* [PATCH 00/10] Alter NOMMU mmap handling [ver #3]
@ 2009-01-08 12:53 David Howells
2009-01-08 12:53 ` [PATCH 01/10] NOMMU: Fix cleanup handling in ramfs_nommu_get_umapped_area() " David Howells
` (9 more replies)
0 siblings, 10 replies; 13+ messages in thread
From: David Howells @ 2009-01-08 12:53 UTC (permalink / raw)
To: torvalds, akpm
Cc: bernds_cb1, dhowells, rgetz, vapier.adi, lethal, linux-kernel
The primary change made by the attached patches alters the NOMMU mmap code so
that VMAs are per-MM rather than being a shared resource with separate per-MM
linkage blocks.
This solves two problems:
(1) In SYSV SHM where nattch for a segment does not reflect the number of
shmat's (and forks) done.
(2) In mmap() where the VMA's vm_mm is set to point to the parent mm by an
exec'ing process when VM_EXECUTABLE is specified, regardless of the fact
that a VMA might be shared and already have its vm_mm assigned to another
process or even a dead process.
And adds a number of features:
(3) Allocations are now made with alloc_pages() rather than kmalloc(). This
simplifies page accounting within a block of pages.
(4) Excess pages can be trimmed from allocations.
(5) VMAs are added for the parent MM's RB-tree and mmap lists, as for MMU-mode
mmap.
(6) Non-anonymous VMAs can now be linked to the inode as for MMU-mode.
(7) mmaps can now be partially unmapped under some circumstances.
(8) Simplified coredumping in ELF-FDPIC.
(9) /proc/meminfo has a line "MmapCopy" that indicates the amount of private
copying mmap() has done.
There are some patches with secondary changes that do not relate to the primary
change:
(*) Ramfs needs to clean up correctly in the case where its get_unmapped_area
call gets fewer pages from find_get_pages() than were asked for.
(*) The askedalloc and realalloc variables are removed as nothing uses the
values computed.
(*) The ELF-FDPIC and FLAT binfmts no longer expand the stack segment to
consume the space that is allocated to them beyond what is asked for.
This removes some users of the contentious kobjsize().
An alternative method would be to to replace the call to kobjsize() with
something that uses the VMA list only.
(*) Fix kobjsize() to check the VMA list.
(*) Support XIP on initramfs under NOMMU conditions by pre-truncating the
ramfs files to get contiguous memory allocated in advance in the page
cache.
The patches can be obtained as a tarball from:
http://people.redhat.com/~dhowells/nommu-permmvmas.tar.bz2
And are available by GIT from:
git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-2.6-nommu.git
These patches have been tested on FRV, Blackfin and SH.
David
^ permalink raw reply [flat|nested] 13+ messages in thread
* [PATCH 01/10] NOMMU: Fix cleanup handling in ramfs_nommu_get_umapped_area() [ver #3]
2009-01-08 12:53 [PATCH 00/10] Alter NOMMU mmap handling [ver #3] David Howells
@ 2009-01-08 12:53 ` David Howells
2009-01-08 12:53 ` [PATCH 02/10] NOMMU: Rename ARM's struct vm_region " David Howells
` (8 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: David Howells @ 2009-01-08 12:53 UTC (permalink / raw)
To: torvalds, akpm
Cc: bernds_cb1, dhowells, rgetz, vapier.adi, lethal, linux-kernel
Fix cleanup handling in ramfs_nommu_get_umapped_area() by only freeing the
number of pages that find_get_pages() said it had returned (nr) rather than
attempting to free the number of pages we asked for (lpages) - thus avoiding
the situation whereby put_page() may be handed NULL pointers if
find_get_pages() returned fewer pages that were requested.
Also avoid a warning about nr being uninitialised and the need for an
if-statement in the cleanup path by using appropriate gotos.
Signed-off-by: David Howells <dhowells@redhat.com>
---
fs/ramfs/file-nommu.c | 21 ++++++++++-----------
1 files changed, 10 insertions(+), 11 deletions(-)
diff --git a/fs/ramfs/file-nommu.c b/fs/ramfs/file-nommu.c
index 76acdbc..b9b567a 100644
--- a/fs/ramfs/file-nommu.c
+++ b/fs/ramfs/file-nommu.c
@@ -262,11 +262,11 @@ unsigned long ramfs_nommu_get_unmapped_area(struct file *file,
ret = -ENOMEM;
pages = kzalloc(lpages * sizeof(struct page *), GFP_KERNEL);
if (!pages)
- goto out;
+ goto out_free;
nr = find_get_pages(inode->i_mapping, pgoff, lpages, pages);
if (nr != lpages)
- goto out; /* leave if some pages were missing */
+ goto out_free_pages; /* leave if some pages were missing */
/* check the pages for physical adjacency */
ptr = pages;
@@ -274,19 +274,18 @@ unsigned long ramfs_nommu_get_unmapped_area(struct file *file,
page++;
for (loop = lpages; loop > 1; loop--)
if (*ptr++ != page++)
- goto out;
+ goto out_free_pages;
/* okay - all conditions fulfilled */
ret = (unsigned long) page_address(pages[0]);
- out:
- if (pages) {
- ptr = pages;
- for (loop = lpages; loop > 0; loop--)
- put_page(*ptr++);
- kfree(pages);
- }
-
+out_free_pages:
+ ptr = pages;
+ for (loop = nr; loop > 0; loop--)
+ put_page(*ptr++);
+out_free:
+ kfree(pages);
+out:
return ret;
}
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [PATCH 02/10] NOMMU: Rename ARM's struct vm_region [ver #3]
2009-01-08 12:53 [PATCH 00/10] Alter NOMMU mmap handling [ver #3] David Howells
2009-01-08 12:53 ` [PATCH 01/10] NOMMU: Fix cleanup handling in ramfs_nommu_get_umapped_area() " David Howells
@ 2009-01-08 12:53 ` David Howells
2009-01-08 12:54 ` [PATCH 03/10] NOMMU: Delete askedalloc and realalloc variables " David Howells
` (7 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: David Howells @ 2009-01-08 12:53 UTC (permalink / raw)
To: torvalds, akpm
Cc: bernds_cb1, dhowells, rgetz, vapier.adi, lethal, linux-kernel
Rename ARM's struct vm_region so that I can introduce my own global version
for NOMMU. It's feasible that the ARM version may wish to use my global one
instead.
The NOMMU vm_region struct defines areas of the physical memory map that are
under mmap. This may include chunks of RAM or regions of memory mapped
devices, such as flash. It is also used to retain copies of file content so
that shareable private memory mappings of files can be made. As such, it may
be compatible with what is described in the banner comment for ARM's vm_region
struct.
Signed-off-by: David Howells <dhowells@redhat.com>
---
arch/arm/mm/dma-mapping.c | 28 ++++++++++++++--------------
1 files changed, 14 insertions(+), 14 deletions(-)
diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
index 6796001..310e479 100644
--- a/arch/arm/mm/dma-mapping.c
+++ b/arch/arm/mm/dma-mapping.c
@@ -71,7 +71,7 @@ static DEFINE_SPINLOCK(consistent_lock);
* the amount of RAM found at boot time.) I would imagine that get_vm_area()
* would have to initialise this each time prior to calling vm_region_alloc().
*/
-struct vm_region {
+struct arm_vm_region {
struct list_head vm_list;
unsigned long vm_start;
unsigned long vm_end;
@@ -79,20 +79,20 @@ struct vm_region {
int vm_active;
};
-static struct vm_region consistent_head = {
+static struct arm_vm_region consistent_head = {
.vm_list = LIST_HEAD_INIT(consistent_head.vm_list),
.vm_start = CONSISTENT_BASE,
.vm_end = CONSISTENT_END,
};
-static struct vm_region *
-vm_region_alloc(struct vm_region *head, size_t size, gfp_t gfp)
+static struct arm_vm_region *
+arm_vm_region_alloc(struct arm_vm_region *head, size_t size, gfp_t gfp)
{
unsigned long addr = head->vm_start, end = head->vm_end - size;
unsigned long flags;
- struct vm_region *c, *new;
+ struct arm_vm_region *c, *new;
- new = kmalloc(sizeof(struct vm_region), gfp);
+ new = kmalloc(sizeof(struct arm_vm_region), gfp);
if (!new)
goto out;
@@ -127,9 +127,9 @@ vm_region_alloc(struct vm_region *head, size_t size, gfp_t gfp)
return NULL;
}
-static struct vm_region *vm_region_find(struct vm_region *head, unsigned long addr)
+static struct arm_vm_region *arm_vm_region_find(struct arm_vm_region *head, unsigned long addr)
{
- struct vm_region *c;
+ struct arm_vm_region *c;
list_for_each_entry(c, &head->vm_list, vm_list) {
if (c->vm_active && c->vm_start == addr)
@@ -149,7 +149,7 @@ __dma_alloc(struct device *dev, size_t size, dma_addr_t *handle, gfp_t gfp,
pgprot_t prot)
{
struct page *page;
- struct vm_region *c;
+ struct arm_vm_region *c;
unsigned long order;
u64 mask = ISA_DMA_THRESHOLD, limit;
@@ -214,7 +214,7 @@ __dma_alloc(struct device *dev, size_t size, dma_addr_t *handle, gfp_t gfp,
/*
* Allocate a virtual address in the consistent mapping region.
*/
- c = vm_region_alloc(&consistent_head, size,
+ c = arm_vm_region_alloc(&consistent_head, size,
gfp & ~(__GFP_DMA | __GFP_HIGHMEM));
if (c) {
pte_t *pte;
@@ -311,13 +311,13 @@ static int dma_mmap(struct device *dev, struct vm_area_struct *vma,
void *cpu_addr, dma_addr_t dma_addr, size_t size)
{
unsigned long flags, user_size, kern_size;
- struct vm_region *c;
+ struct arm_vm_region *c;
int ret = -ENXIO;
user_size = (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
spin_lock_irqsave(&consistent_lock, flags);
- c = vm_region_find(&consistent_head, (unsigned long)cpu_addr);
+ c = arm_vm_region_find(&consistent_head, (unsigned long)cpu_addr);
spin_unlock_irqrestore(&consistent_lock, flags);
if (c) {
@@ -359,7 +359,7 @@ EXPORT_SYMBOL(dma_mmap_writecombine);
*/
void dma_free_coherent(struct device *dev, size_t size, void *cpu_addr, dma_addr_t handle)
{
- struct vm_region *c;
+ struct arm_vm_region *c;
unsigned long flags, addr;
pte_t *ptep;
int idx;
@@ -378,7 +378,7 @@ void dma_free_coherent(struct device *dev, size_t size, void *cpu_addr, dma_addr
size = PAGE_ALIGN(size);
spin_lock_irqsave(&consistent_lock, flags);
- c = vm_region_find(&consistent_head, (unsigned long)cpu_addr);
+ c = arm_vm_region_find(&consistent_head, (unsigned long)cpu_addr);
if (!c)
goto no_area;
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [PATCH 03/10] NOMMU: Delete askedalloc and realalloc variables [ver #3]
2009-01-08 12:53 [PATCH 00/10] Alter NOMMU mmap handling [ver #3] David Howells
2009-01-08 12:53 ` [PATCH 01/10] NOMMU: Fix cleanup handling in ramfs_nommu_get_umapped_area() " David Howells
2009-01-08 12:53 ` [PATCH 02/10] NOMMU: Rename ARM's struct vm_region " David Howells
@ 2009-01-08 12:54 ` David Howells
2009-01-08 12:54 ` [PATCH 04/10] NOMMU: Make VMAs per MM as for MMU-mode linux " David Howells
` (6 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: David Howells @ 2009-01-08 12:54 UTC (permalink / raw)
To: torvalds, akpm
Cc: bernds_cb1, dhowells, rgetz, vapier.adi, lethal, linux-kernel
Delete the askedalloc and realalloc variables as nothing actually uses the
value calculated.
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Mike Frysinger <vapier.adi@gmail.com>
Acked-by: Paul Mundt <lethal@linux-sh.org>
---
mm/nommu.c | 28 +---------------------------
1 files changed, 1 insertions(+), 27 deletions(-)
diff --git a/mm/nommu.c b/mm/nommu.c
index 1c28ea3..23f355b 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -40,7 +40,6 @@ void *high_memory;
struct page *mem_map;
unsigned long max_mapnr;
unsigned long num_physpages;
-unsigned long askedalloc, realalloc;
atomic_long_t vm_committed_space = ATOMIC_LONG_INIT(0);
int sysctl_overcommit_memory = OVERCOMMIT_GUESS; /* heuristic overcommit */
int sysctl_overcommit_ratio = 50; /* default is 50% */
@@ -1042,22 +1041,11 @@ unsigned long do_mmap_pgoff(struct file *file,
/* okay... we have a mapping; now we have to register it */
result = (void *) vma->vm_start;
- if (vma->vm_flags & VM_MAPPED_COPY) {
- realalloc += kobjsize(result);
- askedalloc += len;
- }
-
- realalloc += kobjsize(vma);
- askedalloc += sizeof(*vma);
-
current->mm->total_vm += len >> PAGE_SHIFT;
add_nommu_vma(vma);
shared:
- realalloc += kobjsize(vml);
- askedalloc += sizeof(*vml);
-
add_vma_to_mm(current->mm, vml);
up_write(&nommu_vma_sem);
@@ -1124,14 +1112,8 @@ static void put_vma(struct mm_struct *mm, struct vm_area_struct *vma)
/* IO memory and memory shared directly out of the pagecache from
* ramfs/tmpfs mustn't be released here */
- if (vma->vm_flags & VM_MAPPED_COPY) {
- realalloc -= kobjsize((void *) vma->vm_start);
- askedalloc -= vma->vm_end - vma->vm_start;
+ if (vma->vm_flags & VM_MAPPED_COPY)
kfree((void *) vma->vm_start);
- }
-
- realalloc -= kobjsize(vma);
- askedalloc -= sizeof(*vma);
if (vma->vm_file) {
fput(vma->vm_file);
@@ -1177,8 +1159,6 @@ int do_munmap(struct mm_struct *mm, unsigned long addr, size_t len)
put_vma(mm, vml->vma);
*parent = vml->next;
- realalloc -= kobjsize(vml);
- askedalloc -= sizeof(*vml);
kfree(vml);
update_hiwater_vm(mm);
@@ -1220,9 +1200,6 @@ void exit_mmap(struct mm_struct * mm)
while ((tmp = mm->context.vmlist)) {
mm->context.vmlist = tmp->next;
put_vma(mm, tmp->vma);
-
- realalloc -= kobjsize(tmp);
- askedalloc -= sizeof(*tmp);
kfree(tmp);
}
@@ -1276,9 +1253,6 @@ unsigned long do_mremap(unsigned long addr,
/* all checks complete - do it */
vma->vm_end = vma->vm_start + new_len;
- askedalloc -= old_len;
- askedalloc += new_len;
-
return vma->vm_start;
}
EXPORT_SYMBOL(do_mremap);
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [PATCH 04/10] NOMMU: Make VMAs per MM as for MMU-mode linux [ver #3]
2009-01-08 12:53 [PATCH 00/10] Alter NOMMU mmap handling [ver #3] David Howells
` (2 preceding siblings ...)
2009-01-08 12:54 ` [PATCH 03/10] NOMMU: Delete askedalloc and realalloc variables " David Howells
@ 2009-01-08 12:54 ` David Howells
2009-01-13 7:55 ` Andrew Morton
2009-01-08 12:54 ` [PATCH 05/10] NOMMU: Make mmap allocation page trimming behaviour configurable. " David Howells
` (5 subsequent siblings)
9 siblings, 1 reply; 13+ messages in thread
From: David Howells @ 2009-01-08 12:54 UTC (permalink / raw)
To: torvalds, akpm
Cc: bernds_cb1, dhowells, rgetz, vapier.adi, lethal, linux-kernel
Make VMAs per mm_struct as for MMU-mode linux. This solves two problems:
(1) In SYSV SHM where nattch for a segment does not reflect the number of
shmat's (and forks) done.
(2) In mmap() where the VMA's vm_mm is set to point to the parent mm by an
exec'ing process when VM_EXECUTABLE is specified, regardless of the fact
that a VMA might be shared and already have its vm_mm assigned to another
process or a dead process.
A new struct (vm_region) is introduced to track a mapped region and to remember
the circumstances under which it may be shared and the vm_list_struct structure
is discarded as it's no longer required.
This patch makes the following additional changes:
(1) Regions are now allocated with alloc_pages() rather than kmalloc() and
with no recourse to __GFP_COMP, so the pages are not composite. Instead,
each page has a reference on it held by the region. Anything else that is
interested in such a page will have to get a reference on it to retain it.
When the pages are released due to unmapping, each page is passed to
put_page() and will be freed when the page usage count reaches zero.
(2) Excess pages are trimmed after an allocation as the allocation must be
made as a power-of-2 quantity of pages.
(3) VMAs are added to the parent MM's R/B tree and mmap lists. As an MM may
end up with overlapping VMAs within the tree, the VMA struct address is
appended to the sort key.
(4) Non-anonymous VMAs are now added to the backing inode's prio list.
(5) Holes may be punched in anonymous VMAs with munmap(), releasing parts of
the backing region. The VMA and region structs will be split if
necessary.
(6) sys_shmdt() only releases one attachment to a SYSV IPC shared memory
segment instead of all the attachments at that addresss. Multiple
shmat()'s return the same address under NOMMU-mode instead of different
virtual addresses as under MMU-mode.
(7) Core dumping for ELF-FDPIC requires fewer exceptions for NOMMU-mode.
(8) /proc/maps is now the global list of mapped regions, and may list bits
that aren't actually mapped anywhere.
(9) /proc/meminfo gains a line (tagged "MmapCopy") that indicates the amount
of RAM currently allocated by mmap to hold mappable regions that can't be
mapped directly. These are copies of the backing device or file if not
anonymous.
These changes make NOMMU mode more similar to MMU mode. The downside is that
NOMMU mode requires some extra memory to track things over NOMMU without this
patch (VMAs are no longer shared, and there are now region structs).
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Mike Frysinger <vapier.adi@gmail.com>
Acked-by: Paul Mundt <lethal@linux-sh.org>
---
Documentation/nommu-mmap.txt | 18 -
arch/arm/include/asm/mmu.h | 1
arch/blackfin/include/asm/mmu.h | 1
arch/blackfin/kernel/ptrace.c | 6
arch/blackfin/kernel/traps.c | 11
arch/frv/kernel/ptrace.c | 11
arch/h8300/include/asm/mmu.h | 1
arch/m68knommu/include/asm/mmu.h | 1
arch/sh/include/asm/mmu.h | 1
fs/binfmt_elf_fdpic.c | 27 -
fs/proc/internal.h | 2
fs/proc/meminfo.c | 6
fs/proc/nommu.c | 71 +--
fs/proc/task_nommu.c | 108 +++-
include/asm-frv/mmu.h | 1
include/asm-m32r/mmu.h | 1
include/linux/mm.h | 18 -
include/linux/mm_types.h | 18 +
ipc/shm.c | 12
kernel/fork.c | 4
lib/Kconfig.debug | 7
mm/mmap.c | 10
mm/nommu.c | 960 ++++++++++++++++++++++++++------------
23 files changed, 860 insertions(+), 436 deletions(-)
diff --git a/Documentation/nommu-mmap.txt b/Documentation/nommu-mmap.txt
index 7714f57..02b89dc 100644
--- a/Documentation/nommu-mmap.txt
+++ b/Documentation/nommu-mmap.txt
@@ -109,12 +109,18 @@ and it's also much more restricted in the latter case:
FURTHER NOTES ON NO-MMU MMAP
============================
- (*) A request for a private mapping of less than a page in size may not return
- a page-aligned buffer. This is because the kernel calls kmalloc() to
- allocate the buffer, not get_free_page().
-
- (*) A list of all the mappings on the system is visible through /proc/maps in
- no-MMU mode.
+ (*) A request for a private mapping of a file may return a buffer that is not
+ page-aligned. This is because XIP may take place, and the data may not be
+ paged aligned in the backing store.
+
+ (*) A request for an anonymous mapping will always be page aligned. If
+ possible the size of the request should be a power of two otherwise some
+ of the space may be wasted as the kernel must allocate a power-of-2
+ granule but will only discard the excess if appropriately configured as
+ this has an effect on fragmentation.
+
+ (*) A list of all the private copy and anonymous mappings on the system is
+ visible through /proc/maps in no-MMU mode.
(*) A list of all the mappings in use by a process is visible through
/proc/<pid>/maps in no-MMU mode.
diff --git a/arch/arm/include/asm/mmu.h b/arch/arm/include/asm/mmu.h
index 53099d4..b561584 100644
--- a/arch/arm/include/asm/mmu.h
+++ b/arch/arm/include/asm/mmu.h
@@ -24,7 +24,6 @@ typedef struct {
* modified for 2.6 by Hyok S. Choi <hyok.choi@samsung.com>
*/
typedef struct {
- struct vm_list_struct *vmlist;
unsigned long end_brk;
} mm_context_t;
diff --git a/arch/blackfin/include/asm/mmu.h b/arch/blackfin/include/asm/mmu.h
index 757e439..dbfd686 100644
--- a/arch/blackfin/include/asm/mmu.h
+++ b/arch/blackfin/include/asm/mmu.h
@@ -10,7 +10,6 @@ struct sram_list_struct {
};
typedef struct {
- struct vm_list_struct *vmlist;
unsigned long end_brk;
unsigned long stack_start;
diff --git a/arch/blackfin/kernel/ptrace.c b/arch/blackfin/kernel/ptrace.c
index d2d3885..594e325 100644
--- a/arch/blackfin/kernel/ptrace.c
+++ b/arch/blackfin/kernel/ptrace.c
@@ -160,15 +160,15 @@ put_reg(struct task_struct *task, int regno, unsigned long data)
static inline int is_user_addr_valid(struct task_struct *child,
unsigned long start, unsigned long len)
{
- struct vm_list_struct *vml;
+ struct vm_area_struct *vma;
struct sram_list_struct *sraml;
/* overflow */
if (start + len < start)
return -EIO;
- for (vml = child->mm->context.vmlist; vml; vml = vml->next)
- if (start >= vml->vma->vm_start && start + len < vml->vma->vm_end)
+ vma = find_vma(child->mm, start);
+ if (vma && start >= vma->vm_start && start + len <= vma->vm_end)
return 0;
for (sraml = child->mm->context.sram_list; sraml; sraml = sraml->next)
diff --git a/arch/blackfin/kernel/traps.c b/arch/blackfin/kernel/traps.c
index 17d8e41..5b0667d 100644
--- a/arch/blackfin/kernel/traps.c
+++ b/arch/blackfin/kernel/traps.c
@@ -32,6 +32,7 @@
#include <linux/module.h>
#include <linux/kallsyms.h>
#include <linux/fs.h>
+#include <linux/rbtree.h>
#include <asm/traps.h>
#include <asm/cacheflush.h>
#include <asm/cplb.h>
@@ -83,6 +84,7 @@ static void decode_address(char *buf, unsigned long address)
struct mm_struct *mm;
unsigned long flags, offset;
unsigned char in_atomic = (bfin_read_IPEND() & 0x10) || in_atomic();
+ struct rb_node *n;
#ifdef CONFIG_KALLSYMS
unsigned long symsize;
@@ -128,9 +130,10 @@ static void decode_address(char *buf, unsigned long address)
if (!mm)
continue;
- vml = mm->context.vmlist;
- while (vml) {
- struct vm_area_struct *vma = vml->vma;
+ for (n = rb_first(&mm->mm_rb); n; n = rb_next(n)) {
+ struct vm_area_struct *vma;
+
+ vma = rb_entry(n, struct vm_area_struct, vm_rb);
if (address >= vma->vm_start && address < vma->vm_end) {
char _tmpbuf[256];
@@ -176,8 +179,6 @@ static void decode_address(char *buf, unsigned long address)
goto done;
}
-
- vml = vml->next;
}
if (!in_atomic)
mmput(mm);
diff --git a/arch/frv/kernel/ptrace.c b/arch/frv/kernel/ptrace.c
index 709e9bd..5e7d401 100644
--- a/arch/frv/kernel/ptrace.c
+++ b/arch/frv/kernel/ptrace.c
@@ -69,7 +69,8 @@ static inline int put_reg(struct task_struct *task, int regno,
}
/*
- * check that an address falls within the bounds of the target process's memory mappings
+ * check that an address falls within the bounds of the target process's memory
+ * mappings
*/
static inline int is_user_addr_valid(struct task_struct *child,
unsigned long start, unsigned long len)
@@ -79,11 +80,11 @@ static inline int is_user_addr_valid(struct task_struct *child,
return -EIO;
return 0;
#else
- struct vm_list_struct *vml;
+ struct vm_area_struct *vma;
- for (vml = child->mm->context.vmlist; vml; vml = vml->next)
- if (start >= vml->vma->vm_start && start + len <= vml->vma->vm_end)
- return 0;
+ vma = find_vma(child->mm, start);
+ if (vma && start >= vma->vm_start && start + len <= vma->vm_end)
+ return 0;
return -EIO;
#endif
diff --git a/arch/h8300/include/asm/mmu.h b/arch/h8300/include/asm/mmu.h
index 2ce06ea..3130996 100644
--- a/arch/h8300/include/asm/mmu.h
+++ b/arch/h8300/include/asm/mmu.h
@@ -4,7 +4,6 @@
/* Copyright (C) 2002, David McCullough <davidm@snapgear.com> */
typedef struct {
- struct vm_list_struct *vmlist;
unsigned long end_brk;
} mm_context_t;
diff --git a/arch/m68knommu/include/asm/mmu.h b/arch/m68knommu/include/asm/mmu.h
index 5fa6b68..e2da1e6 100644
--- a/arch/m68knommu/include/asm/mmu.h
+++ b/arch/m68knommu/include/asm/mmu.h
@@ -4,7 +4,6 @@
/* Copyright (C) 2002, David McCullough <davidm@snapgear.com> */
typedef struct {
- struct vm_list_struct *vmlist;
unsigned long end_brk;
} mm_context_t;
diff --git a/arch/sh/include/asm/mmu.h b/arch/sh/include/asm/mmu.h
index fdcb93b..6c43625 100644
--- a/arch/sh/include/asm/mmu.h
+++ b/arch/sh/include/asm/mmu.h
@@ -9,7 +9,6 @@ typedef struct {
mm_context_id_t id;
void *vdso;
#else
- struct vm_list_struct *vmlist;
unsigned long end_brk;
#endif
#ifdef CONFIG_BINFMT_ELF_FDPIC
diff --git a/fs/binfmt_elf_fdpic.c b/fs/binfmt_elf_fdpic.c
index aa5b432..22baf1b 100644
--- a/fs/binfmt_elf_fdpic.c
+++ b/fs/binfmt_elf_fdpic.c
@@ -1567,11 +1567,9 @@ end_coredump:
static int elf_fdpic_dump_segments(struct file *file, size_t *size,
unsigned long *limit, unsigned long mm_flags)
{
- struct vm_list_struct *vml;
-
- for (vml = current->mm->context.vmlist; vml; vml = vml->next) {
- struct vm_area_struct *vma = vml->vma;
+ struct vm_area_struct *vma;
+ for (vma = current->mm->mmap; vma; vma = vma->vm_next) {
if (!maydump(vma, mm_flags))
continue;
@@ -1617,9 +1615,6 @@ static int elf_fdpic_core_dump(long signr, struct pt_regs *regs,
elf_fpxregset_t *xfpu = NULL;
#endif
int thread_status_size = 0;
-#ifndef CONFIG_MMU
- struct vm_list_struct *vml;
-#endif
elf_addr_t *auxv;
unsigned long mm_flags;
@@ -1685,13 +1680,7 @@ static int elf_fdpic_core_dump(long signr, struct pt_regs *regs,
fill_prstatus(prstatus, current, signr);
elf_core_copy_regs(&prstatus->pr_reg, regs);
-#ifdef CONFIG_MMU
segs = current->mm->map_count;
-#else
- segs = 0;
- for (vml = current->mm->context.vmlist; vml; vml = vml->next)
- segs++;
-#endif
#ifdef ELF_CORE_EXTRA_PHDRS
segs += ELF_CORE_EXTRA_PHDRS;
#endif
@@ -1766,20 +1755,10 @@ static int elf_fdpic_core_dump(long signr, struct pt_regs *regs,
mm_flags = current->mm->flags;
/* write program headers for segments dump */
- for (
-#ifdef CONFIG_MMU
- vma = current->mm->mmap; vma; vma = vma->vm_next
-#else
- vml = current->mm->context.vmlist; vml; vml = vml->next
-#endif
- ) {
+ for (vma = current->mm->mmap; vma; vma = vma->vm_next) {
struct elf_phdr phdr;
size_t sz;
-#ifndef CONFIG_MMU
- vma = vml->vma;
-#endif
-
sz = vma->vm_end - vma->vm_start;
phdr.p_type = PT_LOAD;
diff --git a/fs/proc/internal.h b/fs/proc/internal.h
index 3e8aeb8..cd53ff8 100644
--- a/fs/proc/internal.h
+++ b/fs/proc/internal.h
@@ -41,8 +41,6 @@ do { \
(vmi)->used = 0; \
(vmi)->largest_chunk = 0; \
} while(0)
-
-extern int nommu_vma_show(struct seq_file *, struct vm_area_struct *);
#endif
extern int proc_tid_stat(struct seq_file *m, struct pid_namespace *ns,
diff --git a/fs/proc/meminfo.c b/fs/proc/meminfo.c
index b1675c4..43d2394 100644
--- a/fs/proc/meminfo.c
+++ b/fs/proc/meminfo.c
@@ -74,6 +74,9 @@ static int meminfo_proc_show(struct seq_file *m, void *v)
"LowTotal: %8lu kB\n"
"LowFree: %8lu kB\n"
#endif
+#ifndef CONFIG_MMU
+ "MmapCopy: %8lu kB\n"
+#endif
"SwapTotal: %8lu kB\n"
"SwapFree: %8lu kB\n"
"Dirty: %8lu kB\n"
@@ -116,6 +119,9 @@ static int meminfo_proc_show(struct seq_file *m, void *v)
K(i.totalram-i.totalhigh),
K(i.freeram-i.freehigh),
#endif
+#ifndef CONFIG_MMU
+ K((unsigned long) atomic_read(&mmap_pages_allocated)),
+#endif
K(i.totalswap),
K(i.freeswap),
K(global_page_state(NR_FILE_DIRTY)),
diff --git a/fs/proc/nommu.c b/fs/proc/nommu.c
index 3f87d26..b446d7a 100644
--- a/fs/proc/nommu.c
+++ b/fs/proc/nommu.c
@@ -33,33 +33,33 @@
#include "internal.h"
/*
- * display a single VMA to a sequenced file
+ * display a single region to a sequenced file
*/
-int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma)
+static int nommu_region_show(struct seq_file *m, struct vm_region *region)
{
unsigned long ino = 0;
struct file *file;
dev_t dev = 0;
int flags, len;
- flags = vma->vm_flags;
- file = vma->vm_file;
+ flags = region->vm_flags;
+ file = region->vm_file;
if (file) {
- struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
+ struct inode *inode = region->vm_file->f_path.dentry->d_inode;
dev = inode->i_sb->s_dev;
ino = inode->i_ino;
}
seq_printf(m,
"%08lx-%08lx %c%c%c%c %08llx %02x:%02x %lu %n",
- vma->vm_start,
- vma->vm_end,
+ region->vm_start,
+ region->vm_end,
flags & VM_READ ? 'r' : '-',
flags & VM_WRITE ? 'w' : '-',
flags & VM_EXEC ? 'x' : '-',
flags & VM_MAYSHARE ? flags & VM_SHARED ? 'S' : 's' : 'p',
- ((loff_t)vma->vm_pgoff) << PAGE_SHIFT,
+ ((loff_t)region->vm_pgoff) << PAGE_SHIFT,
MAJOR(dev), MINOR(dev), ino, &len);
if (file) {
@@ -75,61 +75,54 @@ int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma)
}
/*
- * display a list of all the VMAs the kernel knows about
+ * display a list of all the REGIONs the kernel knows about
* - nommu kernals have a single flat list
*/
-static int nommu_vma_list_show(struct seq_file *m, void *v)
+static int nommu_region_list_show(struct seq_file *m, void *_p)
{
- struct vm_area_struct *vma;
+ struct rb_node *p = _p;
- vma = rb_entry((struct rb_node *) v, struct vm_area_struct, vm_rb);
- return nommu_vma_show(m, vma);
+ return nommu_region_show(m, rb_entry(p, struct vm_region, vm_rb));
}
-static void *nommu_vma_list_start(struct seq_file *m, loff_t *_pos)
+static void *nommu_region_list_start(struct seq_file *m, loff_t *_pos)
{
- struct rb_node *_rb;
+ struct rb_node *p;
loff_t pos = *_pos;
- void *next = NULL;
- down_read(&nommu_vma_sem);
+ down_read(&nommu_region_sem);
- for (_rb = rb_first(&nommu_vma_tree); _rb; _rb = rb_next(_rb)) {
- if (pos == 0) {
- next = _rb;
- break;
- }
- pos--;
- }
-
- return next;
+ for (p = rb_first(&nommu_region_tree); p; p = rb_next(p))
+ if (pos-- == 0)
+ return p;
+ return NULL;
}
-static void nommu_vma_list_stop(struct seq_file *m, void *v)
+static void nommu_region_list_stop(struct seq_file *m, void *v)
{
- up_read(&nommu_vma_sem);
+ up_read(&nommu_region_sem);
}
-static void *nommu_vma_list_next(struct seq_file *m, void *v, loff_t *pos)
+static void *nommu_region_list_next(struct seq_file *m, void *v, loff_t *pos)
{
(*pos)++;
return rb_next((struct rb_node *) v);
}
-static const struct seq_operations proc_nommu_vma_list_seqop = {
- .start = nommu_vma_list_start,
- .next = nommu_vma_list_next,
- .stop = nommu_vma_list_stop,
- .show = nommu_vma_list_show
+static struct seq_operations proc_nommu_region_list_seqop = {
+ .start = nommu_region_list_start,
+ .next = nommu_region_list_next,
+ .stop = nommu_region_list_stop,
+ .show = nommu_region_list_show
};
-static int proc_nommu_vma_list_open(struct inode *inode, struct file *file)
+static int proc_nommu_region_list_open(struct inode *inode, struct file *file)
{
- return seq_open(file, &proc_nommu_vma_list_seqop);
+ return seq_open(file, &proc_nommu_region_list_seqop);
}
-static const struct file_operations proc_nommu_vma_list_operations = {
- .open = proc_nommu_vma_list_open,
+static const struct file_operations proc_nommu_region_list_operations = {
+ .open = proc_nommu_region_list_open,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release,
@@ -137,7 +130,7 @@ static const struct file_operations proc_nommu_vma_list_operations = {
static int __init proc_nommu_init(void)
{
- proc_create("maps", S_IRUGO, NULL, &proc_nommu_vma_list_operations);
+ proc_create("maps", S_IRUGO, NULL, &proc_nommu_region_list_operations);
return 0;
}
diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
index d4a8be3..ca4a48d 100644
--- a/fs/proc/task_nommu.c
+++ b/fs/proc/task_nommu.c
@@ -15,25 +15,25 @@
*/
void task_mem(struct seq_file *m, struct mm_struct *mm)
{
- struct vm_list_struct *vml;
+ struct vm_area_struct *vma;
+ struct rb_node *p;
unsigned long bytes = 0, sbytes = 0, slack = 0;
down_read(&mm->mmap_sem);
- for (vml = mm->context.vmlist; vml; vml = vml->next) {
- if (!vml->vma)
- continue;
+ for (p = rb_first(&mm->mm_rb); p; p = rb_next(p)) {
+ vma = rb_entry(p, struct vm_area_struct, vm_rb);
- bytes += kobjsize(vml);
+ bytes += kobjsize(vma);
if (atomic_read(&mm->mm_count) > 1 ||
- atomic_read(&vml->vma->vm_usage) > 1
- ) {
- sbytes += kobjsize((void *) vml->vma->vm_start);
- sbytes += kobjsize(vml->vma);
+ vma->vm_region ||
+ vma->vm_flags & VM_MAYSHARE) {
+ sbytes += kobjsize((void *) vma->vm_start);
+ if (vma->vm_region)
+ sbytes += kobjsize(vma->vm_region);
} else {
- bytes += kobjsize((void *) vml->vma->vm_start);
- bytes += kobjsize(vml->vma);
- slack += kobjsize((void *) vml->vma->vm_start) -
- (vml->vma->vm_end - vml->vma->vm_start);
+ bytes += kobjsize((void *) vma->vm_start);
+ slack += kobjsize((void *) vma->vm_start) -
+ (vma->vm_end - vma->vm_start);
}
}
@@ -70,13 +70,14 @@ void task_mem(struct seq_file *m, struct mm_struct *mm)
unsigned long task_vsize(struct mm_struct *mm)
{
- struct vm_list_struct *tbp;
+ struct vm_area_struct *vma;
+ struct rb_node *p;
unsigned long vsize = 0;
down_read(&mm->mmap_sem);
- for (tbp = mm->context.vmlist; tbp; tbp = tbp->next) {
- if (tbp->vma)
- vsize += kobjsize((void *) tbp->vma->vm_start);
+ for (p = rb_first(&mm->mm_rb); p; p = rb_next(p)) {
+ vma = rb_entry(p, struct vm_area_struct, vm_rb);
+ vsize += vma->vm_region->vm_end - vma->vm_region->vm_start;
}
up_read(&mm->mmap_sem);
return vsize;
@@ -85,16 +86,15 @@ unsigned long task_vsize(struct mm_struct *mm)
int task_statm(struct mm_struct *mm, int *shared, int *text,
int *data, int *resident)
{
- struct vm_list_struct *tbp;
+ struct vm_area_struct *vma;
+ struct rb_node *p;
int size = kobjsize(mm);
down_read(&mm->mmap_sem);
- for (tbp = mm->context.vmlist; tbp; tbp = tbp->next) {
- size += kobjsize(tbp);
- if (tbp->vma) {
- size += kobjsize(tbp->vma);
- size += kobjsize((void *) tbp->vma->vm_start);
- }
+ for (p = rb_first(&mm->mm_rb); p; p = rb_next(p)) {
+ vma = rb_entry(p, struct vm_area_struct, vm_rb);
+ size += kobjsize(vma);
+ size += kobjsize((void *) vma->vm_start);
}
size += (*text = mm->end_code - mm->start_code);
@@ -105,20 +105,62 @@ int task_statm(struct mm_struct *mm, int *shared, int *text,
}
/*
+ * display a single VMA to a sequenced file
+ */
+static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma)
+{
+ unsigned long ino = 0;
+ struct file *file;
+ dev_t dev = 0;
+ int flags, len;
+
+ flags = vma->vm_flags;
+ file = vma->vm_file;
+
+ if (file) {
+ struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
+ dev = inode->i_sb->s_dev;
+ ino = inode->i_ino;
+ }
+
+ seq_printf(m,
+ "%08lx-%08lx %c%c%c%c %08lx %02x:%02x %lu %n",
+ vma->vm_start,
+ vma->vm_end,
+ flags & VM_READ ? 'r' : '-',
+ flags & VM_WRITE ? 'w' : '-',
+ flags & VM_EXEC ? 'x' : '-',
+ flags & VM_MAYSHARE ? flags & VM_SHARED ? 'S' : 's' : 'p',
+ vma->vm_pgoff << PAGE_SHIFT,
+ MAJOR(dev), MINOR(dev), ino, &len);
+
+ if (file) {
+ len = 25 + sizeof(void *) * 6 - len;
+ if (len < 1)
+ len = 1;
+ seq_printf(m, "%*c", len, ' ');
+ seq_path(m, &file->f_path, "");
+ }
+
+ seq_putc(m, '\n');
+ return 0;
+}
+
+/*
* display mapping lines for a particular process's /proc/pid/maps
*/
-static int show_map(struct seq_file *m, void *_vml)
+static int show_map(struct seq_file *m, void *_p)
{
- struct vm_list_struct *vml = _vml;
+ struct rb_node *p = _p;
- return nommu_vma_show(m, vml->vma);
+ return nommu_vma_show(m, rb_entry(p, struct vm_area_struct, vm_rb));
}
static void *m_start(struct seq_file *m, loff_t *pos)
{
struct proc_maps_private *priv = m->private;
- struct vm_list_struct *vml;
struct mm_struct *mm;
+ struct rb_node *p;
loff_t n = *pos;
/* pin the task and mm whilst we play with them */
@@ -134,9 +176,9 @@ static void *m_start(struct seq_file *m, loff_t *pos)
}
/* start from the Nth VMA */
- for (vml = mm->context.vmlist; vml; vml = vml->next)
+ for (p = rb_first(&mm->mm_rb); p; p = rb_next(p))
if (n-- == 0)
- return vml;
+ return p;
return NULL;
}
@@ -152,12 +194,12 @@ static void m_stop(struct seq_file *m, void *_vml)
}
}
-static void *m_next(struct seq_file *m, void *_vml, loff_t *pos)
+static void *m_next(struct seq_file *m, void *_p, loff_t *pos)
{
- struct vm_list_struct *vml = _vml;
+ struct rb_node *p = _p;
(*pos)++;
- return vml ? vml->next : NULL;
+ return p ? rb_next(p) : NULL;
}
static const struct seq_operations proc_pid_maps_ops = {
diff --git a/include/asm-frv/mmu.h b/include/asm-frv/mmu.h
index 22c0371..86ca0e8 100644
--- a/include/asm-frv/mmu.h
+++ b/include/asm-frv/mmu.h
@@ -22,7 +22,6 @@ typedef struct {
unsigned long dtlb_ptd_mapping; /* [DAMR5] PTD mapping for dtlb cached PGE */
#else
- struct vm_list_struct *vmlist;
unsigned long end_brk;
#endif
diff --git a/include/asm-m32r/mmu.h b/include/asm-m32r/mmu.h
index d9bd724..150cb92 100644
--- a/include/asm-m32r/mmu.h
+++ b/include/asm-m32r/mmu.h
@@ -4,7 +4,6 @@
#if !defined(CONFIG_MMU)
typedef struct {
- struct vm_list_struct *vmlist;
unsigned long end_brk;
} mm_context_t;
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 4a3d28c..b91a73f 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -56,19 +56,9 @@ extern unsigned long mmap_min_addr;
extern struct kmem_cache *vm_area_cachep;
-/*
- * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
- * disabled, then there's a single shared list of VMAs maintained by the
- * system, and mm's subscribe to these individually
- */
-struct vm_list_struct {
- struct vm_list_struct *next;
- struct vm_area_struct *vma;
-};
-
#ifndef CONFIG_MMU
-extern struct rb_root nommu_vma_tree;
-extern struct rw_semaphore nommu_vma_sem;
+extern struct rb_root nommu_region_tree;
+extern struct rw_semaphore nommu_region_sem;
extern unsigned int kobjsize(const void *objp);
#endif
@@ -1061,6 +1051,7 @@ extern void memmap_init_zone(unsigned long, int, unsigned long,
unsigned long, enum memmap_context);
extern void setup_per_zone_pages_min(void);
extern void mem_init(void);
+extern void __init mmap_init(void);
extern void show_mem(void);
extern void si_meminfo(struct sysinfo * val);
extern void si_meminfo_node(struct sysinfo *val, int nid);
@@ -1072,6 +1063,9 @@ extern void setup_per_cpu_pageset(void);
static inline void setup_per_cpu_pageset(void) {}
#endif
+/* nommu.c */
+extern atomic_t mmap_pages_allocated;
+
/* prio_tree.c */
void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 9cfc9b6..1c1e0d3 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -97,6 +97,22 @@ struct page {
};
/*
+ * A region containing a mapping of a non-memory backed file under NOMMU
+ * conditions. These are held in a global tree and are pinned by the VMAs that
+ * map parts of them.
+ */
+struct vm_region {
+ struct rb_node vm_rb; /* link in global region tree */
+ unsigned long vm_flags; /* VMA vm_flags */
+ unsigned long vm_start; /* start address of region */
+ unsigned long vm_end; /* region initialised to here */
+ unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
+ struct file *vm_file; /* the backing file or NULL */
+
+ atomic_t vm_usage; /* region usage count */
+};
+
+/*
* This struct defines a memory VMM memory area. There is one of these
* per VM-area/task. A VM area is any part of the process virtual memory
* space that has a special rule for the page-fault handlers (ie a shared
@@ -152,7 +168,7 @@ struct vm_area_struct {
unsigned long vm_truncate_count;/* truncate_count or restart_addr */
#ifndef CONFIG_MMU
- atomic_t vm_usage; /* refcount (VMAs shared if !MMU) */
+ struct vm_region *vm_region; /* NOMMU mapping region */
#endif
#ifdef CONFIG_NUMA
struct mempolicy *vm_policy; /* NUMA policy for the VMA */
diff --git a/ipc/shm.c b/ipc/shm.c
index b125b56..d0ab552 100644
--- a/ipc/shm.c
+++ b/ipc/shm.c
@@ -990,6 +990,7 @@ asmlinkage long sys_shmdt(char __user *shmaddr)
*/
vma = find_vma(mm, addr);
+#ifdef CONFIG_MMU
while (vma) {
next = vma->vm_next;
@@ -1034,6 +1035,17 @@ asmlinkage long sys_shmdt(char __user *shmaddr)
vma = next;
}
+#else /* CONFIG_MMU */
+ /* under NOMMU conditions, the exact address to be destroyed must be
+ * given */
+ retval = -EINVAL;
+ if (vma->vm_start == addr && vma->vm_ops == &shm_vm_ops) {
+ do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
+ retval = 0;
+ }
+
+#endif
+
up_write(&mm->mmap_sem);
return retval;
}
diff --git a/kernel/fork.c b/kernel/fork.c
index 7b8f2a7..0bce4a4 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -1481,12 +1481,10 @@ void __init proc_caches_init(void)
fs_cachep = kmem_cache_create("fs_cache",
sizeof(struct fs_struct), 0,
SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
- vm_area_cachep = kmem_cache_create("vm_area_struct",
- sizeof(struct vm_area_struct), 0,
- SLAB_PANIC, NULL);
mm_cachep = kmem_cache_create("mm_struct",
sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
+ mmap_init();
}
/*
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 2e75478..d0a32aa 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -512,6 +512,13 @@ config DEBUG_VIRTUAL
If unsure, say N.
+config DEBUG_NOMMU_REGIONS
+ bool "Debug the global anon/private NOMMU mapping region tree"
+ depends on DEBUG_KERNEL && !MMU
+ help
+ This option causes the global tree of anonymous and private mapping
+ regions to be regularly checked for invalid topology.
+
config DEBUG_WRITECOUNT
bool "Debug filesystem writers count"
depends on DEBUG_KERNEL
diff --git a/mm/mmap.c b/mm/mmap.c
index a910c04..7496231 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2472,3 +2472,13 @@ void mm_drop_all_locks(struct mm_struct *mm)
mutex_unlock(&mm_all_locks_mutex);
}
+
+/*
+ * initialise the VMA slab
+ */
+void __init mmap_init(void)
+{
+ vm_area_cachep = kmem_cache_create("vm_area_struct",
+ sizeof(struct vm_area_struct), 0,
+ SLAB_PANIC, NULL);
+}
diff --git a/mm/nommu.c b/mm/nommu.c
index 23f355b..0d363df 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -6,7 +6,7 @@
*
* See Documentation/nommu-mmap.txt
*
- * Copyright (c) 2004-2005 David Howells <dhowells@redhat.com>
+ * Copyright (c) 2004-2008 David Howells <dhowells@redhat.com>
* Copyright (c) 2000-2003 David McCullough <davidm@snapgear.com>
* Copyright (c) 2000-2001 D Jeff Dionne <jeff@uClinux.org>
* Copyright (c) 2002 Greg Ungerer <gerg@snapgear.com>
@@ -33,6 +33,28 @@
#include <asm/uaccess.h>
#include <asm/tlb.h>
#include <asm/tlbflush.h>
+#include "internal.h"
+
+static inline __attribute__((format(printf, 1, 2)))
+void no_printk(const char *fmt, ...)
+{
+}
+
+#if 0
+#define kenter(FMT, ...) \
+ printk(KERN_DEBUG "==> %s("FMT")\n", __func__, ##__VA_ARGS__)
+#define kleave(FMT, ...) \
+ printk(KERN_DEBUG "<== %s()"FMT"\n", __func__, ##__VA_ARGS__)
+#define kdebug(FMT, ...) \
+ printk(KERN_DEBUG "xxx" FMT"yyy\n", ##__VA_ARGS__)
+#else
+#define kenter(FMT, ...) \
+ no_printk(KERN_DEBUG "==> %s("FMT")\n", __func__, ##__VA_ARGS__)
+#define kleave(FMT, ...) \
+ no_printk(KERN_DEBUG "<== %s()"FMT"\n", __func__, ##__VA_ARGS__)
+#define kdebug(FMT, ...) \
+ no_printk(KERN_DEBUG FMT"\n", ##__VA_ARGS__)
+#endif
#include "internal.h"
@@ -46,12 +68,15 @@ int sysctl_overcommit_ratio = 50; /* default is 50% */
int sysctl_max_map_count = DEFAULT_MAX_MAP_COUNT;
int heap_stack_gap = 0;
+atomic_t mmap_pages_allocated;
+
EXPORT_SYMBOL(mem_map);
EXPORT_SYMBOL(num_physpages);
-/* list of shareable VMAs */
-struct rb_root nommu_vma_tree = RB_ROOT;
-DECLARE_RWSEM(nommu_vma_sem);
+/* list of mapped, potentially shareable regions */
+static struct kmem_cache *vm_region_jar;
+struct rb_root nommu_region_tree = RB_ROOT;
+DECLARE_RWSEM(nommu_region_sem);
struct vm_operations_struct generic_file_vm_ops = {
};
@@ -400,129 +425,174 @@ asmlinkage unsigned long sys_brk(unsigned long brk)
return mm->brk = brk;
}
-#ifdef DEBUG
-static void show_process_blocks(void)
+/*
+ * initialise the VMA and region record slabs
+ */
+void __init mmap_init(void)
{
- struct vm_list_struct *vml;
-
- printk("Process blocks %d:", current->pid);
-
- for (vml = ¤t->mm->context.vmlist; vml; vml = vml->next) {
- printk(" %p: %p", vml, vml->vma);
- if (vml->vma)
- printk(" (%d @%lx #%d)",
- kobjsize((void *) vml->vma->vm_start),
- vml->vma->vm_start,
- atomic_read(&vml->vma->vm_usage));
- printk(vml->next ? " ->" : ".\n");
- }
+ vm_region_jar = kmem_cache_create("vm_region_jar",
+ sizeof(struct vm_region), 0,
+ SLAB_PANIC, NULL);
+ vm_area_cachep = kmem_cache_create("vm_area_struct",
+ sizeof(struct vm_area_struct), 0,
+ SLAB_PANIC, NULL);
}
-#endif /* DEBUG */
/*
- * add a VMA into a process's mm_struct in the appropriate place in the list
- * - should be called with mm->mmap_sem held writelocked
+ * validate the region tree
+ * - the caller must hold the region lock
*/
-static void add_vma_to_mm(struct mm_struct *mm, struct vm_list_struct *vml)
+#ifdef CONFIG_DEBUG_NOMMU_REGIONS
+static noinline void validate_nommu_regions(void)
{
- struct vm_list_struct **ppv;
+ struct vm_region *region, *last;
+ struct rb_node *p, *lastp;
- for (ppv = ¤t->mm->context.vmlist; *ppv; ppv = &(*ppv)->next)
- if ((*ppv)->vma->vm_start > vml->vma->vm_start)
- break;
+ lastp = rb_first(&nommu_region_tree);
+ if (!lastp)
+ return;
+
+ last = rb_entry(lastp, struct vm_region, vm_rb);
+ if (unlikely(last->vm_end <= last->vm_start))
+ BUG();
+
+ while ((p = rb_next(lastp))) {
+ region = rb_entry(p, struct vm_region, vm_rb);
+ last = rb_entry(lastp, struct vm_region, vm_rb);
+
+ if (unlikely(region->vm_end <= region->vm_start))
+ BUG();
+ if (unlikely(region->vm_start < last->vm_end))
+ BUG();
- vml->next = *ppv;
- *ppv = vml;
+ lastp = p;
+ }
}
+#else
+#define validate_nommu_regions() do {} while(0)
+#endif
/*
- * look up the first VMA in which addr resides, NULL if none
- * - should be called with mm->mmap_sem at least held readlocked
+ * add a region into the global tree
*/
-struct vm_area_struct *find_vma(struct mm_struct *mm, unsigned long addr)
+static void add_nommu_region(struct vm_region *region)
{
- struct vm_list_struct *loop, *vml;
+ struct vm_region *pregion;
+ struct rb_node **p, *parent;
- /* search the vm_start ordered list */
- vml = NULL;
- for (loop = mm->context.vmlist; loop; loop = loop->next) {
- if (loop->vma->vm_start > addr)
- break;
- vml = loop;
+ validate_nommu_regions();
+
+ BUG_ON(region->vm_start & ~PAGE_MASK);
+
+ parent = NULL;
+ p = &nommu_region_tree.rb_node;
+ while (*p) {
+ parent = *p;
+ pregion = rb_entry(parent, struct vm_region, vm_rb);
+ if (region->vm_start < pregion->vm_start)
+ p = &(*p)->rb_left;
+ else if (region->vm_start > pregion->vm_start)
+ p = &(*p)->rb_right;
+ else if (pregion == region)
+ return;
+ else
+ BUG();
}
- if (vml && vml->vma->vm_end > addr)
- return vml->vma;
+ rb_link_node(®ion->vm_rb, parent, p);
+ rb_insert_color(®ion->vm_rb, &nommu_region_tree);
- return NULL;
+ validate_nommu_regions();
}
-EXPORT_SYMBOL(find_vma);
/*
- * find a VMA
- * - we don't extend stack VMAs under NOMMU conditions
+ * delete a region from the global tree
*/
-struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr)
+static void delete_nommu_region(struct vm_region *region)
{
- return find_vma(mm, addr);
-}
+ BUG_ON(!nommu_region_tree.rb_node);
-int expand_stack(struct vm_area_struct *vma, unsigned long address)
-{
- return -ENOMEM;
+ validate_nommu_regions();
+ rb_erase(®ion->vm_rb, &nommu_region_tree);
+ validate_nommu_regions();
}
/*
- * look up the first VMA exactly that exactly matches addr
- * - should be called with mm->mmap_sem at least held readlocked
+ * free a contiguous series of pages
*/
-static inline struct vm_area_struct *find_vma_exact(struct mm_struct *mm,
- unsigned long addr)
+static void free_page_series(unsigned long from, unsigned long to)
{
- struct vm_list_struct *vml;
-
- /* search the vm_start ordered list */
- for (vml = mm->context.vmlist; vml; vml = vml->next) {
- if (vml->vma->vm_start == addr)
- return vml->vma;
- if (vml->vma->vm_start > addr)
- break;
+ for (; from < to; from += PAGE_SIZE) {
+ struct page *page = virt_to_page(from);
+
+ kdebug("- free %lx", from);
+ atomic_dec(&mmap_pages_allocated);
+ if (page_count(page) != 1)
+ kdebug("free page %p [%d]", page, page_count(page));
+ put_page(page);
}
-
- return NULL;
}
/*
- * find a VMA in the global tree
+ * release a reference to a region
+ * - the caller must hold the region semaphore, which this releases
+ * - the region may not have been added to the tree yet, in which case vm_end
+ * will equal vm_start
*/
-static inline struct vm_area_struct *find_nommu_vma(unsigned long start)
+static void __put_nommu_region(struct vm_region *region)
+ __releases(nommu_region_sem)
{
- struct vm_area_struct *vma;
- struct rb_node *n = nommu_vma_tree.rb_node;
+ kenter("%p{%d}", region, atomic_read(®ion->vm_usage));
- while (n) {
- vma = rb_entry(n, struct vm_area_struct, vm_rb);
+ BUG_ON(!nommu_region_tree.rb_node);
- if (start < vma->vm_start)
- n = n->rb_left;
- else if (start > vma->vm_start)
- n = n->rb_right;
- else
- return vma;
+ if (atomic_dec_and_test(®ion->vm_usage)) {
+ if (region->vm_end > region->vm_start)
+ delete_nommu_region(region);
+ up_write(&nommu_region_sem);
+
+ if (region->vm_file)
+ fput(region->vm_file);
+
+ /* IO memory and memory shared directly out of the pagecache
+ * from ramfs/tmpfs mustn't be released here */
+ if (region->vm_flags & VM_MAPPED_COPY) {
+ kdebug("free series");
+ free_page_series(region->vm_start, region->vm_end);
+ }
+ kmem_cache_free(vm_region_jar, region);
+ } else {
+ up_write(&nommu_region_sem);
}
+}
- return NULL;
+/*
+ * release a reference to a region
+ */
+static void put_nommu_region(struct vm_region *region)
+{
+ down_write(&nommu_region_sem);
+ __put_nommu_region(region);
}
/*
- * add a VMA in the global tree
+ * add a VMA into a process's mm_struct in the appropriate place in the list
+ * and tree and add to the address space's page tree also if not an anonymous
+ * page
+ * - should be called with mm->mmap_sem held writelocked
*/
-static void add_nommu_vma(struct vm_area_struct *vma)
+static void add_vma_to_mm(struct mm_struct *mm, struct vm_area_struct *vma)
{
- struct vm_area_struct *pvma;
+ struct vm_area_struct *pvma, **pp;
struct address_space *mapping;
- struct rb_node **p = &nommu_vma_tree.rb_node;
- struct rb_node *parent = NULL;
+ struct rb_node **p, *parent;
+
+ kenter(",%p", vma);
+
+ BUG_ON(!vma->vm_region);
+
+ mm->map_count++;
+ vma->vm_mm = mm;
/* add the VMA to the mapping */
if (vma->vm_file) {
@@ -533,42 +603,62 @@ static void add_nommu_vma(struct vm_area_struct *vma)
flush_dcache_mmap_unlock(mapping);
}
- /* add the VMA to the master list */
+ /* add the VMA to the tree */
+ parent = NULL;
+ p = &mm->mm_rb.rb_node;
while (*p) {
parent = *p;
pvma = rb_entry(parent, struct vm_area_struct, vm_rb);
- if (vma->vm_start < pvma->vm_start) {
+ /* sort by: start addr, end addr, VMA struct addr in that order
+ * (the latter is necessary as we may get identical VMAs) */
+ if (vma->vm_start < pvma->vm_start)
p = &(*p)->rb_left;
- }
- else if (vma->vm_start > pvma->vm_start) {
+ else if (vma->vm_start > pvma->vm_start)
p = &(*p)->rb_right;
- }
- else {
- /* mappings are at the same address - this can only
- * happen for shared-mem chardevs and shared file
- * mappings backed by ramfs/tmpfs */
- BUG_ON(!(pvma->vm_flags & VM_SHARED));
-
- if (vma < pvma)
- p = &(*p)->rb_left;
- else if (vma > pvma)
- p = &(*p)->rb_right;
- else
- BUG();
- }
+ else if (vma->vm_end < pvma->vm_end)
+ p = &(*p)->rb_left;
+ else if (vma->vm_end > pvma->vm_end)
+ p = &(*p)->rb_right;
+ else if (vma < pvma)
+ p = &(*p)->rb_left;
+ else if (vma > pvma)
+ p = &(*p)->rb_right;
+ else
+ BUG();
}
rb_link_node(&vma->vm_rb, parent, p);
- rb_insert_color(&vma->vm_rb, &nommu_vma_tree);
+ rb_insert_color(&vma->vm_rb, &mm->mm_rb);
+
+ /* add VMA to the VMA list also */
+ for (pp = &mm->mmap; (pvma = *pp); pp = &(*pp)->vm_next) {
+ if (pvma->vm_start > vma->vm_start)
+ break;
+ if (pvma->vm_start < vma->vm_start)
+ continue;
+ if (pvma->vm_end < vma->vm_end)
+ break;
+ }
+
+ vma->vm_next = *pp;
+ *pp = vma;
}
/*
- * delete a VMA from the global list
+ * delete a VMA from its owning mm_struct and address space
*/
-static void delete_nommu_vma(struct vm_area_struct *vma)
+static void delete_vma_from_mm(struct vm_area_struct *vma)
{
+ struct vm_area_struct **pp;
struct address_space *mapping;
+ struct mm_struct *mm = vma->vm_mm;
+
+ kenter("%p", vma);
+
+ mm->map_count--;
+ if (mm->mmap_cache == vma)
+ mm->mmap_cache = NULL;
/* remove the VMA from the mapping */
if (vma->vm_file) {
@@ -579,8 +669,115 @@ static void delete_nommu_vma(struct vm_area_struct *vma)
flush_dcache_mmap_unlock(mapping);
}
- /* remove from the master list */
- rb_erase(&vma->vm_rb, &nommu_vma_tree);
+ /* remove from the MM's tree and list */
+ rb_erase(&vma->vm_rb, &mm->mm_rb);
+ for (pp = &mm->mmap; *pp; pp = &(*pp)->vm_next) {
+ if (*pp == vma) {
+ *pp = vma->vm_next;
+ break;
+ }
+ }
+
+ vma->vm_mm = NULL;
+}
+
+/*
+ * destroy a VMA record
+ */
+static void delete_vma(struct mm_struct *mm, struct vm_area_struct *vma)
+{
+ kenter("%p", vma);
+ if (vma->vm_ops && vma->vm_ops->close)
+ vma->vm_ops->close(vma);
+ if (vma->vm_file) {
+ fput(vma->vm_file);
+ if (vma->vm_flags & VM_EXECUTABLE)
+ removed_exe_file_vma(mm);
+ }
+ put_nommu_region(vma->vm_region);
+ kmem_cache_free(vm_area_cachep, vma);
+}
+
+/*
+ * look up the first VMA in which addr resides, NULL if none
+ * - should be called with mm->mmap_sem at least held readlocked
+ */
+struct vm_area_struct *find_vma(struct mm_struct *mm, unsigned long addr)
+{
+ struct vm_area_struct *vma;
+ struct rb_node *n = mm->mm_rb.rb_node;
+
+ /* check the cache first */
+ vma = mm->mmap_cache;
+ if (vma && vma->vm_start <= addr && vma->vm_end > addr)
+ return vma;
+
+ /* trawl the tree (there may be multiple mappings in which addr
+ * resides) */
+ for (n = rb_first(&mm->mm_rb); n; n = rb_next(n)) {
+ vma = rb_entry(n, struct vm_area_struct, vm_rb);
+ if (vma->vm_start > addr)
+ return NULL;
+ if (vma->vm_end > addr) {
+ mm->mmap_cache = vma;
+ return vma;
+ }
+ }
+
+ return NULL;
+}
+EXPORT_SYMBOL(find_vma);
+
+/*
+ * find a VMA
+ * - we don't extend stack VMAs under NOMMU conditions
+ */
+struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr)
+{
+ return find_vma(mm, addr);
+}
+
+/*
+ * expand a stack to a given address
+ * - not supported under NOMMU conditions
+ */
+int expand_stack(struct vm_area_struct *vma, unsigned long address)
+{
+ return -ENOMEM;
+}
+
+/*
+ * look up the first VMA exactly that exactly matches addr
+ * - should be called with mm->mmap_sem at least held readlocked
+ */
+static struct vm_area_struct *find_vma_exact(struct mm_struct *mm,
+ unsigned long addr,
+ unsigned long len)
+{
+ struct vm_area_struct *vma;
+ struct rb_node *n = mm->mm_rb.rb_node;
+ unsigned long end = addr + len;
+
+ /* check the cache first */
+ vma = mm->mmap_cache;
+ if (vma && vma->vm_start == addr && vma->vm_end == end)
+ return vma;
+
+ /* trawl the tree (there may be multiple mappings in which addr
+ * resides) */
+ for (n = rb_first(&mm->mm_rb); n; n = rb_next(n)) {
+ vma = rb_entry(n, struct vm_area_struct, vm_rb);
+ if (vma->vm_start < addr)
+ continue;
+ if (vma->vm_start > addr)
+ return NULL;
+ if (vma->vm_end == end) {
+ mm->mmap_cache = vma;
+ return vma;
+ }
+ }
+
+ return NULL;
}
/*
@@ -595,7 +792,7 @@ static int validate_mmap_request(struct file *file,
unsigned long pgoff,
unsigned long *_capabilities)
{
- unsigned long capabilities;
+ unsigned long capabilities, rlen;
unsigned long reqprot = prot;
int ret;
@@ -615,12 +812,12 @@ static int validate_mmap_request(struct file *file,
return -EINVAL;
/* Careful about overflows.. */
- len = PAGE_ALIGN(len);
- if (!len || len > TASK_SIZE)
+ rlen = PAGE_ALIGN(len);
+ if (!rlen || rlen > TASK_SIZE)
return -ENOMEM;
/* offset overflow? */
- if ((pgoff + (len >> PAGE_SHIFT)) < pgoff)
+ if ((pgoff + (rlen >> PAGE_SHIFT)) < pgoff)
return -EOVERFLOW;
if (file) {
@@ -794,9 +991,10 @@ static unsigned long determine_vm_flags(struct file *file,
}
/*
- * set up a shared mapping on a file
+ * set up a shared mapping on a file (the driver or filesystem provides and
+ * pins the storage)
*/
-static int do_mmap_shared_file(struct vm_area_struct *vma, unsigned long len)
+static int do_mmap_shared_file(struct vm_area_struct *vma)
{
int ret;
@@ -814,10 +1012,14 @@ static int do_mmap_shared_file(struct vm_area_struct *vma, unsigned long len)
/*
* set up a private mapping or an anonymous shared mapping
*/
-static int do_mmap_private(struct vm_area_struct *vma, unsigned long len)
+static int do_mmap_private(struct vm_area_struct *vma,
+ struct vm_region *region,
+ unsigned long len)
{
+ struct page *pages;
+ unsigned long total, point, n, rlen;
void *base;
- int ret;
+ int ret, order;
/* invoke the file's mapping function so that it can keep track of
* shared mappings on devices or memory
@@ -836,23 +1038,46 @@ static int do_mmap_private(struct vm_area_struct *vma, unsigned long len)
* make a private copy of the data and map that instead */
}
+ rlen = PAGE_ALIGN(len);
+
/* allocate some memory to hold the mapping
* - note that this may not return a page-aligned address if the object
* we're allocating is smaller than a page
*/
- base = kmalloc(len, GFP_KERNEL|__GFP_COMP);
- if (!base)
+ order = get_order(rlen);
+ kdebug("alloc order %d for %lx", order, len);
+
+ pages = alloc_pages(GFP_KERNEL, order);
+ if (!pages)
goto enomem;
- vma->vm_start = (unsigned long) base;
- vma->vm_end = vma->vm_start + len;
- vma->vm_flags |= VM_MAPPED_COPY;
+ /* we allocated a power-of-2 sized page set, so we need to trim off the
+ * excess */
+ total = 1 << order;
+ atomic_add(total, &mmap_pages_allocated);
+
+ point = rlen >> PAGE_SHIFT;
+ while (total > point) {
+ order = ilog2(total - point);
+ n = 1 << order;
+ kdebug("shave %lu/%lu @%lu", n, total - point, total);
+ atomic_sub(n, &mmap_pages_allocated);
+ total -= n;
+ set_page_refcounted(pages + total);
+ __free_pages(pages + total, order);
+ }
+
+ total = rlen >> PAGE_SHIFT;
+ for (point = 1; point < total; point++)
+ set_page_refcounted(&pages[point]);
-#ifdef WARN_ON_SLACK
- if (len + WARN_ON_SLACK <= kobjsize(result))
- printk("Allocation of %lu bytes from process %d has %lu bytes of slack\n",
- len, current->pid, kobjsize(result) - len);
-#endif
+ base = page_address(pages);
+ region->vm_flags = vma->vm_flags |= VM_MAPPED_COPY;
+ region->vm_start = (unsigned long) base;
+ region->vm_end = region->vm_start + rlen;
+
+ vma->vm_start = region->vm_start;
+ vma->vm_end = region->vm_start + len;
if (vma->vm_file) {
/* read the contents of a file into the copy */
@@ -864,26 +1089,27 @@ static int do_mmap_private(struct vm_area_struct *vma, unsigned long len)
old_fs = get_fs();
set_fs(KERNEL_DS);
- ret = vma->vm_file->f_op->read(vma->vm_file, base, len, &fpos);
+ ret = vma->vm_file->f_op->read(vma->vm_file, base, rlen, &fpos);
set_fs(old_fs);
if (ret < 0)
goto error_free;
/* clear the last little bit */
- if (ret < len)
- memset(base + ret, 0, len - ret);
+ if (ret < rlen)
+ memset(base + ret, 0, rlen - ret);
} else {
/* if it's an anonymous mapping, then just clear it */
- memset(base, 0, len);
+ memset(base, 0, rlen);
}
return 0;
error_free:
- kfree(base);
- vma->vm_start = 0;
+ free_page_series(region->vm_start, region->vm_end);
+ region->vm_start = vma->vm_start = 0;
+ region->vm_end = vma->vm_end = 0;
return ret;
enomem:
@@ -903,13 +1129,14 @@ unsigned long do_mmap_pgoff(struct file *file,
unsigned long flags,
unsigned long pgoff)
{
- struct vm_list_struct *vml = NULL;
- struct vm_area_struct *vma = NULL;
+ struct vm_area_struct *vma;
+ struct vm_region *region;
struct rb_node *rb;
- unsigned long capabilities, vm_flags;
- void *result;
+ unsigned long capabilities, vm_flags, result;
int ret;
+ kenter(",%lx,%lx,%lx,%lx,%lx", addr, len, prot, flags, pgoff);
+
if (!(flags & MAP_FIXED))
addr = round_hint_to_min(addr);
@@ -917,73 +1144,120 @@ unsigned long do_mmap_pgoff(struct file *file,
* mapping */
ret = validate_mmap_request(file, addr, len, prot, flags, pgoff,
&capabilities);
- if (ret < 0)
+ if (ret < 0) {
+ kleave(" = %d [val]", ret);
return ret;
+ }
/* we've determined that we can make the mapping, now translate what we
* now know into VMA flags */
vm_flags = determine_vm_flags(file, prot, flags, capabilities);
- /* we're going to need to record the mapping if it works */
- vml = kzalloc(sizeof(struct vm_list_struct), GFP_KERNEL);
- if (!vml)
- goto error_getting_vml;
+ /* we're going to need to record the mapping */
+ region = kmem_cache_zalloc(vm_region_jar, GFP_KERNEL);
+ if (!region)
+ goto error_getting_region;
+
+ vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
+ if (!vma)
+ goto error_getting_vma;
+
+ atomic_set(®ion->vm_usage, 1);
+ region->vm_flags = vm_flags;
+ region->vm_pgoff = pgoff;
- down_write(&nommu_vma_sem);
+ INIT_LIST_HEAD(&vma->anon_vma_node);
+ vma->vm_flags = vm_flags;
+ vma->vm_pgoff = pgoff;
- /* if we want to share, we need to check for VMAs created by other
+ if (file) {
+ region->vm_file = file;
+ get_file(file);
+ vma->vm_file = file;
+ get_file(file);
+ if (vm_flags & VM_EXECUTABLE) {
+ added_exe_file_vma(current->mm);
+ vma->vm_mm = current->mm;
+ }
+ }
+
+ down_write(&nommu_region_sem);
+
+ /* if we want to share, we need to check for regions created by other
* mmap() calls that overlap with our proposed mapping
- * - we can only share with an exact match on most regular files
+ * - we can only share with a superset match on most regular files
* - shared mappings on character devices and memory backed files are
* permitted to overlap inexactly as far as we are concerned for in
* these cases, sharing is handled in the driver or filesystem rather
* than here
*/
if (vm_flags & VM_MAYSHARE) {
- unsigned long pglen = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
- unsigned long vmpglen;
+ struct vm_region *pregion;
+ unsigned long pglen, rpglen, pgend, rpgend, start;
- /* suppress VMA sharing for shared regions */
- if (vm_flags & VM_SHARED &&
- capabilities & BDI_CAP_MAP_DIRECT)
- goto dont_share_VMAs;
+ pglen = (len + PAGE_SIZE - 1) >> PAGE_SHIFT;
+ pgend = pgoff + pglen;
- for (rb = rb_first(&nommu_vma_tree); rb; rb = rb_next(rb)) {
- vma = rb_entry(rb, struct vm_area_struct, vm_rb);
+ for (rb = rb_first(&nommu_region_tree); rb; rb = rb_next(rb)) {
+ pregion = rb_entry(rb, struct vm_region, vm_rb);
- if (!(vma->vm_flags & VM_MAYSHARE))
+ if (!(pregion->vm_flags & VM_MAYSHARE))
continue;
/* search for overlapping mappings on the same file */
- if (vma->vm_file->f_path.dentry->d_inode != file->f_path.dentry->d_inode)
+ if (pregion->vm_file->f_path.dentry->d_inode !=
+ file->f_path.dentry->d_inode)
continue;
- if (vma->vm_pgoff >= pgoff + pglen)
+ if (pregion->vm_pgoff >= pgend)
continue;
- vmpglen = vma->vm_end - vma->vm_start + PAGE_SIZE - 1;
- vmpglen >>= PAGE_SHIFT;
- if (pgoff >= vma->vm_pgoff + vmpglen)
+ rpglen = pregion->vm_end - pregion->vm_start;
+ rpglen = (rpglen + PAGE_SIZE - 1) >> PAGE_SHIFT;
+ rpgend = pregion->vm_pgoff + rpglen;
+ if (pgoff >= rpgend)
continue;
- /* handle inexactly overlapping matches between mappings */
- if (vma->vm_pgoff != pgoff || vmpglen != pglen) {
+ /* handle inexactly overlapping matches between
+ * mappings */
+ if ((pregion->vm_pgoff != pgoff || rpglen != pglen) &&
+ !(pgoff >= pregion->vm_pgoff && pgend <= rpgend)) {
+ /* new mapping is not a subset of the region */
if (!(capabilities & BDI_CAP_MAP_DIRECT))
goto sharing_violation;
continue;
}
- /* we've found a VMA we can share */
- atomic_inc(&vma->vm_usage);
-
- vml->vma = vma;
- result = (void *) vma->vm_start;
- goto shared;
+ /* we've found a region we can share */
+ atomic_inc(&pregion->vm_usage);
+ vma->vm_region = pregion;
+ start = pregion->vm_start;
+ start += (pgoff - pregion->vm_pgoff) << PAGE_SHIFT;
+ vma->vm_start = start;
+ vma->vm_end = start + len;
+
+ if (pregion->vm_flags & VM_MAPPED_COPY) {
+ kdebug("share copy");
+ vma->vm_flags |= VM_MAPPED_COPY;
+ } else {
+ kdebug("share mmap");
+ ret = do_mmap_shared_file(vma);
+ if (ret < 0) {
+ vma->vm_region = NULL;
+ vma->vm_start = 0;
+ vma->vm_end = 0;
+ atomic_dec(&pregion->vm_usage);
+ pregion = NULL;
+ goto error_just_free;
+ }
+ }
+ fput(region->vm_file);
+ kmem_cache_free(vm_region_jar, region);
+ region = pregion;
+ result = start;
+ goto share;
}
- dont_share_VMAs:
- vma = NULL;
-
/* obtain the address at which to make a shared mapping
* - this is the hook for quasi-memory character devices to
* tell us the location of a shared mapping
@@ -994,102 +1268,93 @@ unsigned long do_mmap_pgoff(struct file *file,
if (IS_ERR((void *) addr)) {
ret = addr;
if (ret != (unsigned long) -ENOSYS)
- goto error;
+ goto error_just_free;
/* the driver refused to tell us where to site
* the mapping so we'll have to attempt to copy
* it */
ret = (unsigned long) -ENODEV;
if (!(capabilities & BDI_CAP_MAP_COPY))
- goto error;
+ goto error_just_free;
capabilities &= ~BDI_CAP_MAP_DIRECT;
+ } else {
+ vma->vm_start = region->vm_start = addr;
+ vma->vm_end = region->vm_end = addr + len;
}
}
}
- /* we're going to need a VMA struct as well */
- vma = kzalloc(sizeof(struct vm_area_struct), GFP_KERNEL);
- if (!vma)
- goto error_getting_vma;
-
- INIT_LIST_HEAD(&vma->anon_vma_node);
- atomic_set(&vma->vm_usage, 1);
- if (file) {
- get_file(file);
- if (vm_flags & VM_EXECUTABLE) {
- added_exe_file_vma(current->mm);
- vma->vm_mm = current->mm;
- }
- }
- vma->vm_file = file;
- vma->vm_flags = vm_flags;
- vma->vm_start = addr;
- vma->vm_end = addr + len;
- vma->vm_pgoff = pgoff;
-
- vml->vma = vma;
+ vma->vm_region = region;
/* set up the mapping */
if (file && vma->vm_flags & VM_SHARED)
- ret = do_mmap_shared_file(vma, len);
+ ret = do_mmap_shared_file(vma);
else
- ret = do_mmap_private(vma, len);
+ ret = do_mmap_private(vma, region, len);
if (ret < 0)
- goto error;
+ goto error_put_region;
+
+ add_nommu_region(region);
/* okay... we have a mapping; now we have to register it */
- result = (void *) vma->vm_start;
+ result = vma->vm_start;
current->mm->total_vm += len >> PAGE_SHIFT;
- add_nommu_vma(vma);
+share:
+ add_vma_to_mm(current->mm, vma);
- shared:
- add_vma_to_mm(current->mm, vml);
-
- up_write(&nommu_vma_sem);
+ up_write(&nommu_region_sem);
if (prot & PROT_EXEC)
- flush_icache_range((unsigned long) result,
- (unsigned long) result + len);
+ flush_icache_range(result, result + len);
-#ifdef DEBUG
- printk("do_mmap:\n");
- show_process_blocks();
-#endif
+ kleave(" = %lx", result);
+ return result;
- return (unsigned long) result;
-
- error:
- up_write(&nommu_vma_sem);
- kfree(vml);
+error_put_region:
+ __put_nommu_region(region);
if (vma) {
if (vma->vm_file) {
fput(vma->vm_file);
if (vma->vm_flags & VM_EXECUTABLE)
removed_exe_file_vma(vma->vm_mm);
}
- kfree(vma);
+ kmem_cache_free(vm_area_cachep, vma);
}
+ kleave(" = %d [pr]", ret);
return ret;
- sharing_violation:
- up_write(&nommu_vma_sem);
- printk("Attempt to share mismatched mappings\n");
- kfree(vml);
- return -EINVAL;
+error_just_free:
+ up_write(&nommu_region_sem);
+error:
+ fput(region->vm_file);
+ kmem_cache_free(vm_region_jar, region);
+ fput(vma->vm_file);
+ if (vma->vm_flags & VM_EXECUTABLE)
+ removed_exe_file_vma(vma->vm_mm);
+ kmem_cache_free(vm_area_cachep, vma);
+ kleave(" = %d", ret);
+ return ret;
+
+sharing_violation:
+ up_write(&nommu_region_sem);
+ printk(KERN_WARNING "Attempt to share mismatched mappings\n");
+ ret = -EINVAL;
+ goto error;
- error_getting_vma:
- up_write(&nommu_vma_sem);
- kfree(vml);
- printk("Allocation of vma for %lu byte allocation from process %d failed\n",
+error_getting_vma:
+ kmem_cache_free(vm_region_jar, region);
+ printk(KERN_WARNING "Allocation of vma for %lu byte allocation"
+ " from process %d failed\n",
len, current->pid);
show_free_areas();
return -ENOMEM;
- error_getting_vml:
- printk("Allocation of vml for %lu byte allocation from process %d failed\n",
+error_getting_region:
+ printk(KERN_WARNING "Allocation of vm region for %lu byte allocation"
+ " from process %d failed\n",
len, current->pid);
show_free_areas();
return -ENOMEM;
@@ -1097,77 +1362,180 @@ unsigned long do_mmap_pgoff(struct file *file,
EXPORT_SYMBOL(do_mmap_pgoff);
/*
- * handle mapping disposal for uClinux
+ * split a vma into two pieces at address 'addr', a new vma is allocated either
+ * for the first part or the tail.
*/
-static void put_vma(struct mm_struct *mm, struct vm_area_struct *vma)
+int split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
+ unsigned long addr, int new_below)
{
- if (vma) {
- down_write(&nommu_vma_sem);
+ struct vm_area_struct *new;
+ struct vm_region *region;
+ unsigned long npages;
- if (atomic_dec_and_test(&vma->vm_usage)) {
- delete_nommu_vma(vma);
+ kenter("");
- if (vma->vm_ops && vma->vm_ops->close)
- vma->vm_ops->close(vma);
+ /* we're only permitted to split anonymous regions that have a single
+ * owner */
+ if (vma->vm_file ||
+ atomic_read(&vma->vm_region->vm_usage) != 1)
+ return -ENOMEM;
- /* IO memory and memory shared directly out of the pagecache from
- * ramfs/tmpfs mustn't be released here */
- if (vma->vm_flags & VM_MAPPED_COPY)
- kfree((void *) vma->vm_start);
+ if (mm->map_count >= sysctl_max_map_count)
+ return -ENOMEM;
- if (vma->vm_file) {
- fput(vma->vm_file);
- if (vma->vm_flags & VM_EXECUTABLE)
- removed_exe_file_vma(mm);
- }
- kfree(vma);
- }
+ region = kmem_cache_alloc(vm_region_jar, GFP_KERNEL);
+ if (!region)
+ return -ENOMEM;
+
+ new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
+ if (!new) {
+ kmem_cache_free(vm_region_jar, region);
+ return -ENOMEM;
+ }
+
+ /* most fields are the same, copy all, and then fixup */
+ *new = *vma;
+ *region = *vma->vm_region;
+ new->vm_region = region;
+
+ npages = (addr - vma->vm_start) >> PAGE_SHIFT;
+
+ if (new_below) {
+ region->vm_end = new->vm_end = addr;
+ } else {
+ region->vm_start = new->vm_start = addr;
+ region->vm_pgoff = new->vm_pgoff += npages;
+ }
- up_write(&nommu_vma_sem);
+ if (new->vm_ops && new->vm_ops->open)
+ new->vm_ops->open(new);
+
+ delete_vma_from_mm(vma);
+ down_write(&nommu_region_sem);
+ delete_nommu_region(vma->vm_region);
+ if (new_below) {
+ vma->vm_region->vm_start = vma->vm_start = addr;
+ vma->vm_region->vm_pgoff = vma->vm_pgoff += npages;
+ } else {
+ vma->vm_region->vm_end = vma->vm_end = addr;
}
+ add_nommu_region(vma->vm_region);
+ add_nommu_region(new->vm_region);
+ up_write(&nommu_region_sem);
+ add_vma_to_mm(mm, vma);
+ add_vma_to_mm(mm, new);
+ return 0;
}
/*
- * release a mapping
- * - under NOMMU conditions the parameters must match exactly to the mapping to
- * be removed
+ * shrink a VMA by removing the specified chunk from either the beginning or
+ * the end
*/
-int do_munmap(struct mm_struct *mm, unsigned long addr, size_t len)
+static int shrink_vma(struct mm_struct *mm,
+ struct vm_area_struct *vma,
+ unsigned long from, unsigned long to)
{
- struct vm_list_struct *vml, **parent;
- unsigned long end = addr + len;
+ struct vm_region *region;
-#ifdef DEBUG
- printk("do_munmap:\n");
-#endif
+ kenter("");
- for (parent = &mm->context.vmlist; *parent; parent = &(*parent)->next) {
- if ((*parent)->vma->vm_start > addr)
- break;
- if ((*parent)->vma->vm_start == addr &&
- ((len == 0) || ((*parent)->vma->vm_end == end)))
- goto found;
- }
+ /* adjust the VMA's pointers, which may reposition it in the MM's tree
+ * and list */
+ delete_vma_from_mm(vma);
+ if (from > vma->vm_start)
+ vma->vm_end = from;
+ else
+ vma->vm_start = to;
+ add_vma_to_mm(mm, vma);
- printk("munmap of non-mmaped memory by process %d (%s): %p\n",
- current->pid, current->comm, (void *) addr);
- return -EINVAL;
+ /* cut the backing region down to size */
+ region = vma->vm_region;
+ BUG_ON(atomic_read(®ion->vm_usage) != 1);
- found:
- vml = *parent;
+ down_write(&nommu_region_sem);
+ delete_nommu_region(region);
+ if (from > region->vm_start)
+ region->vm_end = from;
+ else
+ region->vm_start = to;
+ add_nommu_region(region);
+ up_write(&nommu_region_sem);
- put_vma(mm, vml->vma);
+ free_page_series(from, to);
+ return 0;
+}
- *parent = vml->next;
- kfree(vml);
+/*
+ * release a mapping
+ * - under NOMMU conditions the chunk to be unmapped must be backed by a single
+ * VMA, though it need not cover the whole VMA
+ */
+int do_munmap(struct mm_struct *mm, unsigned long start, size_t len)
+{
+ struct vm_area_struct *vma;
+ struct rb_node *rb;
+ unsigned long end = start + len;
+ int ret;
- update_hiwater_vm(mm);
- mm->total_vm -= len >> PAGE_SHIFT;
+ kenter(",%lx,%zx", start, len);
-#ifdef DEBUG
- show_process_blocks();
-#endif
+ if (len == 0)
+ return -EINVAL;
+
+ /* find the first potentially overlapping VMA */
+ vma = find_vma(mm, start);
+ if (!vma) {
+ printk(KERN_WARNING
+ "munmap of memory not mmapped by process %d (%s):"
+ " 0x%lx-0x%lx\n",
+ current->pid, current->comm, start, start + len - 1);
+ return -EINVAL;
+ }
+ /* we're allowed to split an anonymous VMA but not a file-backed one */
+ if (vma->vm_file) {
+ do {
+ if (start > vma->vm_start) {
+ kleave(" = -EINVAL [miss]");
+ return -EINVAL;
+ }
+ if (end == vma->vm_end)
+ goto erase_whole_vma;
+ rb = rb_next(&vma->vm_rb);
+ vma = rb_entry(rb, struct vm_area_struct, vm_rb);
+ } while (rb);
+ kleave(" = -EINVAL [split file]");
+ return -EINVAL;
+ } else {
+ /* the chunk must be a subset of the VMA found */
+ if (start == vma->vm_start && end == vma->vm_end)
+ goto erase_whole_vma;
+ if (start < vma->vm_start || end > vma->vm_end) {
+ kleave(" = -EINVAL [superset]");
+ return -EINVAL;
+ }
+ if (start & ~PAGE_MASK) {
+ kleave(" = -EINVAL [unaligned start]");
+ return -EINVAL;
+ }
+ if (end != vma->vm_end && end & ~PAGE_MASK) {
+ kleave(" = -EINVAL [unaligned split]");
+ return -EINVAL;
+ }
+ if (start != vma->vm_start && end != vma->vm_end) {
+ ret = split_vma(mm, vma, start, 1);
+ if (ret < 0) {
+ kleave(" = %d [split]", ret);
+ return ret;
+ }
+ }
+ return shrink_vma(mm, vma, start, end);
+ }
+
+erase_whole_vma:
+ delete_vma_from_mm(vma);
+ delete_vma(mm, vma);
+ kleave(" = 0");
return 0;
}
EXPORT_SYMBOL(do_munmap);
@@ -1184,29 +1552,26 @@ asmlinkage long sys_munmap(unsigned long addr, size_t len)
}
/*
- * Release all mappings
+ * release all the mappings made in a process's VM space
*/
-void exit_mmap(struct mm_struct * mm)
+void exit_mmap(struct mm_struct *mm)
{
- struct vm_list_struct *tmp;
+ struct vm_area_struct *vma;
- if (mm) {
-#ifdef DEBUG
- printk("Exit_mmap:\n");
-#endif
+ if (!mm)
+ return;
- mm->total_vm = 0;
+ kenter("");
- while ((tmp = mm->context.vmlist)) {
- mm->context.vmlist = tmp->next;
- put_vma(mm, tmp->vma);
- kfree(tmp);
- }
+ mm->total_vm = 0;
-#ifdef DEBUG
- show_process_blocks();
-#endif
+ while ((vma = mm->mmap)) {
+ mm->mmap = vma->vm_next;
+ delete_vma_from_mm(vma);
+ delete_vma(mm, vma);
}
+
+ kleave("");
}
unsigned long do_brk(unsigned long addr, unsigned long len)
@@ -1219,8 +1584,8 @@ unsigned long do_brk(unsigned long addr, unsigned long len)
* time (controlled by the MREMAP_MAYMOVE flag and available VM space)
*
* under NOMMU conditions, we only permit changing a mapping's size, and only
- * as long as it stays within the hole allocated by the kmalloc() call in
- * do_mmap_pgoff() and the block is not shareable
+ * as long as it stays within the region allocated by do_mmap_private() and the
+ * block is not shareable
*
* MREMAP_FIXED is not supported under NOMMU conditions
*/
@@ -1231,13 +1596,16 @@ unsigned long do_mremap(unsigned long addr,
struct vm_area_struct *vma;
/* insanity checks first */
- if (new_len == 0)
+ if (old_len == 0 || new_len == 0)
return (unsigned long) -EINVAL;
+ if (addr & ~PAGE_MASK)
+ return -EINVAL;
+
if (flags & MREMAP_FIXED && new_addr != addr)
return (unsigned long) -EINVAL;
- vma = find_vma_exact(current->mm, addr);
+ vma = find_vma_exact(current->mm, addr, old_len);
if (!vma)
return (unsigned long) -EINVAL;
@@ -1247,19 +1615,19 @@ unsigned long do_mremap(unsigned long addr,
if (vma->vm_flags & VM_MAYSHARE)
return (unsigned long) -EPERM;
- if (new_len > kobjsize((void *) addr))
+ if (new_len > vma->vm_region->vm_end - vma->vm_region->vm_start)
return (unsigned long) -ENOMEM;
/* all checks complete - do it */
vma->vm_end = vma->vm_start + new_len;
-
return vma->vm_start;
}
EXPORT_SYMBOL(do_mremap);
-asmlinkage unsigned long sys_mremap(unsigned long addr,
- unsigned long old_len, unsigned long new_len,
- unsigned long flags, unsigned long new_addr)
+asmlinkage
+unsigned long sys_mremap(unsigned long addr,
+ unsigned long old_len, unsigned long new_len,
+ unsigned long flags, unsigned long new_addr)
{
unsigned long ret;
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [PATCH 05/10] NOMMU: Make mmap allocation page trimming behaviour configurable. [ver #3]
2009-01-08 12:53 [PATCH 00/10] Alter NOMMU mmap handling [ver #3] David Howells
` (3 preceding siblings ...)
2009-01-08 12:54 ` [PATCH 04/10] NOMMU: Make VMAs per MM as for MMU-mode linux " David Howells
@ 2009-01-08 12:54 ` David Howells
2009-01-08 12:54 ` [PATCH 06/10] NOMMU: Improve procfs output using per-MM VMAs " David Howells
` (4 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: David Howells @ 2009-01-08 12:54 UTC (permalink / raw)
To: torvalds, akpm
Cc: bernds_cb1, dhowells, rgetz, vapier.adi, lethal, linux-kernel
From: Paul Mundt <lethal@linux-sh.org>
NOMMU mmap allocates a piece of memory for an mmap that's rounded up in size to
the nearest power-of-2 number of pages. Currently it then discards the excess
pages back to the page allocator, making that memory available for use by other
things. This can, however, cause greater amount of fragmentation.
To counter this, a sysctl is added in order to fine-tune the trimming
behaviour. The default behaviour remains to trim pages aggressively, while
this can either be disabled completely or set to a higher page-granular
watermark in order to have finer-grained control.
vm region vm_top bits taken from an earlier patch by David Howells.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Mike Frysinger <vapier.adi@gmail.com>
---
Documentation/nommu-mmap.txt | 15 ++++++++++
Documentation/sysctl/vm.txt | 18 ++++++++++++
include/linux/mm_types.h | 1 +
kernel/sysctl.c | 14 +++++++++
mm/nommu.c | 65 +++++++++++++++++++++++++++---------------
5 files changed, 90 insertions(+), 23 deletions(-)
diff --git a/Documentation/nommu-mmap.txt b/Documentation/nommu-mmap.txt
index 02b89dc..b565e82 100644
--- a/Documentation/nommu-mmap.txt
+++ b/Documentation/nommu-mmap.txt
@@ -248,3 +248,18 @@ PROVIDING SHAREABLE BLOCK DEVICE SUPPORT
Provision of shared mappings on block device files is exactly the same as for
character devices. If there isn't a real device underneath, then the driver
should allocate sufficient contiguous memory to honour any supported mapping.
+
+
+=================================
+ADJUSTING PAGE TRIMMING BEHAVIOUR
+=================================
+
+NOMMU mmap automatically rounds up to the nearest power-of-2 number of pages
+when performing an allocation. This can have adverse effects on memory
+fragmentation, and as such, is left configurable. The default behaviour is to
+aggressively trim allocations and discard any excess pages back in to the page
+allocator. In order to retain finer-grained control over fragmentation, this
+behaviour can either be disabled completely, or bumped up to a higher page
+watermark where trimming begins.
+
+Page trimming behaviour is configurable via the sysctl `vm.nr_trim_pages'.
diff --git a/Documentation/sysctl/vm.txt b/Documentation/sysctl/vm.txt
index cd05994..a341507 100644
--- a/Documentation/sysctl/vm.txt
+++ b/Documentation/sysctl/vm.txt
@@ -38,6 +38,7 @@ Currently, these files are in /proc/sys/vm:
- numa_zonelist_order
- nr_hugepages
- nr_overcommit_hugepages
+- nr_trim_pages (only if CONFIG_MMU=n)
==============================================================
@@ -348,3 +349,20 @@ Change the maximum size of the hugepage pool. The maximum is
nr_hugepages + nr_overcommit_hugepages.
See Documentation/vm/hugetlbpage.txt
+
+==============================================================
+
+nr_trim_pages
+
+This is available only on NOMMU kernels.
+
+This value adjusts the excess page trimming behaviour of power-of-2 aligned
+NOMMU mmap allocations.
+
+A value of 0 disables trimming of allocations entirely, while a value of 1
+trims excess pages aggressively. Any value >= 1 acts as the watermark where
+trimming of allocations is initiated.
+
+The default value is 1.
+
+See Documentation/nommu-mmap.txt for more information.
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 1c1e0d3..92915e8 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -106,6 +106,7 @@ struct vm_region {
unsigned long vm_flags; /* VMA vm_flags */
unsigned long vm_start; /* start address of region */
unsigned long vm_end; /* region initialised to here */
+ unsigned long vm_top; /* region allocated to here */
unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
struct file *vm_file; /* the backing file or NULL */
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 92f6e5b..89d7443 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -82,6 +82,9 @@ extern int percpu_pagelist_fraction;
extern int compat_log;
extern int latencytop_enabled;
extern int sysctl_nr_open_min, sysctl_nr_open_max;
+#ifndef CONFIG_MMU
+extern int sysctl_nr_trim_pages;
+#endif
#ifdef CONFIG_RCU_TORTURE_TEST
extern int rcutorture_runnable;
#endif /* #ifdef CONFIG_RCU_TORTURE_TEST */
@@ -1102,6 +1105,17 @@ static struct ctl_table vm_table[] = {
.mode = 0644,
.proc_handler = &proc_dointvec
},
+#else
+ {
+ .ctl_name = CTL_UNNUMBERED,
+ .procname = "nr_trim_pages",
+ .data = &sysctl_nr_trim_pages,
+ .maxlen = sizeof(sysctl_nr_trim_pages),
+ .mode = 0644,
+ .proc_handler = &proc_dointvec_minmax,
+ .strategy = &sysctl_intvec,
+ .extra1 = &zero,
+ },
#endif
{
.ctl_name = VM_LAPTOP_MODE,
diff --git a/mm/nommu.c b/mm/nommu.c
index 0d363df..a6e8ccf 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -10,7 +10,7 @@
* Copyright (c) 2000-2003 David McCullough <davidm@snapgear.com>
* Copyright (c) 2000-2001 D Jeff Dionne <jeff@uClinux.org>
* Copyright (c) 2002 Greg Ungerer <gerg@snapgear.com>
- * Copyright (c) 2007 Paul Mundt <lethal@linux-sh.org>
+ * Copyright (c) 2007-2008 Paul Mundt <lethal@linux-sh.org>
*/
#include <linux/module.h>
@@ -66,6 +66,7 @@ atomic_long_t vm_committed_space = ATOMIC_LONG_INIT(0);
int sysctl_overcommit_memory = OVERCOMMIT_GUESS; /* heuristic overcommit */
int sysctl_overcommit_ratio = 50; /* default is 50% */
int sysctl_max_map_count = DEFAULT_MAX_MAP_COUNT;
+int sysctl_nr_trim_pages = 1; /* page trimming behaviour */
int heap_stack_gap = 0;
atomic_t mmap_pages_allocated;
@@ -455,6 +456,8 @@ static noinline void validate_nommu_regions(void)
last = rb_entry(lastp, struct vm_region, vm_rb);
if (unlikely(last->vm_end <= last->vm_start))
BUG();
+ if (unlikely(last->vm_top < last->vm_end))
+ BUG();
while ((p = rb_next(lastp))) {
region = rb_entry(p, struct vm_region, vm_rb);
@@ -462,7 +465,9 @@ static noinline void validate_nommu_regions(void)
if (unlikely(region->vm_end <= region->vm_start))
BUG();
- if (unlikely(region->vm_start < last->vm_end))
+ if (unlikely(region->vm_top < region->vm_end))
+ BUG();
+ if (unlikely(region->vm_start < last->vm_top))
BUG();
lastp = p;
@@ -536,7 +541,7 @@ static void free_page_series(unsigned long from, unsigned long to)
/*
* release a reference to a region
* - the caller must hold the region semaphore, which this releases
- * - the region may not have been added to the tree yet, in which case vm_end
+ * - the region may not have been added to the tree yet, in which case vm_top
* will equal vm_start
*/
static void __put_nommu_region(struct vm_region *region)
@@ -547,7 +552,7 @@ static void __put_nommu_region(struct vm_region *region)
BUG_ON(!nommu_region_tree.rb_node);
if (atomic_dec_and_test(®ion->vm_usage)) {
- if (region->vm_end > region->vm_start)
+ if (region->vm_top > region->vm_start)
delete_nommu_region(region);
up_write(&nommu_region_sem);
@@ -558,7 +563,7 @@ static void __put_nommu_region(struct vm_region *region)
* from ramfs/tmpfs mustn't be released here */
if (region->vm_flags & VM_MAPPED_COPY) {
kdebug("free series");
- free_page_series(region->vm_start, region->vm_end);
+ free_page_series(region->vm_start, region->vm_top);
}
kmem_cache_free(vm_region_jar, region);
} else {
@@ -999,6 +1004,10 @@ static int do_mmap_shared_file(struct vm_area_struct *vma)
int ret;
ret = vma->vm_file->f_op->mmap(vma->vm_file, vma);
+ if (ret == 0) {
+ vma->vm_region->vm_top = vma->vm_region->vm_end;
+ return ret;
+ }
if (ret != -ENOSYS)
return ret;
@@ -1027,11 +1036,14 @@ static int do_mmap_private(struct vm_area_struct *vma,
*/
if (vma->vm_file) {
ret = vma->vm_file->f_op->mmap(vma->vm_file, vma);
- if (ret != -ENOSYS) {
+ if (ret == 0) {
/* shouldn't return success if we're not sharing */
- BUG_ON(ret == 0 && !(vma->vm_flags & VM_MAYSHARE));
- return ret; /* success or a real error */
+ BUG_ON(!(vma->vm_flags & VM_MAYSHARE));
+ vma->vm_region->vm_top = vma->vm_region->vm_end;
+ return ret;
}
+ if (ret != -ENOSYS)
+ return ret;
/* getting an ENOSYS error indicates that direct mmap isn't
* possible (as opposed to tried but failed) so we'll try to
@@ -1051,23 +1063,25 @@ static int do_mmap_private(struct vm_area_struct *vma,
if (!pages)
goto enomem;
- /* we allocated a power-of-2 sized page set, so we need to trim off the
- * excess */
total = 1 << order;
atomic_add(total, &mmap_pages_allocated);
point = rlen >> PAGE_SHIFT;
- while (total > point) {
- order = ilog2(total - point);
- n = 1 << order;
- kdebug("shave %lu/%lu @%lu", n, total - point, total);
- atomic_sub(n, &mmap_pages_allocated);
- total -= n;
- set_page_refcounted(pages + total);
- __free_pages(pages + total, order);
+
+ /* we allocated a power-of-2 sized page set, so we may want to trim off
+ * the excess */
+ if (sysctl_nr_trim_pages && total - point >= sysctl_nr_trim_pages) {
+ while (total > point) {
+ order = ilog2(total - point);
+ n = 1 << order;
+ kdebug("shave %lu/%lu @%lu", n, total - point, total);
+ atomic_sub(n, &mmap_pages_allocated);
+ total -= n;
+ set_page_refcounted(pages + total);
+ __free_pages(pages + total, order);
+ }
}
- total = rlen >> PAGE_SHIFT;
for (point = 1; point < total; point++)
set_page_refcounted(&pages[point]);
@@ -1075,6 +1089,7 @@ static int do_mmap_private(struct vm_area_struct *vma,
region->vm_flags = vma->vm_flags |= VM_MAPPED_COPY;
region->vm_start = (unsigned long) base;
region->vm_end = region->vm_start + rlen;
+ region->vm_top = region->vm_start + (total << PAGE_SHIFT);
vma->vm_start = region->vm_start;
vma->vm_end = region->vm_start + len;
@@ -1110,6 +1125,7 @@ error_free:
free_page_series(region->vm_start, region->vm_end);
region->vm_start = vma->vm_start = 0;
region->vm_end = vma->vm_end = 0;
+ region->vm_top = 0;
return ret;
enomem:
@@ -1401,7 +1417,7 @@ int split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
npages = (addr - vma->vm_start) >> PAGE_SHIFT;
if (new_below) {
- region->vm_end = new->vm_end = addr;
+ region->vm_top = region->vm_end = new->vm_end = addr;
} else {
region->vm_start = new->vm_start = addr;
region->vm_pgoff = new->vm_pgoff += npages;
@@ -1418,6 +1434,7 @@ int split_vma(struct mm_struct *mm, struct vm_area_struct *vma,
vma->vm_region->vm_pgoff = vma->vm_pgoff += npages;
} else {
vma->vm_region->vm_end = vma->vm_end = addr;
+ vma->vm_region->vm_top = addr;
}
add_nommu_region(vma->vm_region);
add_nommu_region(new->vm_region);
@@ -1454,10 +1471,12 @@ static int shrink_vma(struct mm_struct *mm,
down_write(&nommu_region_sem);
delete_nommu_region(region);
- if (from > region->vm_start)
- region->vm_end = from;
- else
+ if (from > region->vm_start) {
+ to = region->vm_top;
+ region->vm_top = region->vm_end = from;
+ } else {
region->vm_start = to;
+ }
add_nommu_region(region);
up_write(&nommu_region_sem);
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [PATCH 06/10] NOMMU: Improve procfs output using per-MM VMAs [ver #3]
2009-01-08 12:53 [PATCH 00/10] Alter NOMMU mmap handling [ver #3] David Howells
` (4 preceding siblings ...)
2009-01-08 12:54 ` [PATCH 05/10] NOMMU: Make mmap allocation page trimming behaviour configurable. " David Howells
@ 2009-01-08 12:54 ` David Howells
2009-01-08 12:54 ` [PATCH 07/10] FDPIC: Don't attempt to expand the userspace stack to fill the space allocated " David Howells
` (3 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: David Howells @ 2009-01-08 12:54 UTC (permalink / raw)
To: torvalds, akpm
Cc: bernds_cb1, dhowells, rgetz, vapier.adi, lethal, linux-kernel
Improve procfs output using per-MM VMAs for process memory accounting.
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Mike Frysinger <vapier.adi@gmail.com>
Acked-by: Paul Mundt <lethal@linux-sh.org>
---
fs/proc/task_nommu.c | 32 ++++++++++++++++++++++----------
1 files changed, 22 insertions(+), 10 deletions(-)
diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
index ca4a48d..343ea12 100644
--- a/fs/proc/task_nommu.c
+++ b/fs/proc/task_nommu.c
@@ -16,24 +16,31 @@
void task_mem(struct seq_file *m, struct mm_struct *mm)
{
struct vm_area_struct *vma;
+ struct vm_region *region;
struct rb_node *p;
- unsigned long bytes = 0, sbytes = 0, slack = 0;
+ unsigned long bytes = 0, sbytes = 0, slack = 0, size;
down_read(&mm->mmap_sem);
for (p = rb_first(&mm->mm_rb); p; p = rb_next(p)) {
vma = rb_entry(p, struct vm_area_struct, vm_rb);
bytes += kobjsize(vma);
+
+ region = vma->vm_region;
+ if (region) {
+ size = kobjsize(region);
+ size += region->vm_end - region->vm_start;
+ } else {
+ size = vma->vm_end - vma->vm_start;
+ }
+
if (atomic_read(&mm->mm_count) > 1 ||
- vma->vm_region ||
vma->vm_flags & VM_MAYSHARE) {
- sbytes += kobjsize((void *) vma->vm_start);
- if (vma->vm_region)
- sbytes += kobjsize(vma->vm_region);
+ sbytes += size;
} else {
- bytes += kobjsize((void *) vma->vm_start);
- slack += kobjsize((void *) vma->vm_start) -
- (vma->vm_end - vma->vm_start);
+ bytes += size;
+ if (region)
+ slack = region->vm_end - vma->vm_end;
}
}
@@ -77,7 +84,7 @@ unsigned long task_vsize(struct mm_struct *mm)
down_read(&mm->mmap_sem);
for (p = rb_first(&mm->mm_rb); p; p = rb_next(p)) {
vma = rb_entry(p, struct vm_area_struct, vm_rb);
- vsize += vma->vm_region->vm_end - vma->vm_region->vm_start;
+ vsize += vma->vm_end - vma->vm_start;
}
up_read(&mm->mmap_sem);
return vsize;
@@ -87,6 +94,7 @@ int task_statm(struct mm_struct *mm, int *shared, int *text,
int *data, int *resident)
{
struct vm_area_struct *vma;
+ struct vm_region *region;
struct rb_node *p;
int size = kobjsize(mm);
@@ -94,7 +102,11 @@ int task_statm(struct mm_struct *mm, int *shared, int *text,
for (p = rb_first(&mm->mm_rb); p; p = rb_next(p)) {
vma = rb_entry(p, struct vm_area_struct, vm_rb);
size += kobjsize(vma);
- size += kobjsize((void *) vma->vm_start);
+ region = vma->vm_region;
+ if (region) {
+ size += kobjsize(region);
+ size += region->vm_end - region->vm_start;
+ }
}
size += (*text = mm->end_code - mm->start_code);
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [PATCH 07/10] FDPIC: Don't attempt to expand the userspace stack to fill the space allocated [ver #3]
2009-01-08 12:53 [PATCH 00/10] Alter NOMMU mmap handling [ver #3] David Howells
` (5 preceding siblings ...)
2009-01-08 12:54 ` [PATCH 06/10] NOMMU: Improve procfs output using per-MM VMAs " David Howells
@ 2009-01-08 12:54 ` David Howells
2009-01-08 12:54 ` [PATCH 08/10] FLAT: " David Howells
` (2 subsequent siblings)
9 siblings, 0 replies; 13+ messages in thread
From: David Howells @ 2009-01-08 12:54 UTC (permalink / raw)
To: torvalds, akpm
Cc: bernds_cb1, dhowells, rgetz, vapier.adi, lethal, linux-kernel
Stop the ELF-FDPIC binfmt from attempting to expand the userspace stack and brk
segments to fill the space actually allocated for it. The space allocated may
be rounded up by mmap(), and may be wasted.
However, finding out how much space we actually obtained uses the contentious
kobjsize() function which we'd like to get rid of as it doesn't necessarily
work for all slab allocators.
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Mike Frysinger <vapier.adi@gmail.com>
Acked-by: Paul Mundt <lethal@linux-sh.org>
---
fs/binfmt_elf_fdpic.c | 8 --------
1 files changed, 0 insertions(+), 8 deletions(-)
diff --git a/fs/binfmt_elf_fdpic.c b/fs/binfmt_elf_fdpic.c
index 22baf1b..f3e72c5 100644
--- a/fs/binfmt_elf_fdpic.c
+++ b/fs/binfmt_elf_fdpic.c
@@ -168,9 +168,6 @@ static int load_elf_fdpic_binary(struct linux_binprm *bprm,
struct elf_fdpic_params exec_params, interp_params;
struct elf_phdr *phdr;
unsigned long stack_size, entryaddr;
-#ifndef CONFIG_MMU
- unsigned long fullsize;
-#endif
#ifdef ELF_FDPIC_PLAT_INIT
unsigned long dynaddr;
#endif
@@ -390,11 +387,6 @@ static int load_elf_fdpic_binary(struct linux_binprm *bprm,
goto error_kill;
}
- /* expand the stack mapping to use up the entire allocation granule */
- fullsize = kobjsize((char *) current->mm->start_brk);
- if (!IS_ERR_VALUE(do_mremap(current->mm->start_brk, stack_size,
- fullsize, 0, 0)))
- stack_size = fullsize;
up_write(¤t->mm->mmap_sem);
current->mm->brk = current->mm->start_brk;
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [PATCH 08/10] FLAT: Don't attempt to expand the userspace stack to fill the space allocated [ver #3]
2009-01-08 12:53 [PATCH 00/10] Alter NOMMU mmap handling [ver #3] David Howells
` (6 preceding siblings ...)
2009-01-08 12:54 ` [PATCH 07/10] FDPIC: Don't attempt to expand the userspace stack to fill the space allocated " David Howells
@ 2009-01-08 12:54 ` David Howells
2009-01-08 12:54 ` [PATCH 09/10] NOMMU: Teach kobjsize() about VMA regions. " David Howells
2009-01-08 12:54 ` [PATCH 10/10] NOMMU: Support XIP on initramfs " David Howells
9 siblings, 0 replies; 13+ messages in thread
From: David Howells @ 2009-01-08 12:54 UTC (permalink / raw)
To: torvalds, akpm
Cc: bernds_cb1, dhowells, rgetz, vapier.adi, lethal, linux-kernel
Stop the FLAT binfmt from attempting to expand the userspace stack and brk
segments to fill the space actually allocated for it. The space allocated may
be rounded up by mmap(), and may be wasted.
However, finding out how much space we actually obtained uses the contentious
kobjsize() function which we'd like to get rid of as it doesn't necessarily
work for all slab allocators.
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Mike Frysinger <vapier.adi@gmail.com>
Acked-by: Paul Mundt <lethal@linux-sh.org>
---
fs/binfmt_flat.c | 34 +++++++++++-----------------------
1 files changed, 11 insertions(+), 23 deletions(-)
diff --git a/fs/binfmt_flat.c b/fs/binfmt_flat.c
index 7bbd5c6..5cebf0b 100644
--- a/fs/binfmt_flat.c
+++ b/fs/binfmt_flat.c
@@ -417,8 +417,8 @@ static int load_flat_file(struct linux_binprm * bprm,
unsigned long textpos = 0, datapos = 0, result;
unsigned long realdatastart = 0;
unsigned long text_len, data_len, bss_len, stack_len, flags;
- unsigned long len, reallen, memp = 0;
- unsigned long extra, rlim;
+ unsigned long len, memp = 0;
+ unsigned long memp_size, extra, rlim;
unsigned long *reloc = 0, *rp;
struct inode *inode;
int i, rev, relocs = 0;
@@ -543,17 +543,10 @@ static int load_flat_file(struct linux_binprm * bprm,
}
len = data_len + extra + MAX_SHARED_LIBS * sizeof(unsigned long);
+ len = PAGE_ALIGN(len);
down_write(¤t->mm->mmap_sem);
realdatastart = do_mmap(0, 0, len,
PROT_READ|PROT_WRITE|PROT_EXEC, MAP_PRIVATE, 0);
- /* Remap to use all availabe slack region space */
- if (realdatastart && (realdatastart < (unsigned long)-4096)) {
- reallen = kobjsize((void *)realdatastart);
- if (reallen > len) {
- realdatastart = do_mremap(realdatastart, len,
- reallen, MREMAP_FIXED, realdatastart);
- }
- }
up_write(¤t->mm->mmap_sem);
if (realdatastart == 0 || realdatastart >= (unsigned long)-4096) {
@@ -591,21 +584,14 @@ static int load_flat_file(struct linux_binprm * bprm,
reloc = (unsigned long *) (datapos+(ntohl(hdr->reloc_start)-text_len));
memp = realdatastart;
-
+ memp_size = len;
} else {
len = text_len + data_len + extra + MAX_SHARED_LIBS * sizeof(unsigned long);
+ len = PAGE_ALIGN(len);
down_write(¤t->mm->mmap_sem);
textpos = do_mmap(0, 0, len,
PROT_READ | PROT_EXEC | PROT_WRITE, MAP_PRIVATE, 0);
- /* Remap to use all availabe slack region space */
- if (textpos && (textpos < (unsigned long) -4096)) {
- reallen = kobjsize((void *)textpos);
- if (reallen > len) {
- textpos = do_mremap(textpos, len, reallen,
- MREMAP_FIXED, textpos);
- }
- }
up_write(¤t->mm->mmap_sem);
if (!textpos || textpos >= (unsigned long) -4096) {
@@ -622,7 +608,7 @@ static int load_flat_file(struct linux_binprm * bprm,
reloc = (unsigned long *) (textpos + ntohl(hdr->reloc_start) +
MAX_SHARED_LIBS * sizeof(unsigned long));
memp = textpos;
-
+ memp_size = len;
#ifdef CONFIG_BINFMT_ZFLAT
/*
* load it all in and treat it like a RAM load from now on
@@ -680,10 +666,12 @@ static int load_flat_file(struct linux_binprm * bprm,
* set up the brk stuff, uses any slack left in data/bss/stack
* allocation. We put the brk after the bss (between the bss
* and stack) like other platforms.
+ * Userspace code relies on the stack pointer starting out at
+ * an address right at the end of a page.
*/
current->mm->start_brk = datapos + data_len + bss_len;
current->mm->brk = (current->mm->start_brk + 3) & ~3;
- current->mm->context.end_brk = memp + kobjsize((void *) memp) - stack_len;
+ current->mm->context.end_brk = memp + memp_size - stack_len;
}
if (flags & FLAT_FLAG_KTRACE)
@@ -790,8 +778,8 @@ static int load_flat_file(struct linux_binprm * bprm,
/* zero the BSS, BRK and stack areas */
memset((void*)(datapos + data_len), 0, bss_len +
- (memp + kobjsize((void *) memp) - stack_len - /* end brk */
- libinfo->lib_list[id].start_brk) + /* start brk */
+ (memp + memp_size - stack_len - /* end brk */
+ libinfo->lib_list[id].start_brk) + /* start brk */
stack_len);
return 0;
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [PATCH 09/10] NOMMU: Teach kobjsize() about VMA regions. [ver #3]
2009-01-08 12:53 [PATCH 00/10] Alter NOMMU mmap handling [ver #3] David Howells
` (7 preceding siblings ...)
2009-01-08 12:54 ` [PATCH 08/10] FLAT: " David Howells
@ 2009-01-08 12:54 ` David Howells
2009-01-08 12:54 ` [PATCH 10/10] NOMMU: Support XIP on initramfs " David Howells
9 siblings, 0 replies; 13+ messages in thread
From: David Howells @ 2009-01-08 12:54 UTC (permalink / raw)
To: torvalds, akpm
Cc: bernds_cb1, dhowells, rgetz, vapier.adi, lethal, linux-kernel
From: Paul Mundt <lethal@linux-sh.org>
Now that we no longer use compound pages for all large allocations,
kobjsize() actively breaks things like binfmt_flat by always handing
back PAGE_SIZE for mmap'ed regions. Fix this up by looking up the
VMA region for non-compounds.
Ideally binfmt_flat wants to get rid of kobjsize() completely, but
this is an incremental step.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Mike Frysinger <vapier.adi@gmail.com>
---
mm/nommu.c | 14 ++++++++++++++
1 files changed, 14 insertions(+), 0 deletions(-)
diff --git a/mm/nommu.c b/mm/nommu.c
index a6e8ccf..60ed837 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -149,6 +149,20 @@ unsigned int kobjsize(const void *objp)
return ksize(objp);
/*
+ * If it's not a compound page, see if we have a matching VMA
+ * region. This test is intentionally done in reverse order,
+ * so if there's no VMA, we still fall through and hand back
+ * PAGE_SIZE for 0-order pages.
+ */
+ if (!PageCompound(page)) {
+ struct vm_area_struct *vma;
+
+ vma = find_vma(current->mm, (unsigned long)objp);
+ if (vma)
+ return vma->vm_end - vma->vm_start;
+ }
+
+ /*
* The ksize() function is only guaranteed to work for pointers
* returned by kmalloc(). So handle arbitrary pointers here.
*/
^ permalink raw reply related [flat|nested] 13+ messages in thread
* [PATCH 10/10] NOMMU: Support XIP on initramfs [ver #3]
2009-01-08 12:53 [PATCH 00/10] Alter NOMMU mmap handling [ver #3] David Howells
` (8 preceding siblings ...)
2009-01-08 12:54 ` [PATCH 09/10] NOMMU: Teach kobjsize() about VMA regions. " David Howells
@ 2009-01-08 12:54 ` David Howells
9 siblings, 0 replies; 13+ messages in thread
From: David Howells @ 2009-01-08 12:54 UTC (permalink / raw)
To: torvalds, akpm
Cc: bernds_cb1, dhowells, rgetz, vapier.adi, lethal, linux-kernel
Support XIP on files unpacked from the initramfs image on NOMMU systems. This
simply requires the length of the file to be preset so that the ramfs fs can
attempt to garner sufficient contiguous storage to store the file (NOMMU mmap
can only map contiguous RAM).
All the other bits to do XIP on initramfs files are present:
(1) ramfs's truncate attempts to allocate a contiguous run of pages when a
file is truncated upwards from nothing.
(2) ramfs sets BDI on its files to indicate direct mapping is possible, and
that its files can be mapped for read, write and exec.
(3) NOMMU mmap() will use the above bits to determine that it can do XIP.
Possibly this needs better controls, because it will _always_ try and do
XIP.
One disadvantage of this very simplistic approach is that sufficient space
will be allocated to store the whole file, and not just the bit that would be
XIP'd. To deal with this, though, the initramfs unpacker would have to be
able to parse the file contents.
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Paul Mundt <lethal@linux-sh.org>
---
init/initramfs.c | 1 +
1 files changed, 1 insertions(+), 0 deletions(-)
diff --git a/init/initramfs.c b/init/initramfs.c
index 4f5ba75..d9c941c 100644
--- a/init/initramfs.c
+++ b/init/initramfs.c
@@ -317,6 +317,7 @@ static int __init do_name(void)
if (wfd >= 0) {
sys_fchown(wfd, uid, gid);
sys_fchmod(wfd, mode);
+ sys_ftruncate(wfd, body_len);
vcollected = kstrdup(collected, GFP_KERNEL);
state = CopyFile;
}
^ permalink raw reply related [flat|nested] 13+ messages in thread
* Re: [PATCH 04/10] NOMMU: Make VMAs per MM as for MMU-mode linux [ver #3]
2009-01-08 12:54 ` [PATCH 04/10] NOMMU: Make VMAs per MM as for MMU-mode linux " David Howells
@ 2009-01-13 7:55 ` Andrew Morton
2009-01-13 10:51 ` David Howells
0 siblings, 1 reply; 13+ messages in thread
From: Andrew Morton @ 2009-01-13 7:55 UTC (permalink / raw)
To: David Howells
Cc: torvalds, bernds_cb1, rgetz, vapier.adi, lethal, linux-kernel
On Thu, 08 Jan 2009 12:54:07 +0000 David Howells <dhowells@redhat.com> wrote:
> Make VMAs per mm_struct as for MMU-mode linux.
>
>
> ...
>
> --- a/fs/proc/meminfo.c
> +++ b/fs/proc/meminfo.c
> @@ -74,6 +74,9 @@ static int meminfo_proc_show(struct seq_file *m, void *v)
> "LowTotal: %8lu kB\n"
> "LowFree: %8lu kB\n"
> #endif
> +#ifndef CONFIG_MMU
> + "MmapCopy: %8lu kB\n"
> +#endif
> "SwapTotal: %8lu kB\n"
> "SwapFree: %8lu kB\n"
> "Dirty: %8lu kB\n"
> @@ -116,6 +119,9 @@ static int meminfo_proc_show(struct seq_file *m, void *v)
> K(i.totalram-i.totalhigh),
> K(i.freeram-i.freehigh),
> #endif
> +#ifndef CONFIG_MMU
> + K((unsigned long) atomic_read(&mmap_pages_allocated)),
> +#endif
This is slightly wrong, as atomic_read() returns "int". It will show
weird results on 64-bit nommu machines which allocate more than 2G pages :)
> K(i.totalswap),
> K(i.freeswap),
> K(global_page_state(NR_FILE_DIRTY)),
>
> ...
>
> +static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma)
> +{
> + unsigned long ino = 0;
> + struct file *file;
> + dev_t dev = 0;
> + int flags, len;
> +
> + flags = vma->vm_flags;
> + file = vma->vm_file;
> +
> + if (file) {
> + struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
> + dev = inode->i_sb->s_dev;
> + ino = inode->i_ino;
> + }
> +
> + seq_printf(m,
> + "%08lx-%08lx %c%c%c%c %08lx %02x:%02x %lu %n",
> + vma->vm_start,
> + vma->vm_end,
> + flags & VM_READ ? 'r' : '-',
> + flags & VM_WRITE ? 'w' : '-',
> + flags & VM_EXEC ? 'x' : '-',
> + flags & VM_MAYSHARE ? flags & VM_SHARED ? 'S' : 's' : 'p',
> + vma->vm_pgoff << PAGE_SHIFT,
This will overflow at file offsets >4G?
> + MAJOR(dev), MINOR(dev), ino, &len);
> +
> + if (file) {
> + len = 25 + sizeof(void *) * 6 - len;
> + if (len < 1)
> + len = 1;
> + seq_printf(m, "%*c", len, ' ');
> + seq_path(m, &file->f_path, "");
> + }
> +
> + seq_putc(m, '\n');
> + return 0;
> +}
> +
>
> ...
>
> --- a/mm/mmap.c
> +++ b/mm/mmap.c
> @@ -2472,3 +2472,13 @@ void mm_drop_all_locks(struct mm_struct *mm)
>
> mutex_unlock(&mm_all_locks_mutex);
> }
> +
> +/*
> + * initialise the VMA slab
> + */
> +void __init mmap_init(void)
> +{
> + vm_area_cachep = kmem_cache_create("vm_area_struct",
> + sizeof(struct vm_area_struct), 0,
> + SLAB_PANIC, NULL);
We have the KMEM_CACHE() macro for this.
It was initially added because we'd been through a period of changing
the kmem_cache_create() arguments every ten minutes and we got tired of
patching zillions of callsites.
(I see this was copy-n-pasted from fork.c. Why was it moved?)
>
> ...
>
> + vm_region_jar = kmem_cache_create("vm_region_jar",
> + sizeof(struct vm_region), 0,
> + SLAB_PANIC, NULL);
> + vm_area_cachep = kmem_cache_create("vm_area_struct",
> + sizeof(struct vm_area_struct), 0,
> + SLAB_PANIC, NULL);
dittoes
> -#endif /* DEBUG */
>
> /*
> - * add a VMA into a process's mm_struct in the appropriate place in the list
> - * - should be called with mm->mmap_sem held writelocked
> + * validate the region tree
> + * - the caller must hold the region lock
> */
> -static void add_vma_to_mm(struct mm_struct *mm, struct vm_list_struct *vml)
> +#ifdef CONFIG_DEBUG_NOMMU_REGIONS
> +static noinline void validate_nommu_regions(void)
> {
> - struct vm_list_struct **ppv;
> + struct vm_region *region, *last;
> + struct rb_node *p, *lastp;
>
> - for (ppv = ¤t->mm->context.vmlist; *ppv; ppv = &(*ppv)->next)
> - if ((*ppv)->vma->vm_start > vml->vma->vm_start)
> - break;
> + lastp = rb_first(&nommu_region_tree);
> + if (!lastp)
> + return;
> +
> + last = rb_entry(lastp, struct vm_region, vm_rb);
> + if (unlikely(last->vm_end <= last->vm_start))
> + BUG();
open-coded BUG_ON()
> + while ((p = rb_next(lastp))) {
> + region = rb_entry(p, struct vm_region, vm_rb);
`region' could be local to this block, if one feels so inclined.
> + last = rb_entry(lastp, struct vm_region, vm_rb);
> +
> + if (unlikely(region->vm_end <= region->vm_start))
> + BUG();
> + if (unlikely(region->vm_start < last->vm_end))
> + BUG();
BUG_ON()
> - vml->next = *ppv;
> - *ppv = vml;
> + lastp = p;
> + }
> }
> +#else
> +#define validate_nommu_regions() do {} while(0)
There's no need to implement this in cpp...
>
> ...
>
> -static inline struct vm_area_struct *find_vma_exact(struct mm_struct *mm,
> - unsigned long addr)
> +static void free_page_series(unsigned long from, unsigned long to)
> {
> - struct vm_list_struct *vml;
> -
> - /* search the vm_start ordered list */
> - for (vml = mm->context.vmlist; vml; vml = vml->next) {
> - if (vml->vma->vm_start == addr)
> - return vml->vma;
> - if (vml->vma->vm_start > addr)
> - break;
> + for (; from < to; from += PAGE_SIZE) {
> + struct page *page = virt_to_page(from);
> +
> + kdebug("- free %lx", from);
> + atomic_dec(&mmap_pages_allocated);
This isn't counting "pages allocated". It's counting page *refcounts*.
Now, perhaps these pages will always have a refcount of 1 (why?), in
which case things happen to work out. But if so, the next line isn't
needed:
> + if (page_count(page) != 1)
should be "== 1", surely?
> + kdebug("free page %p [%d]", page, page_count(page));
> + put_page(page);
> }
> -
> - return NULL;
> }
>
> /*
> - * find a VMA in the global tree
> + * release a reference to a region
> + * - the caller must hold the region semaphore, which this releases
"for writing"
>
> ...
>
> +int do_munmap(struct mm_struct *mm, unsigned long start, size_t len)
> +{
> + struct vm_area_struct *vma;
> + struct rb_node *rb;
> + unsigned long end = start + len;
> + int ret;
>
> - update_hiwater_vm(mm);
> - mm->total_vm -= len >> PAGE_SHIFT;
> + kenter(",%lx,%zx", start, len);
>
> -#ifdef DEBUG
> - show_process_blocks();
> -#endif
> + if (len == 0)
> + return -EINVAL;
> +
> + /* find the first potentially overlapping VMA */
> + vma = find_vma(mm, start);
> + if (!vma) {
> + printk(KERN_WARNING
> + "munmap of memory not mmapped by process %d (%s):"
> + " 0x%lx-0x%lx\n",
> + current->pid, current->comm, start, start + len - 1);
Unprivileged userspace can spam logs? Perhaps not a concern on typical
nommu systems. But buggy userspace can spam logs, too, which _is_ a
problem?
> + return -EINVAL;
> + }
>
> + /* we're allowed to split an anonymous VMA but not a file-backed one */
> + if (vma->vm_file) {
> + do {
> + if (start > vma->vm_start) {
> + kleave(" = -EINVAL [miss]");
> + return -EINVAL;
> + }
> + if (end == vma->vm_end)
> + goto erase_whole_vma;
> + rb = rb_next(&vma->vm_rb);
> + vma = rb_entry(rb, struct vm_area_struct, vm_rb);
> + } while (rb);
> + kleave(" = -EINVAL [split file]");
> + return -EINVAL;
> + } else {
> + /* the chunk must be a subset of the VMA found */
> + if (start == vma->vm_start && end == vma->vm_end)
> + goto erase_whole_vma;
> + if (start < vma->vm_start || end > vma->vm_end) {
> + kleave(" = -EINVAL [superset]");
> + return -EINVAL;
> + }
> + if (start & ~PAGE_MASK) {
> + kleave(" = -EINVAL [unaligned start]");
> + return -EINVAL;
> + }
> + if (end != vma->vm_end && end & ~PAGE_MASK) {
> + kleave(" = -EINVAL [unaligned split]");
> + return -EINVAL;
> + }
> + if (start != vma->vm_start && end != vma->vm_end) {
> + ret = split_vma(mm, vma, start, 1);
> + if (ret < 0) {
> + kleave(" = %d [split]", ret);
> + return ret;
> + }
> + }
> + return shrink_vma(mm, vma, start, end);
> + }
> +
> +erase_whole_vma:
> + delete_vma_from_mm(vma);
> + delete_vma(mm, vma);
> + kleave(" = 0");
> return 0;
> }
>
> ...
>
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [PATCH 04/10] NOMMU: Make VMAs per MM as for MMU-mode linux [ver #3]
2009-01-13 7:55 ` Andrew Morton
@ 2009-01-13 10:51 ` David Howells
0 siblings, 0 replies; 13+ messages in thread
From: David Howells @ 2009-01-13 10:51 UTC (permalink / raw)
To: Andrew Morton
Cc: dhowells, torvalds, bernds_cb1, rgetz, vapier.adi, lethal,
linux-kernel
Andrew Morton <akpm@linux-foundation.org> wrote:
> > + K((unsigned long) atomic_read(&mmap_pages_allocated)),
>
> This is slightly wrong, as atomic_read() returns "int". It will show
> weird results on 64-bit nommu machines which allocate more than 2G pages :)
Okay... I'll make mmap_pages_allocated an atomic_long. That should be
exactly the same on a 32-bit system.
> > + vma->vm_pgoff << PAGE_SHIFT,
>
> This will overflow at file offsets >4G?
True. I'll cast it to unsigned long long before shifting and adjust the print
format to %08llx.
> > + vm_area_cachep = kmem_cache_create("vm_area_struct",
> > + sizeof(struct vm_area_struct), 0,
> > + SLAB_PANIC, NULL);
>
> We have the KMEM_CACHE() macro for this.
Not that fork.c seems to use it.
> (I see this was copy-n-pasted from fork.c. Why was it moved?)
Because I wanted to do something different for vm_area_struct under nommu for
a time. I'll move it back.
> open-coded BUG_ON()
All changed.
> `region' could be local to this block, if one feels so inclined.
But then you can't look in the variable with gdb.
> > +#define validate_nommu_regions() do {} while(0)
>
> There's no need to implement this in cpp...
True. There's also no need to implement it as a static inline either.
> > +static void free_page_series(unsigned long from, unsigned long to)
> > ...
> > + atomic_dec(&mmap_pages_allocated);
>
> This isn't counting "pages allocated". It's counting page *refcounts*.
> Now, perhaps these pages will always have a refcount of 1 (why?), in
> which case things happen to work out. But if so, the next line isn't
> needed:
It's actually keeping track of the number of pages we have recorded in
vm_region structs as being allocated. Once we put the pages here, we can't
track them anymore.
Furthermore, they should have a refcount of 1 at this point, unless there's a
bug in ptrace or direct I/O.
> > + if (page_count(page) != 1)
>
> should be "== 1", surely?
Why? That would be silly. I want to warn _if_ the page count isn't what I
expect, not if it is. I'll adjust the text of the warning to make it clear.
> > + * - the caller must hold the region semaphore, which this releases
>
> "for writing"
Yep.
> > + printk(KERN_WARNING
> > + "munmap of memory not mmapped by process %d (%s):"
> > + " 0x%lx-0x%lx\n",
> > + current->pid, current->comm, start, start + len - 1);
>
> Unprivileged userspace can spam logs? Perhaps not a concern on typical
> nommu systems. But buggy userspace can spam logs, too, which _is_ a
> problem?
I'll limit it to 5 printks. But this is NOMMU: if userspace wants to corrupt
the kernel, it can do so at will.
Patch attached. Now I have to go and catch a plane to Australia.
David
---
From: David Howells <dhowells@redhat.com>
Subject: [PATCH] NOMMU: Fix a number of issues with the per-MM VMA patch
Fix a number of issues with the per-MM VMA patch:
(1) Make mmap_pages_allocated an atomic_long_t, just in case this is used on
a NOMMU system with more than 2G pages. Makes no difference on a 32-bit
system.
(2) Report vma->vm_pgoff * PAGE_SIZE as a 64-bit value, not a 32-bit value,
lest it overflow.
(3) Move the allocation of the vm_area_struct slab back for fork.c.
(4) Use KMEM_CACHE() for both vm_area_struct and vm_region slabs.
(5) Use BUG_ON() rather than if () BUG().
(6) Make the default validate_nommu_regions() a static inline rather than a
#define.
(7) Make free_page_series()'s objection to pages with a refcount != 1 more
informative.
(8) Adjust the __put_nommu_region() banner comment to indicate that the
semaphore must be held for writing.
(9) Limit the number of warnings about munmaps of non-mmapped regions.
Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David Howells <dhowells@redhat.com>
---
fs/proc/meminfo.c | 2 +-
fs/proc/task_nommu.c | 4 ++--
include/linux/mm.h | 2 +-
kernel/fork.c | 1 +
mm/mmap.c | 3 ---
mm/nommu.c | 52 ++++++++++++++++++++++++--------------------------
6 files changed, 30 insertions(+), 34 deletions(-)
diff --git a/fs/proc/meminfo.c b/fs/proc/meminfo.c
index 43d2394..74ea974 100644
--- a/fs/proc/meminfo.c
+++ b/fs/proc/meminfo.c
@@ -120,7 +120,7 @@ static int meminfo_proc_show(struct seq_file *m, void *v)
K(i.freeram-i.freehigh),
#endif
#ifndef CONFIG_MMU
- K((unsigned long) atomic_read(&mmap_pages_allocated)),
+ K((unsigned long) atomic_long_read(&mmap_pages_allocated)),
#endif
K(i.totalswap),
K(i.freeswap),
diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
index 343ea12..370be0a 100644
--- a/fs/proc/task_nommu.c
+++ b/fs/proc/task_nommu.c
@@ -136,14 +136,14 @@ static int nommu_vma_show(struct seq_file *m, struct vm_area_struct *vma)
}
seq_printf(m,
- "%08lx-%08lx %c%c%c%c %08lx %02x:%02x %lu %n",
+ "%08lx-%08lx %c%c%c%c %08llx %02x:%02x %lu %n",
vma->vm_start,
vma->vm_end,
flags & VM_READ ? 'r' : '-',
flags & VM_WRITE ? 'w' : '-',
flags & VM_EXEC ? 'x' : '-',
flags & VM_MAYSHARE ? flags & VM_SHARED ? 'S' : 's' : 'p',
- vma->vm_pgoff << PAGE_SHIFT,
+ (unsigned long long) vma->vm_pgoff << PAGE_SHIFT,
MAJOR(dev), MINOR(dev), ino, &len);
if (file) {
diff --git a/include/linux/mm.h b/include/linux/mm.h
index b91a73f..d9a832c 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -1064,7 +1064,7 @@ static inline void setup_per_cpu_pageset(void) {}
#endif
/* nommu.c */
-extern atomic_t mmap_pages_allocated;
+extern atomic_long_t mmap_pages_allocated;
/* prio_tree.c */
void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
diff --git a/kernel/fork.c b/kernel/fork.c
index 1d68f12..119b6ed 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -1484,6 +1484,7 @@ void __init proc_caches_init(void)
mm_cachep = kmem_cache_create("mm_struct",
sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
+ vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
mmap_init();
}
diff --git a/mm/mmap.c b/mm/mmap.c
index 7496231..46d126c 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2478,7 +2478,4 @@ void mm_drop_all_locks(struct mm_struct *mm)
*/
void __init mmap_init(void)
{
- vm_area_cachep = kmem_cache_create("vm_area_struct",
- sizeof(struct vm_area_struct), 0,
- SLAB_PANIC, NULL);
}
diff --git a/mm/nommu.c b/mm/nommu.c
index 60ed837..427d8c4 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -69,7 +69,7 @@ int sysctl_max_map_count = DEFAULT_MAX_MAP_COUNT;
int sysctl_nr_trim_pages = 1; /* page trimming behaviour */
int heap_stack_gap = 0;
-atomic_t mmap_pages_allocated;
+atomic_long_t mmap_pages_allocated;
EXPORT_SYMBOL(mem_map);
EXPORT_SYMBOL(num_physpages);
@@ -445,12 +445,7 @@ asmlinkage unsigned long sys_brk(unsigned long brk)
*/
void __init mmap_init(void)
{
- vm_region_jar = kmem_cache_create("vm_region_jar",
- sizeof(struct vm_region), 0,
- SLAB_PANIC, NULL);
- vm_area_cachep = kmem_cache_create("vm_area_struct",
- sizeof(struct vm_area_struct), 0,
- SLAB_PANIC, NULL);
+ vm_region_jar = KMEM_CACHE(vm_region, SLAB_PANIC);
}
/*
@@ -468,27 +463,24 @@ static noinline void validate_nommu_regions(void)
return;
last = rb_entry(lastp, struct vm_region, vm_rb);
- if (unlikely(last->vm_end <= last->vm_start))
- BUG();
- if (unlikely(last->vm_top < last->vm_end))
- BUG();
+ BUG_ON(unlikely(last->vm_end <= last->vm_start));
+ BUG_ON(unlikely(last->vm_top < last->vm_end));
while ((p = rb_next(lastp))) {
region = rb_entry(p, struct vm_region, vm_rb);
last = rb_entry(lastp, struct vm_region, vm_rb);
- if (unlikely(region->vm_end <= region->vm_start))
- BUG();
- if (unlikely(region->vm_top < region->vm_end))
- BUG();
- if (unlikely(region->vm_start < last->vm_top))
- BUG();
+ BUG_ON(unlikely(region->vm_end <= region->vm_start));
+ BUG_ON(unlikely(region->vm_top < region->vm_end));
+ BUG_ON(unlikely(region->vm_start < last->vm_top));
lastp = p;
}
}
#else
-#define validate_nommu_regions() do {} while(0)
+static void validate_nommu_regions(void)
+{
+}
#endif
/*
@@ -545,16 +537,17 @@ static void free_page_series(unsigned long from, unsigned long to)
struct page *page = virt_to_page(from);
kdebug("- free %lx", from);
- atomic_dec(&mmap_pages_allocated);
+ atomic_long_dec(&mmap_pages_allocated);
if (page_count(page) != 1)
- kdebug("free page %p [%d]", page, page_count(page));
+ kdebug("free page %p: refcount not one: %d",
+ page, page_count(page));
put_page(page);
}
}
/*
* release a reference to a region
- * - the caller must hold the region semaphore, which this releases
+ * - the caller must hold the region semaphore for writing, which this releases
* - the region may not have been added to the tree yet, in which case vm_top
* will equal vm_start
*/
@@ -1078,7 +1071,7 @@ static int do_mmap_private(struct vm_area_struct *vma,
goto enomem;
total = 1 << order;
- atomic_add(total, &mmap_pages_allocated);
+ atomic_long_add(total, &mmap_pages_allocated);
point = rlen >> PAGE_SHIFT;
@@ -1089,7 +1082,7 @@ static int do_mmap_private(struct vm_area_struct *vma,
order = ilog2(total - point);
n = 1 << order;
kdebug("shave %lu/%lu @%lu", n, total - point, total);
- atomic_sub(n, &mmap_pages_allocated);
+ atomic_long_sub(n, &mmap_pages_allocated);
total -= n;
set_page_refcounted(pages + total);
__free_pages(pages + total, order);
@@ -1518,10 +1511,15 @@ int do_munmap(struct mm_struct *mm, unsigned long start, size_t len)
/* find the first potentially overlapping VMA */
vma = find_vma(mm, start);
if (!vma) {
- printk(KERN_WARNING
- "munmap of memory not mmapped by process %d (%s):"
- " 0x%lx-0x%lx\n",
- current->pid, current->comm, start, start + len - 1);
+ static int limit = 0;
+ if (limit < 5) {
+ printk(KERN_WARNING
+ "munmap of memory not mmapped by process %d"
+ " (%s): 0x%lx-0x%lx\n",
+ current->pid, current->comm,
+ start, start + len - 1);
+ limit++;
+ }
return -EINVAL;
}
^ permalink raw reply related [flat|nested] 13+ messages in thread
end of thread, other threads:[~2009-01-13 10:53 UTC | newest]
Thread overview: 13+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-01-08 12:53 [PATCH 00/10] Alter NOMMU mmap handling [ver #3] David Howells
2009-01-08 12:53 ` [PATCH 01/10] NOMMU: Fix cleanup handling in ramfs_nommu_get_umapped_area() " David Howells
2009-01-08 12:53 ` [PATCH 02/10] NOMMU: Rename ARM's struct vm_region " David Howells
2009-01-08 12:54 ` [PATCH 03/10] NOMMU: Delete askedalloc and realalloc variables " David Howells
2009-01-08 12:54 ` [PATCH 04/10] NOMMU: Make VMAs per MM as for MMU-mode linux " David Howells
2009-01-13 7:55 ` Andrew Morton
2009-01-13 10:51 ` David Howells
2009-01-08 12:54 ` [PATCH 05/10] NOMMU: Make mmap allocation page trimming behaviour configurable. " David Howells
2009-01-08 12:54 ` [PATCH 06/10] NOMMU: Improve procfs output using per-MM VMAs " David Howells
2009-01-08 12:54 ` [PATCH 07/10] FDPIC: Don't attempt to expand the userspace stack to fill the space allocated " David Howells
2009-01-08 12:54 ` [PATCH 08/10] FLAT: " David Howells
2009-01-08 12:54 ` [PATCH 09/10] NOMMU: Teach kobjsize() about VMA regions. " David Howells
2009-01-08 12:54 ` [PATCH 10/10] NOMMU: Support XIP on initramfs " David Howells
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox