000000000000 0000000000000103 > [ 1.773937][ C23] GPR08: 0000000000000003 653a0d962a590300 0000000000000000 0000000000000000 > [ 1.773937][ C23] GPR12: c000000fffa6f280 0000000000000000 c0000000000084a4 0000000000000000 > [ 1.773937][ C23] GPR16: 0000000053474552 0000000000000000 c00000000003d8c0 c000000fffa6f280 > [ 1.773937][ C23] GPR20: c000000001f5f000 c000000fffa6f340 c000000fffa6f2e8 0000000000000000 > [ 1.773937][ C23] GPR24: 0007fffffecf0000 c0000000065bbb80 0000000000550102 c000000002172b20 > [ 1.773937][ C23] GPR28: 0000000000000000 0000000053474552 0000000000000000 c000000ffffc6d80 > [ 1.773982][ C23] NIP [0000000000000000] 0x0 > [ 1.773988][ C23] LR [0000000000000000] 0x0 > [ 1.773990][ C23] Call Trace: > [ 1.773991][ C23] [c000000fffa6efe0] [c000000001f5f000] .TOC.+0x0/0xa1000 (unreliable) > [ 1.773999][ C23] Code: XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX > [ 1.774021][ C23] ---[ end trace 0000000000000000 ]--- > > Something has gone badly wrong. > > That was a test kernel with some other commits, but nothing that should > cause that. Removing percpu_alloc=page fix it. So, when I try this without my patch "Avoid nmi_enter/nmi_exit in real mode interrupt", I see this getting recreated. However, I was not able to recrate this even once with my changes. Are you able to see this crash with my patch ? Thanks, -Mahesh.
diff --git a/arch/powerpc/include/asm/interrupt.h b/arch/powerpc/include/asm/interrupt.h index a4196ab1d0167..0b96464ff0339 100644 --- a/arch/powerpc/include/asm/interrupt.h +++ b/arch/powerpc/include/asm/interrupt.h @@ -336,6 +336,14 @@ static inline void interrupt_nmi_enter_prepare(struct pt_regs *regs, struct inte if (IS_ENABLED(CONFIG_KASAN)) return; + /* + * Likewise, do not use it in real mode if percpu first chunk is not + * embedded. With CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there + * are chances where percpu allocation can come from vmalloc area. + */ + if (IS_ENABLED(CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK) && !is_embed_first_chunk) + return; + /* Otherwise, it should be safe to call it */ nmi_enter(); } @@ -351,6 +359,8 @@ static inline void interrupt_nmi_exit_prepare(struct pt_regs *regs, struct inter // no nmi_exit for a pseries hash guest taking a real mode exception } else if (IS_ENABLED(CONFIG_KASAN)) { // no nmi_exit for KASAN in real mode + } else if (IS_ENABLED(CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK) && !is_embed_first_chunk) { + // no nmi_exit if percpu first chunk is not embedded } else { nmi_exit(); } diff --git a/arch/powerpc/include/asm/percpu.h b/arch/powerpc/include/asm/percpu.h index 8e5b7d0b851c6..e24063eb0b33b 100644 --- a/arch/powerpc/include/asm/percpu.h +++ b/arch/powerpc/include/asm/percpu.h @@ -15,6 +15,16 @@ #endif /* CONFIG_SMP */ #endif /* __powerpc64__ */ +#ifdef CONFIG_PPC64 +#include <linux/jump_label.h> +DECLARE_STATIC_KEY_FALSE(__percpu_embed_first_chunk); + +#define is_embed_first_chunk \ + (static_key_enabled(&__percpu_embed_first_chunk.key)) +#else +#define is_embed_first_chunk true +#endif /* CONFIG_PPC64 */ + #include <asm-generic/percpu.h> #include <asm/paca.h> diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c index 2f19d5e944852..e04f0ff69d4b6 100644 --- a/arch/powerpc/kernel/setup_64.c +++ b/arch/powerpc/kernel/setup_64.c @@ -834,6 +834,7 @@ static __init int pcpu_cpu_to_node(int cpu) unsigned long __per_cpu_offset[NR_CPUS] __read_mostly; EXPORT_SYMBOL(__per_cpu_offset); +DEFINE_STATIC_KEY_FALSE(__percpu_embed_first_chunk); void __init setup_per_cpu_areas(void) { @@ -869,6 +870,8 @@ void __init setup_per_cpu_areas(void) pr_warn("PERCPU: %s allocator failed (%d), " "falling back to page size\n", pcpu_fc_names[pcpu_chosen_fc], rc); + else + static_key_enable(&__percpu_embed_first_chunk.key); } if (rc < 0)
nmi_enter()/nmi_exit() touches per cpu variables which can lead to kernel crash when invoked during real mode interrupt handling (e.g. early HMI/MCE interrupt handler) if percpu allocation comes from vmalloc area. Early HMI/MCE handlers are called through DEFINE_INTERRUPT_HANDLER_NMI() wrapper which invokes nmi_enter/nmi_exit calls. We don't see any issue when percpu allocation is from the embedded first chunk. However with CONFIG_NEED_PER_CPU_PAGE_FIRST_CHUNK enabled there are chances where percpu allocation can come from the vmalloc area. With kernel command line "percpu_alloc=page" we can force percpu allocation to come from vmalloc area and can see kernel crash in machine_check_early: [ 1.215714] NIP [c000000000e49eb4] rcu_nmi_enter+0x24/0x110 [ 1.215717] LR [c0000000000461a0] machine_check_early+0xf0/0x2c0 [ 1.215719] --- interrupt: 200 [ 1.215720] [c000000fffd73180] [0000000000000000] 0x0 (unreliable) [ 1.215722] [c000000fffd731b0] [0000000000000000] 0x0 [ 1.215724] [c000000fffd73210] [c000000000008364] machine_check_early_common+0x134/0x1f8 Fix this by avoiding use of nmi_enter()/nmi_exit() in real mode if percpu first chunk is not embedded. Signed-off-by: Mahesh Salgaonkar <mahesh@linux.ibm.com> --- Changes in v4: - Fix coding style issues. Changes in v3: - Address comments from Christophe Leroy to avoid using #ifdefs in the code - v2 at https://lore.kernel.org/linuxppc-dev/20240205053647.1763446-1-mahesh@linux.ibm.com/ Changes in v2: - Rebase to upstream master - Use jump_labels, if CONFIG_JUMP_LABEL is enabled, to avoid redoing the embed first chunk test at each interrupt entry. - v1 is at https://lore.kernel.org/linuxppc-dev/164578465828.74956.6065296024817333750.stgit@jupiter/ --- arch/powerpc/include/asm/interrupt.h | 10 ++++++++++ arch/powerpc/include/asm/percpu.h | 10 ++++++++++ arch/powerpc/kernel/setup_64.c | 3 +++ 3 files changed, 23 insertions(+)