* [patch 1/2] spufs: convert nopfn to fault
@ 2008-05-13 7:47 Nick Piggin
2008-05-13 7:48 ` [patch 2/2] mm: remove nopfn Nick Piggin
0 siblings, 1 reply; 5+ messages in thread
From: Nick Piggin @ 2008-05-13 7:47 UTC (permalink / raw)
To: Andrew Morton, Linux Memory Management List, Jeremy Kerr
This version fixes the 64K page size fault problem that Jeremy found in
testing.
--
spufs: convert nopfn to fault
From: Nick Piggin <npiggin@suse.de>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Jeremy Kerr <jk@ozlabs.org>
---
arch/powerpc/platforms/cell/spufs/file.c | 91 ++++++++++++---------------
arch/powerpc/platforms/cell/spufs/sputrace.c | 8 +-
2 files changed, 46 insertions(+), 53 deletions(-)
Index: linux-2.6/arch/powerpc/platforms/cell/spufs/file.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/file.c
+++ linux-2.6/arch/powerpc/platforms/cell/spufs/file.c
@@ -237,11 +237,13 @@ spufs_mem_write(struct file *file, const
return size;
}
-static unsigned long spufs_mem_mmap_nopfn(struct vm_area_struct *vma,
- unsigned long address)
+static int
+spufs_mem_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
struct spu_context *ctx = vma->vm_file->private_data;
- unsigned long pfn, offset, addr0 = address;
+ unsigned long address = (unsigned long)vmf->virtual_address;
+ unsigned long pfn, offset;
+
#ifdef CONFIG_SPU_FS_64K_LS
struct spu_state *csa = &ctx->csa;
int psize;
@@ -259,15 +261,15 @@ static unsigned long spufs_mem_mmap_nopf
}
#endif /* CONFIG_SPU_FS_64K_LS */
- offset = (address - vma->vm_start) + (vma->vm_pgoff << PAGE_SHIFT);
+ offset = vmf->pgoff << PAGE_SHIFT;
if (offset >= LS_SIZE)
- return NOPFN_SIGBUS;
+ return VM_FAULT_SIGBUS;
- pr_debug("spufs_mem_mmap_nopfn address=0x%lx -> 0x%lx, offset=0x%lx\n",
- addr0, address, offset);
+ pr_debug("spufs_mem_mmap_fault address=0x%lx, offset=0x%lx\n",
+ address, offset);
if (spu_acquire(ctx))
- return NOPFN_REFAULT;
+ return VM_FAULT_NOPAGE;
if (ctx->state == SPU_STATE_SAVED) {
vma->vm_page_prot = __pgprot(pgprot_val(vma->vm_page_prot)
@@ -282,12 +284,12 @@ static unsigned long spufs_mem_mmap_nopf
spu_release(ctx);
- return NOPFN_REFAULT;
+ return VM_FAULT_NOPAGE;
}
static struct vm_operations_struct spufs_mem_mmap_vmops = {
- .nopfn = spufs_mem_mmap_nopfn,
+ .fault = spufs_mem_mmap_fault,
};
static int spufs_mem_mmap(struct file *file, struct vm_area_struct *vma)
@@ -350,20 +352,19 @@ static const struct file_operations spuf
#endif
};
-static unsigned long spufs_ps_nopfn(struct vm_area_struct *vma,
- unsigned long address,
+static int spufs_ps_fault(struct vm_area_struct *vma,
+ struct vm_fault *vmf,
unsigned long ps_offs,
unsigned long ps_size)
{
struct spu_context *ctx = vma->vm_file->private_data;
- unsigned long area, offset = address - vma->vm_start;
+ unsigned long area, offset = vmf->pgoff << PAGE_SHIFT;
int ret = 0;
- spu_context_nospu_trace(spufs_ps_nopfn__enter, ctx);
+ spu_context_nospu_trace(spufs_ps_fault__enter, ctx);
- offset += vma->vm_pgoff << PAGE_SHIFT;
if (offset >= ps_size)
- return NOPFN_SIGBUS;
+ return VM_FAULT_SIGBUS;
/*
* Because we release the mmap_sem, the context may be destroyed while
@@ -377,7 +378,7 @@ static unsigned long spufs_ps_nopfn(stru
* pages to hand out to the user, but we don't want to wait
* with the mmap_sem held.
* It is possible to drop the mmap_sem here, but then we need
- * to return NOPFN_REFAULT because the mappings may have
+ * to return VM_FAULT_NOPAGE because the mappings may have
* hanged.
*/
if (spu_acquire(ctx))
@@ -385,14 +386,15 @@ static unsigned long spufs_ps_nopfn(stru
if (ctx->state == SPU_STATE_SAVED) {
up_read(¤t->mm->mmap_sem);
- spu_context_nospu_trace(spufs_ps_nopfn__sleep, ctx);
+ spu_context_nospu_trace(spufs_ps_fault__sleep, ctx);
ret = spufs_wait(ctx->run_wq, ctx->state == SPU_STATE_RUNNABLE);
- spu_context_trace(spufs_ps_nopfn__wake, ctx, ctx->spu);
+ spu_context_trace(spufs_ps_fault__wake, ctx, ctx->spu);
down_read(¤t->mm->mmap_sem);
} else {
area = ctx->spu->problem_phys + ps_offs;
- vm_insert_pfn(vma, address, (area + offset) >> PAGE_SHIFT);
- spu_context_trace(spufs_ps_nopfn__insert, ctx, ctx->spu);
+ vm_insert_pfn(vma, (unsigned long)vmf->virtual_address,
+ (area + offset) >> PAGE_SHIFT);
+ spu_context_trace(spufs_ps_fault__insert, ctx, ctx->spu);
}
if (!ret)
@@ -400,18 +402,18 @@ static unsigned long spufs_ps_nopfn(stru
refault:
put_spu_context(ctx);
- return NOPFN_REFAULT;
+ return VM_FAULT_NOPAGE;
}
#if SPUFS_MMAP_4K
-static unsigned long spufs_cntl_mmap_nopfn(struct vm_area_struct *vma,
- unsigned long address)
+static int spufs_cntl_mmap_fault(struct vm_area_struct *vma,
+ struct vm_fault *vmf)
{
- return spufs_ps_nopfn(vma, address, 0x4000, 0x1000);
+ return spufs_ps_fault(vma, vmf, 0x4000, 0x1000);
}
static struct vm_operations_struct spufs_cntl_mmap_vmops = {
- .nopfn = spufs_cntl_mmap_nopfn,
+ .fault = spufs_cntl_mmap_fault,
};
/*
@@ -1096,23 +1098,23 @@ static ssize_t spufs_signal1_write(struc
return 4;
}
-static unsigned long spufs_signal1_mmap_nopfn(struct vm_area_struct *vma,
- unsigned long address)
+static int
+spufs_signal1_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
#if PAGE_SIZE == 0x1000
- return spufs_ps_nopfn(vma, address, 0x14000, 0x1000);
+ return spufs_ps_fault(vma, vmf, 0x14000, 0x1000);
#elif PAGE_SIZE == 0x10000
/* For 64k pages, both signal1 and signal2 can be used to mmap the whole
* signal 1 and 2 area
*/
- return spufs_ps_nopfn(vma, address, 0x10000, 0x10000);
+ return spufs_ps_fault(vma, vmf, 0x10000, 0x10000);
#else
#error unsupported page size
#endif
}
static struct vm_operations_struct spufs_signal1_mmap_vmops = {
- .nopfn = spufs_signal1_mmap_nopfn,
+ .fault = spufs_signal1_mmap_fault,
};
static int spufs_signal1_mmap(struct file *file, struct vm_area_struct *vma)
@@ -1233,23 +1235,23 @@ static ssize_t spufs_signal2_write(struc
}
#if SPUFS_MMAP_4K
-static unsigned long spufs_signal2_mmap_nopfn(struct vm_area_struct *vma,
- unsigned long address)
+static int
+spufs_signal2_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
#if PAGE_SIZE == 0x1000
- return spufs_ps_nopfn(vma, address, 0x1c000, 0x1000);
+ return spufs_ps_fault(vma, vmf, 0x1c000, 0x1000);
#elif PAGE_SIZE == 0x10000
/* For 64k pages, both signal1 and signal2 can be used to mmap the whole
* signal 1 and 2 area
*/
- return spufs_ps_nopfn(vma, address, 0x10000, 0x10000);
+ return spufs_ps_fault(vma, vmf, 0x10000, 0x10000);
#else
#error unsupported page size
#endif
}
static struct vm_operations_struct spufs_signal2_mmap_vmops = {
- .nopfn = spufs_signal2_mmap_nopfn,
+ .fault = spufs_signal2_mmap_fault,
};
static int spufs_signal2_mmap(struct file *file, struct vm_area_struct *vma)
@@ -1361,14 +1363,14 @@ DEFINE_SPUFS_ATTRIBUTE(spufs_signal2_typ
spufs_signal2_type_set, "%llu\n", SPU_ATTR_ACQUIRE);
#if SPUFS_MMAP_4K
-static unsigned long spufs_mss_mmap_nopfn(struct vm_area_struct *vma,
- unsigned long address)
+static int
+spufs_mss_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
- return spufs_ps_nopfn(vma, address, 0x0000, 0x1000);
+ return spufs_ps_fault(vma, vmf, 0x0000, 0x1000);
}
static struct vm_operations_struct spufs_mss_mmap_vmops = {
- .nopfn = spufs_mss_mmap_nopfn,
+ .fault = spufs_mss_mmap_fault,
};
/*
@@ -1423,14 +1425,14 @@ static const struct file_operations spuf
.mmap = spufs_mss_mmap,
};
-static unsigned long spufs_psmap_mmap_nopfn(struct vm_area_struct *vma,
- unsigned long address)
+static int
+spufs_psmap_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
- return spufs_ps_nopfn(vma, address, 0x0000, 0x20000);
+ return spufs_ps_fault(vma, vmf, 0x0000, 0x20000);
}
static struct vm_operations_struct spufs_psmap_mmap_vmops = {
- .nopfn = spufs_psmap_mmap_nopfn,
+ .fault = spufs_psmap_mmap_fault,
};
/*
@@ -1483,14 +1485,14 @@ static const struct file_operations spuf
#if SPUFS_MMAP_4K
-static unsigned long spufs_mfc_mmap_nopfn(struct vm_area_struct *vma,
- unsigned long address)
+static int
+spufs_mfc_mmap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
- return spufs_ps_nopfn(vma, address, 0x3000, 0x1000);
+ return spufs_ps_fault(vma, vmf, 0x3000, 0x1000);
}
static struct vm_operations_struct spufs_mfc_mmap_vmops = {
- .nopfn = spufs_mfc_mmap_nopfn,
+ .fault = spufs_mfc_mmap_fault,
};
/*
Index: linux-2.6/arch/powerpc/platforms/cell/spufs/sputrace.c
===================================================================
--- linux-2.6.orig/arch/powerpc/platforms/cell/spufs/sputrace.c
+++ linux-2.6/arch/powerpc/platforms/cell/spufs/sputrace.c
@@ -182,10 +182,10 @@ struct spu_probe spu_probes[] = {
{ "spu_yield__enter", "ctx %p", spu_context_nospu_event },
{ "spu_deactivate__enter", "ctx %p", spu_context_nospu_event },
{ "__spu_deactivate__unload", "ctx %p spu %p", spu_context_event },
- { "spufs_ps_nopfn__enter", "ctx %p", spu_context_nospu_event },
- { "spufs_ps_nopfn__sleep", "ctx %p", spu_context_nospu_event },
- { "spufs_ps_nopfn__wake", "ctx %p spu %p", spu_context_event },
- { "spufs_ps_nopfn__insert", "ctx %p spu %p", spu_context_event },
+ { "spufs_ps_fault__enter", "ctx %p", spu_context_nospu_event },
+ { "spufs_ps_fault__sleep", "ctx %p", spu_context_nospu_event },
+ { "spufs_ps_fault__wake", "ctx %p spu %p", spu_context_event },
+ { "spufs_ps_fault__insert", "ctx %p spu %p", spu_context_event },
{ "spu_acquire_saved__enter", "ctx %p", spu_context_nospu_event },
{ "destroy_spu_context__enter", "ctx %p", spu_context_nospu_event },
{ "spufs_stop_callback__enter", "ctx %p spu %p", spu_context_event },
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply [flat|nested] 5+ messages in thread* [patch 2/2] mm: remove nopfn
2008-05-13 7:47 [patch 1/2] spufs: convert nopfn to fault Nick Piggin
@ 2008-05-13 7:48 ` Nick Piggin
2008-05-13 15:48 ` Jack Steiner
0 siblings, 1 reply; 5+ messages in thread
From: Nick Piggin @ 2008-05-13 7:48 UTC (permalink / raw)
To: Andrew Morton, Linux Memory Management List
There are no users of nopfn in the tree. Remove it.
Signed-off-by: Nick Piggin <npiggin@suse.de>
---
include/linux/mm.h | 9 -------
mm/memory.c | 61 ++++-------------------------------------------------
2 files changed, 5 insertions(+), 65 deletions(-)
Index: linux-2.6/include/linux/mm.h
===================================================================
--- linux-2.6.orig/include/linux/mm.h
+++ linux-2.6/include/linux/mm.h
@@ -165,8 +165,6 @@ struct vm_operations_struct {
void (*open)(struct vm_area_struct * area);
void (*close)(struct vm_area_struct * area);
int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
- unsigned long (*nopfn)(struct vm_area_struct *area,
- unsigned long address);
/* notification that a previously read-only page is about to become
* writable, if an error is returned it will cause a SIGBUS */
@@ -674,13 +672,6 @@ static inline int page_mapped(struct pag
}
/*
- * Error return values for the *_nopfn functions
- */
-#define NOPFN_SIGBUS ((unsigned long) -1)
-#define NOPFN_OOM ((unsigned long) -2)
-#define NOPFN_REFAULT ((unsigned long) -3)
-
-/*
* Different kinds of faults, as returned by handle_mm_fault().
* Used to decide whether a process gets delivered SIGBUS or
* just gets major/minor fault counters bumped up.
Index: linux-2.6/mm/memory.c
===================================================================
--- linux-2.6.orig/mm/memory.c
+++ linux-2.6/mm/memory.c
@@ -1290,6 +1290,11 @@ out:
*
* This function should only be called from a vm_ops->fault handler, and
* in that case the handler should return NULL.
+ *
+ * vma cannot be a COW mapping.
+ *
+ * As this is called only for pages that do not currently exist, we
+ * do not need to flush old virtual caches or the TLB.
*/
int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
unsigned long pfn)
@@ -2416,59 +2421,6 @@ static int do_linear_fault(struct mm_str
return __do_fault(mm, vma, address, pmd, pgoff, flags, orig_pte);
}
-
-/*
- * do_no_pfn() tries to create a new page mapping for a page without
- * a struct_page backing it
- *
- * As this is called only for pages that do not currently exist, we
- * do not need to flush old virtual caches or the TLB.
- *
- * We enter with non-exclusive mmap_sem (to exclude vma changes,
- * but allow concurrent faults), and pte mapped but not yet locked.
- * We return with mmap_sem still held, but pte unmapped and unlocked.
- *
- * It is expected that the ->nopfn handler always returns the same pfn
- * for a given virtual mapping.
- *
- * Mark this `noinline' to prevent it from bloating the main pagefault code.
- */
-static noinline int do_no_pfn(struct mm_struct *mm, struct vm_area_struct *vma,
- unsigned long address, pte_t *page_table, pmd_t *pmd,
- int write_access)
-{
- spinlock_t *ptl;
- pte_t entry;
- unsigned long pfn;
-
- pte_unmap(page_table);
- BUG_ON(!(vma->vm_flags & (VM_PFNMAP|VM_MIXEDMAP)));
- BUG_ON((vma->vm_flags & VM_PFNMAP) && is_cow_mapping(vma->vm_flags));
-
- pfn = vma->vm_ops->nopfn(vma, address & PAGE_MASK);
-
- BUG_ON((vma->vm_flags & VM_MIXEDMAP) && pfn_valid(pfn));
-
- if (unlikely(pfn == NOPFN_OOM))
- return VM_FAULT_OOM;
- else if (unlikely(pfn == NOPFN_SIGBUS))
- return VM_FAULT_SIGBUS;
- else if (unlikely(pfn == NOPFN_REFAULT))
- return 0;
-
- page_table = pte_offset_map_lock(mm, pmd, address, &ptl);
-
- /* Only go through if we didn't race with anybody else... */
- if (pte_none(*page_table)) {
- entry = pfn_pte(pfn, vma->vm_page_prot);
- if (write_access)
- entry = maybe_mkwrite(pte_mkdirty(entry), vma);
- set_pte_at(mm, address, page_table, entry);
- }
- pte_unmap_unlock(page_table, ptl);
- return 0;
-}
-
/*
* Fault of a previously existing named mapping. Repopulate the pte
* from the encoded file_pte if possible. This enables swappable
@@ -2529,9 +2481,6 @@ static inline int handle_pte_fault(struc
if (likely(vma->vm_ops->fault))
return do_linear_fault(mm, vma, address,
pte, pmd, write_access, entry);
- if (unlikely(vma->vm_ops->nopfn))
- return do_no_pfn(mm, vma, address, pte,
- pmd, write_access);
}
return do_anonymous_page(mm, vma, address,
pte, pmd, write_access);
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: [patch 2/2] mm: remove nopfn
2008-05-13 7:48 ` [patch 2/2] mm: remove nopfn Nick Piggin
@ 2008-05-13 15:48 ` Jack Steiner
[not found] ` <20080513162046.GA22407@sgi.com>
0 siblings, 1 reply; 5+ messages in thread
From: Jack Steiner @ 2008-05-13 15:48 UTC (permalink / raw)
To: Nick Piggin, Robin Holt; +Cc: Andrew Morton, Linux Memory Management List
On Tue, May 13, 2008 at 09:48:29AM +0200, Nick Piggin wrote:
> There are no users of nopfn in the tree. Remove it.
>
The SGI mspec driver use to use the nopfn callout. I see that this
was recently changed but the new code fails with:
kernel BUG at mm/memory.c:2278!
fop1[5887]: bugcheck! 0 [1]
Modules linked in:
Call Trace:
[<a000000100012740>] show_stack+0x40/0xa0
[<a000000100013050>] show_regs+0x850/0x8a0
[<a000000100036210>] die+0x1b0/0x2c0
[<a000000100036370>] die_if_kernel+0x50/0x80
[<a000000100037a50>] ia64_bad_break+0x230/0x460
[<a00000010000a2a0>] ia64_leave_kernel+0x0/0x270
[<a000000100141650>] __do_fault+0xb0/0xa20
[<a000000100145a50>] handle_mm_fault+0x2f0/0xf40
[<a000000100059160>] ia64_do_page_fault+0x220/0xa40
[<a00000010000a2a0>] ia64_leave_kernel+0x0/0x270
The mspec driver is tripping the bugcheck in __do_fault()
BUG_ON(vma->vm_flags & VM_PFNMAP);
The driver does not create pte entries at map time. Instead, it
relies on the nopfn (now fault) callout to assign resources
and create the ptes. It is intentionally done this way in order to
ensure that node-local resources are assigned.
What should the driver be doing to avoid this problem??
Also, the new GRU driver will have a similar problem. It currently
uses the nopfn callout since it needs to be able to assign resources
at fault, not mmap. The driver is not currently in-tree but will be
posted as soon as mmu_notifiers are available. I can post the current
version if it is helpful.....
--- jack
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2008-05-14 1:22 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-05-13 7:47 [patch 1/2] spufs: convert nopfn to fault Nick Piggin
2008-05-13 7:48 ` [patch 2/2] mm: remove nopfn Nick Piggin
2008-05-13 15:48 ` Jack Steiner
[not found] ` <20080513162046.GA22407@sgi.com>
2008-05-14 0:44 ` Nick Piggin
2008-05-14 1:22 ` Jack Steiner
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).