Commit 6dd29b3d authored by Ingo Molnar's avatar Ingo Molnar
Browse files

Revert "x86/mm/gup: Switch GUP to the generic get_user_page_fast() implementation"

This reverts commit 2947ba05.

Dan Williams reported dax-pmem kernel warnings with the following signature:

   WARNING: CPU: 8 PID: 245 at lib/percpu-refcount.c:155 percpu_ref_switch_to_atomic_rcu+0x1f5/0x200
   percpu ref (dax_pmem_percpu_release [dax_pmem]) <= 0 (0) after switching to atomic

... and bisected it to this commit, which suggests possible memory corruption
caused by the x86 fast-GUP conversion.

He also pointed out:

 "
  This is similar to the backtrace when we were not properly handling
  pud faults and was fixed with this commit: 220ced16

 "mm: fix
  get_user_pages() vs device-dax pud mappings"

  I've found some missing _devmap checks in the generic
  get_user_pages_fast() path, but this does not fix the regression
  [...]
 "

So given that there are known bugs, and a pretty robust looking bisection
points to this commit suggesting that are unknown bugs in the conversion
as well, revert it for the time being - we'll re-try in v4.13.
Reported-by: default avatarDan Williams <dan.j.williams@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: aneesh.kumar@linux.vnet.ibm.com
Cc: dann.frazier@canonical.com
Cc: dave.hansen@intel.com
Cc: steve.capper@linaro.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: default avatarIngo Molnar <mingo@kernel.org>
parent ace2fb5a
......@@ -1666,7 +1666,7 @@ config ARCH_SELECT_MEMORY_MODEL
config HAVE_ARCH_PFN_VALID
def_bool ARCH_HAS_HOLES_MEMORYMODEL || !SPARSEMEM
config HAVE_GENERIC_GUP
config HAVE_GENERIC_RCU_GUP
def_bool y
depends on ARM_LPAE
......
......@@ -205,7 +205,7 @@ config GENERIC_CALIBRATE_DELAY
config ZONE_DMA
def_bool y
config HAVE_GENERIC_GUP
config HAVE_GENERIC_RCU_GUP
def_bool y
config ARCH_DMA_ADDR_T_64BIT
......
......@@ -135,7 +135,7 @@ config PPC
select HAVE_FUNCTION_GRAPH_TRACER
select HAVE_FUNCTION_TRACER
select HAVE_GCC_PLUGINS
select HAVE_GENERIC_GUP
select HAVE_GENERIC_RCU_GUP
select HAVE_HW_BREAKPOINT if PERF_EVENTS && (PPC_BOOK3S || PPC_8xx)
select HAVE_IDE
select HAVE_IOREMAP_PROT
......
......@@ -2789,9 +2789,6 @@ config X86_DMA_REMAP
bool
depends on STA2X11
config HAVE_GENERIC_GUP
def_bool y
source "net/Kconfig"
source "drivers/Kconfig"
......
......@@ -220,6 +220,18 @@ static inline int vma_pkey(struct vm_area_struct *vma)
}
#endif
static inline bool __pkru_allows_pkey(u16 pkey, bool write)
{
u32 pkru = read_pkru();
if (!__pkru_allows_read(pkru, pkey))
return false;
if (write && !__pkru_allows_write(pkru, pkey))
return false;
return true;
}
/*
* We only want to enforce protection keys on the current process
* because we effectively have no access to PKRU for other
......
......@@ -212,51 +212,4 @@ static inline pud_t native_pudp_get_and_clear(pud_t *pudp)
#define __pte_to_swp_entry(pte) ((swp_entry_t){ (pte).pte_high })
#define __swp_entry_to_pte(x) ((pte_t){ { .pte_high = (x).val } })
#define gup_get_pte gup_get_pte
/*
* WARNING: only to be used in the get_user_pages_fast() implementation.
*
* With get_user_pages_fast(), we walk down the pagetables without taking
* any locks. For this we would like to load the pointers atomically,
* but that is not possible (without expensive cmpxchg8b) on PAE. What
* we do have is the guarantee that a PTE will only either go from not
* present to present, or present to not present or both -- it will not
* switch to a completely different present page without a TLB flush in
* between; something that we are blocking by holding interrupts off.
*
* Setting ptes from not present to present goes:
*
* ptep->pte_high = h;
* smp_wmb();
* ptep->pte_low = l;
*
* And present to not present goes:
*
* ptep->pte_low = 0;
* smp_wmb();
* ptep->pte_high = 0;
*
* We must ensure here that the load of pte_low sees 'l' iff pte_high
* sees 'h'. We load pte_high *after* loading pte_low, which ensures we
* don't see an older value of pte_high. *Then* we recheck pte_low,
* which ensures that we haven't picked up a changed pte high. We might
* have gotten rubbish values from pte_low and pte_high, but we are
* guaranteed that pte_low will not have the present bit set *unless*
* it is 'l'. Because get_user_pages_fast() only operates on present ptes
* we're safe.
*/
static inline pte_t gup_get_pte(pte_t *ptep)
{
pte_t pte;
do {
pte.pte_low = ptep->pte_low;
smp_rmb();
pte.pte_high = ptep->pte_high;
smp_rmb();
} while (unlikely(pte.pte_low != ptep->pte_low));
return pte;
}
#endif /* _ASM_X86_PGTABLE_3LEVEL_H */
......@@ -244,11 +244,6 @@ static inline int pud_devmap(pud_t pud)
return 0;
}
#endif
static inline int pgd_devmap(pgd_t pgd)
{
return 0;
}
#endif
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
......@@ -1190,54 +1185,6 @@ static inline u16 pte_flags_pkey(unsigned long pte_flags)
#endif
}
static inline bool __pkru_allows_pkey(u16 pkey, bool write)
{
u32 pkru = read_pkru();
if (!__pkru_allows_read(pkru, pkey))
return false;
if (write && !__pkru_allows_write(pkru, pkey))
return false;
return true;
}
/*
* 'pteval' can come from a PTE, PMD or PUD. We only check
* _PAGE_PRESENT, _PAGE_USER, and _PAGE_RW in here which are the
* same value on all 3 types.
*/
static inline bool __pte_access_permitted(unsigned long pteval, bool write)
{
unsigned long need_pte_bits = _PAGE_PRESENT|_PAGE_USER;
if (write)
need_pte_bits |= _PAGE_RW;
if ((pteval & need_pte_bits) != need_pte_bits)
return 0;
return __pkru_allows_pkey(pte_flags_pkey(pteval), write);
}
#define pte_access_permitted pte_access_permitted
static inline bool pte_access_permitted(pte_t pte, bool write)
{
return __pte_access_permitted(pte_val(pte), write);
}
#define pmd_access_permitted pmd_access_permitted
static inline bool pmd_access_permitted(pmd_t pmd, bool write)
{
return __pte_access_permitted(pmd_val(pmd), write);
}
#define pud_access_permitted pud_access_permitted
static inline bool pud_access_permitted(pud_t pud, bool write)
{
return __pte_access_permitted(pud_val(pud), write);
}
#include <asm-generic/pgtable.h>
#endif /* __ASSEMBLY__ */
......
......@@ -227,20 +227,6 @@ extern void cleanup_highmap(void);
extern void init_extra_mapping_uc(unsigned long phys, unsigned long size);
extern void init_extra_mapping_wb(unsigned long phys, unsigned long size);
#define gup_fast_permitted gup_fast_permitted
static inline bool gup_fast_permitted(unsigned long start, int nr_pages,
int write)
{
unsigned long len, end;
len = (unsigned long)nr_pages << PAGE_SHIFT;
end = start + len;
if (end < start)
return false;
if (end >> __VIRTUAL_MASK_SHIFT)
return false;
return true;
}
#endif /* !__ASSEMBLY__ */
#endif /* _ASM_X86_PGTABLE_64_H */
......@@ -2,7 +2,7 @@
KCOV_INSTRUMENT_tlb.o := n
obj-y := init.o init_$(BITS).o fault.o ioremap.o extable.o pageattr.o mmap.o \
pat.o pgtable.o physaddr.o setup_nx.o tlb.o
pat.o pgtable.o physaddr.o gup.o setup_nx.o tlb.o
# Make sure __phys_addr has no stackprotector
nostackp := $(call cc-option, -fno-stack-protector)
......
/*
* Lockless get_user_pages_fast for x86
*
* Copyright (C) 2008 Nick Piggin
* Copyright (C) 2008 Novell Inc.
*/
#include <linux/sched.h>
#include <linux/mm.h>
#include <linux/vmstat.h>
#include <linux/highmem.h>
#include <linux/swap.h>
#include <linux/memremap.h>
#include <asm/mmu_context.h>
#include <asm/pgtable.h>
static inline pte_t gup_get_pte(pte_t *ptep)
{
#ifndef CONFIG_X86_PAE
return READ_ONCE(*ptep);
#else
/*
* With get_user_pages_fast, we walk down the pagetables without taking
* any locks. For this we would like to load the pointers atomically,
* but that is not possible (without expensive cmpxchg8b) on PAE. What
* we do have is the guarantee that a pte will only either go from not
* present to present, or present to not present or both -- it will not
* switch to a completely different present page without a TLB flush in
* between; something that we are blocking by holding interrupts off.
*
* Setting ptes from not present to present goes:
* ptep->pte_high = h;
* smp_wmb();
* ptep->pte_low = l;
*
* And present to not present goes:
* ptep->pte_low = 0;
* smp_wmb();
* ptep->pte_high = 0;
*
* We must ensure here that the load of pte_low sees l iff pte_high
* sees h. We load pte_high *after* loading pte_low, which ensures we
* don't see an older value of pte_high. *Then* we recheck pte_low,
* which ensures that we haven't picked up a changed pte high. We might
* have got rubbish values from pte_low and pte_high, but we are
* guaranteed that pte_low will not have the present bit set *unless*
* it is 'l'. And get_user_pages_fast only operates on present ptes, so
* we're safe.
*
* gup_get_pte should not be used or copied outside gup.c without being
* very careful -- it does not atomically load the pte or anything that
* is likely to be useful for you.
*/
pte_t pte;
retry:
pte.pte_low = ptep->pte_low;
smp_rmb();
pte.pte_high = ptep->pte_high;
smp_rmb();
if (unlikely(pte.pte_low != ptep->pte_low))
goto retry;
return pte;
#endif
}
static void undo_dev_pagemap(int *nr, int nr_start, struct page **pages)
{
while ((*nr) - nr_start) {
struct page *page = pages[--(*nr)];
ClearPageReferenced(page);
put_page(page);
}
}
/*
* 'pteval' can come from a pte, pmd, pud or p4d. We only check
* _PAGE_PRESENT, _PAGE_USER, and _PAGE_RW in here which are the
* same value on all 4 types.
*/
static inline int pte_allows_gup(unsigned long pteval, int write)
{
unsigned long need_pte_bits = _PAGE_PRESENT|_PAGE_USER;
if (write)
need_pte_bits |= _PAGE_RW;
if ((pteval & need_pte_bits) != need_pte_bits)
return 0;
/* Check memory protection keys permissions. */
if (!__pkru_allows_pkey(pte_flags_pkey(pteval), write))
return 0;
return 1;
}
/*
* The performance critical leaf functions are made noinline otherwise gcc
* inlines everything into a single function which results in too much
* register pressure.
*/
static noinline int gup_pte_range(pmd_t pmd, unsigned long addr,
unsigned long end, int write, struct page **pages, int *nr)
{
struct dev_pagemap *pgmap = NULL;
int nr_start = *nr, ret = 0;
pte_t *ptep, *ptem;
/*
* Keep the original mapped PTE value (ptem) around since we
* might increment ptep off the end of the page when finishing
* our loop iteration.
*/
ptem = ptep = pte_offset_map(&pmd, addr);
do {
pte_t pte = gup_get_pte(ptep);
struct page *page;
/* Similar to the PMD case, NUMA hinting must take slow path */
if (pte_protnone(pte))
break;
if (!pte_allows_gup(pte_val(pte), write))
break;
if (pte_devmap(pte)) {
pgmap = get_dev_pagemap(pte_pfn(pte), pgmap);
if (unlikely(!pgmap)) {
undo_dev_pagemap(nr, nr_start, pages);
break;
}
} else if (pte_special(pte))
break;
VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
page = pte_page(pte);
get_page(page);
put_dev_pagemap(pgmap);
SetPageReferenced(page);
pages[*nr] = page;
(*nr)++;
} while (ptep++, addr += PAGE_SIZE, addr != end);
if (addr == end)
ret = 1;
pte_unmap(ptem);
return ret;
}
static inline void get_head_page_multiple(struct page *page, int nr)
{
VM_BUG_ON_PAGE(page != compound_head(page), page);
VM_BUG_ON_PAGE(page_count(page) == 0, page);
page_ref_add(page, nr);
SetPageReferenced(page);
}
static int __gup_device_huge(unsigned long pfn, unsigned long addr,
unsigned long end, struct page **pages, int *nr)
{
int nr_start = *nr;
struct dev_pagemap *pgmap = NULL;
do {
struct page *page = pfn_to_page(pfn);
pgmap = get_dev_pagemap(pfn, pgmap);
if (unlikely(!pgmap)) {
undo_dev_pagemap(nr, nr_start, pages);
return 0;
}
SetPageReferenced(page);
pages[*nr] = page;
get_page(page);
put_dev_pagemap(pgmap);
(*nr)++;
pfn++;
} while (addr += PAGE_SIZE, addr != end);
return 1;
}
static int __gup_device_huge_pmd(pmd_t pmd, unsigned long addr,
unsigned long end, struct page **pages, int *nr)
{
unsigned long fault_pfn;
fault_pfn = pmd_pfn(pmd) + ((addr & ~PMD_MASK) >> PAGE_SHIFT);
return __gup_device_huge(fault_pfn, addr, end, pages, nr);
}
static int __gup_device_huge_pud(pud_t pud, unsigned long addr,
unsigned long end, struct page **pages, int *nr)
{
unsigned long fault_pfn;
fault_pfn = pud_pfn(pud) + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
return __gup_device_huge(fault_pfn, addr, end, pages, nr);
}
static noinline int gup_huge_pmd(pmd_t pmd, unsigned long addr,
unsigned long end, int write, struct page **pages, int *nr)
{
struct page *head, *page;
int refs;
if (!pte_allows_gup(pmd_val(pmd), write))
return 0;
VM_BUG_ON(!pfn_valid(pmd_pfn(pmd)));
if (pmd_devmap(pmd))
return __gup_device_huge_pmd(pmd, addr, end, pages, nr);
/* hugepages are never "special" */
VM_BUG_ON(pmd_flags(pmd) & _PAGE_SPECIAL);
refs = 0;
head = pmd_page(pmd);
page = head + ((addr & ~PMD_MASK) >> PAGE_SHIFT);
do {
VM_BUG_ON_PAGE(compound_head(page) != head, page);
pages[*nr] = page;
(*nr)++;
page++;
refs++;
} while (addr += PAGE_SIZE, addr != end);
get_head_page_multiple(head, refs);
return 1;
}
static int gup_pmd_range(pud_t pud, unsigned long addr, unsigned long end,
int write, struct page **pages, int *nr)
{
unsigned long next;
pmd_t *pmdp;
pmdp = pmd_offset(&pud, addr);
do {
pmd_t pmd = *pmdp;
next = pmd_addr_end(addr, end);
if (pmd_none(pmd))
return 0;
if (unlikely(pmd_large(pmd) || !pmd_present(pmd))) {
/*
* NUMA hinting faults need to be handled in the GUP
* slowpath for accounting purposes and so that they
* can be serialised against THP migration.
*/
if (pmd_protnone(pmd))
return 0;
if (!gup_huge_pmd(pmd, addr, next, write, pages, nr))
return 0;
} else {
if (!gup_pte_range(pmd, addr, next, write, pages, nr))
return 0;
}
} while (pmdp++, addr = next, addr != end);
return 1;
}
static noinline int gup_huge_pud(pud_t pud, unsigned long addr,
unsigned long end, int write, struct page **pages, int *nr)
{
struct page *head, *page;
int refs;
if (!pte_allows_gup(pud_val(pud), write))
return 0;
VM_BUG_ON(!pfn_valid(pud_pfn(pud)));
if (pud_devmap(pud))
return __gup_device_huge_pud(pud, addr, end, pages, nr);
/* hugepages are never "special" */
VM_BUG_ON(pud_flags(pud) & _PAGE_SPECIAL);
refs = 0;
head = pud_page(pud);
page = head + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
do {
VM_BUG_ON_PAGE(compound_head(page) != head, page);
pages[*nr] = page;
(*nr)++;
page++;
refs++;
} while (addr += PAGE_SIZE, addr != end);
get_head_page_multiple(head, refs);
return 1;
}
static int gup_pud_range(p4d_t p4d, unsigned long addr, unsigned long end,
int write, struct page **pages, int *nr)
{
unsigned long next;
pud_t *pudp;
pudp = pud_offset(&p4d, addr);
do {
pud_t pud = *pudp;
next = pud_addr_end(addr, end);
if (pud_none(pud))
return 0;
if (unlikely(pud_large(pud))) {
if (!gup_huge_pud(pud, addr, next, write, pages, nr))
return 0;
} else {
if (!gup_pmd_range(pud, addr, next, write, pages, nr))
return 0;
}
} while (pudp++, addr = next, addr != end);
return 1;
}
static int gup_p4d_range(pgd_t pgd, unsigned long addr, unsigned long end,
int write, struct page **pages, int *nr)
{
unsigned long next;
p4d_t *p4dp;
p4dp = p4d_offset(&pgd, addr);
do {
p4d_t p4d = *p4dp;
next = p4d_addr_end(addr, end);
if (p4d_none(p4d))
return 0;
BUILD_BUG_ON(p4d_large(p4d));
if (!gup_pud_range(p4d, addr, next, write, pages, nr))
return 0;
} while (p4dp++, addr = next, addr != end);
return 1;
}
/*
* Like get_user_pages_fast() except its IRQ-safe in that it won't fall
* back to the regular GUP.
*/
int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
struct page **pages)
{
struct mm_struct *mm = current->mm;
unsigned long addr, len, end;
unsigned long next;
unsigned long flags;
pgd_t *pgdp;
int nr = 0;
start &= PAGE_MASK;
addr = start;
len = (unsigned long) nr_pages << PAGE_SHIFT;
end = start + len;
if (unlikely(!access_ok(write ? VERIFY_WRITE : VERIFY_READ,
(void __user *)start, len)))
return 0;
/*
* XXX: batch / limit 'nr', to avoid large irq off latency
* needs some instrumenting to determine the common sizes used by
* important workloads (eg. DB2), and whether limiting the batch size
* will decrease performance.
*
* It seems like we're in the clear for the moment. Direct-IO is
* the main guy that batches up lots of get_user_pages, and even
* they are limited to 64-at-a-time which is not so many.
*/
/*
* This doesn't prevent pagetable teardown, but does prevent
* the pagetables and pages from being freed on x86.
*
* So long as we atomically load page table pointers versus teardown
* (which we do on x86, with the above PAE exception), we can follow the
* address down to the the page and take a ref on it.
*/
local_irq_save(flags);
pgdp = pgd_offset(mm, addr);
do {
pgd_t pgd = *pgdp;
next = pgd_addr_end(addr, end);
if (pgd_none(pgd))
break;
if (!gup_p4d_range(pgd, addr, next, write, pages, &nr))
break;
} while (pgdp++, addr = next, addr != end);
local_irq_restore(flags);
return nr;
}
/**
* get_user_pages_fast() - pin user pages in memory
* @start: starting user address
* @nr_pages: number of pages from start to pin
* @write: whether pages will be written to
* @pages: array that receives pointers to the pages pinned.
* Should be at least nr_pages long.
*
* Attempt to pin user pages in memory without taking mm->mmap_sem.
* If not successful, it will fall back to taking the lock and
* calling get_user_pages().
*