The patch titled Subject: x86/mm: Split vmalloc_sync_all() has been added to the -mm tree. Its filename is x86-mm-split-vmalloc_sync_all.patch
This patch should soon appear at http://ozlabs.org/~akpm/mmots/broken-out/x86-mm-split-vmalloc_sync_all.patch and later at http://ozlabs.org/~akpm/mmotm/broken-out/x86-mm-split-vmalloc_sync_all.patch
Before you just go and hit "reply", please: a) Consider who else should be cc'ed b) Prefer to cc a suitable mailing list as well c) Ideally: find the original patch on the mailing list and do a reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next and is updated there every 3-4 working days
------------------------------------------------------ From: Joerg Roedel jroedel@suse.de Subject: x86/mm: Split vmalloc_sync_all()
Commit 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in __purge_vmap_area_lazy()") introduced a call to vmalloc_sync_all() in the vunmap() code-path. While this change was necessary to maintain correctness on x86-32-pae kernels, it also adds additional cycles for architectures that don't need it.
Specifically on x86-64 with CONFIG_VMAP_STACK=y some people reported severe performance regressions in micro-benchmarks because it now also calls the x86-64 implementation of vmalloc_sync_all() on vunmap(). But the vmalloc_sync_all() implementation on x86-64 is only needed for newly created mappings.
To avoid the unnecessary work on x86-64 and to gain the performance back, split up vmalloc_sync_all() into two functions:
* vmalloc_sync_mappings(), and * vmalloc_sync_unmappings()
Most call-sites to vmalloc_sync_all() only care about new mappings being synchronized. The only exception is the new call-site added in the above mentioned commit.
Shile Zhang directed us to a report of an 80% regression in reaim throughput.
Link: http://lkml.kernel.org/r/20191009124418.8286-1-joro@8bytes.org Link: https://lists.01.org/hyperkitty/list/lkp@lists.01.org/thread/4D3JPPHBNOSPFK2... Link: http://lkml.kernel.org/r/20191113095530.228959-1-shile.zhang@linux.alibaba.c... Fixes: 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in __purge_vmap_area_lazy()") Signed-off-by: Joerg Roedel jroedel@suse.de Reported-by: kernel test robot oliver.sang@intel.com Reported-by: Shile Zhang shile.zhang@linux.alibaba.com Acked-by: Rafael J. Wysocki rafael.j.wysocki@intel.com [GHES] Cc: Dave Hansen dave.hansen@linux.intel.com Cc: Andy Lutomirski luto@kernel.org Cc: Peter Zijlstra peterz@infradead.org Cc: Thomas Gleixner tglx@linutronix.de Cc: Ingo Molnar mingo@redhat.com Cc: Borislav Petkov bp@alien8.de Cc: stable@vger.kernel.org Signed-off-by: Andrew Morton akpm@linux-foundation.org ---
arch/x86/mm/fault.c | 26 ++++++++++++++++++++++++-- drivers/acpi/apei/ghes.c | 2 +- include/linux/vmalloc.h | 5 +++-- kernel/notifier.c | 2 +- mm/nommu.c | 10 +++++++--- mm/vmalloc.c | 11 +++++++---- 6 files changed, 43 insertions(+), 13 deletions(-)
--- a/arch/x86/mm/fault.c~x86-mm-split-vmalloc_sync_all +++ a/arch/x86/mm/fault.c @@ -189,7 +189,7 @@ static inline pmd_t *vmalloc_sync_one(pg return pmd_k; }
-void vmalloc_sync_all(void) +static void vmalloc_sync(void) { unsigned long address;
@@ -216,6 +216,16 @@ void vmalloc_sync_all(void) } }
+void vmalloc_sync_mappings(void) +{ + vmalloc_sync(); +} + +void vmalloc_sync_unmappings(void) +{ + vmalloc_sync(); +} + /* * 32-bit: * @@ -318,11 +328,23 @@ out:
#else /* CONFIG_X86_64: */
-void vmalloc_sync_all(void) +void vmalloc_sync_mappings(void) { + /* + * 64-bit mappings might allocate new p4d/pud pages + * that need to be propagated to all tasks' PGDs. + */ sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END); }
+void vmalloc_sync_unmappings(void) +{ + /* + * Unmappings never allocate or free p4d/pud pages. + * No work is required here. + */ +} + /* * 64-bit: * --- a/drivers/acpi/apei/ghes.c~x86-mm-split-vmalloc_sync_all +++ a/drivers/acpi/apei/ghes.c @@ -171,7 +171,7 @@ int ghes_estatus_pool_init(int num_ghes) * New allocation must be visible in all pgd before it can be found by * an NMI allocating from the pool. */ - vmalloc_sync_all(); + vmalloc_sync_mappings();
rc = gen_pool_add(ghes_estatus_pool, addr, PAGE_ALIGN(len), -1); if (rc) --- a/include/linux/vmalloc.h~x86-mm-split-vmalloc_sync_all +++ a/include/linux/vmalloc.h @@ -126,8 +126,9 @@ extern int remap_vmalloc_range_partial(s
extern int remap_vmalloc_range(struct vm_area_struct *vma, void *addr, unsigned long pgoff); -void vmalloc_sync_all(void); - +void vmalloc_sync_mappings(void); +void vmalloc_sync_unmappings(void); + /* * Lowlevel-APIs (not for driver use!) */ --- a/kernel/notifier.c~x86-mm-split-vmalloc_sync_all +++ a/kernel/notifier.c @@ -554,7 +554,7 @@ NOKPROBE_SYMBOL(notify_die);
int register_die_notifier(struct notifier_block *nb) { - vmalloc_sync_all(); + vmalloc_sync_mappings(); return atomic_notifier_chain_register(&die_chain, nb); } EXPORT_SYMBOL_GPL(register_die_notifier); --- a/mm/nommu.c~x86-mm-split-vmalloc_sync_all +++ a/mm/nommu.c @@ -359,10 +359,14 @@ void vm_unmap_aliases(void) EXPORT_SYMBOL_GPL(vm_unmap_aliases);
/* - * Implement a stub for vmalloc_sync_all() if the architecture chose not to - * have one. + * Implement a stub for vmalloc_sync_[un]mapping() if the architecture + * chose not to have one. */ -void __weak vmalloc_sync_all(void) +void __weak vmalloc_sync_mappings(void) +{ +} + +void __weak vmalloc_sync_unmappings(void) { }
--- a/mm/vmalloc.c~x86-mm-split-vmalloc_sync_all +++ a/mm/vmalloc.c @@ -1259,7 +1259,7 @@ static bool __purge_vmap_area_lazy(unsig * First make sure the mappings are removed from all page-tables * before they are freed. */ - vmalloc_sync_all(); + vmalloc_sync_unmappings();
/* * TODO: to calculate a flush range without looping. @@ -3050,16 +3050,19 @@ int remap_vmalloc_range(struct vm_area_s EXPORT_SYMBOL(remap_vmalloc_range);
/* - * Implement a stub for vmalloc_sync_all() if the architecture chose not to - * have one. + * Implement stubs for vmalloc_sync_[un]mappings () if the architecture chose + * not to have one. * * The purpose of this function is to make sure the vmalloc area * mappings are identical in all page-tables in the system. */ -void __weak vmalloc_sync_all(void) +void __weak vmalloc_sync_mappings(void) { }
+void __weak vmalloc_sync_unmappings(void) +{ +}
static int f(pte_t *pte, unsigned long addr, void *data) { _
Patches currently in -mm which might be from jroedel@suse.de are
x86-mm-split-vmalloc_sync_all.patch