vmap_pages() (4th resend)
Subsystems (3)
PATCH 1/1 vmap_pages() (4th resend)
The vmap_pages() functions allows to map an array of virtually non-continguos pages into the kernel virtual memory. The implementation is very simple and a small variation of vmalloc() - instead of allocating new pages in alloc_area_pte() uncondintionally a pointer to a page array is passed down all through vmalloc_area_pages => alloc_area_pmd => alloc_area_pte and if it is non-null no pages are allocated but the reference count on the existing ones is incremented.
The old vmalloc_area_pages is renamed to __vmap_area_pages and vmalloc_area_pages is a small wrapper around it, passing in an NULL page array. Similarly __vmalloc is renamed to vmap_pages and a small wrapper is added.
In addition I've removed th unused vmalloc_dma and cleaned up vmalloc.h a little - this could need more cleanup (and kdoc documentation for the vmalloc.c stuff), but I will do this later in an incremental patch.
This patch is needed by certain areas in XFS, i.e. lock recovery and in the directory btree code for strange mkfs flags to operate on large contingous buffers (>PAGE_SIZE).
The patch has been reviewed by akpm.
drivers/char/mem.c +3 -2
@@ -42,6 +42,9 @@#if defined(CONFIG_S390_TAPE ) && defined(CONFIG_S390_TAPE_CHAR )extern void tapechar_init (void );#endif + +extern long vwrite (char *buf,char *addr,unsigned long count); +extern long vread (char *buf,char *addr,unsigned long count);static ssize_t do_write_mem (struct file * file,void *p,unsigned long realp,const char * buf,size_t count,loff_t *ppos)
@@ -273,8 +276,6 @@return virtr + read; } -extern long vwrite (char *buf,char *addr,unsigned long count); -/* * This function writes to the *virtual* memory as seen by the kernel. */
kernel/ksyms.c +1 -0
@@ -112,6 +112,7 @@EXPORT_SYMBOL (vmalloc);EXPORT_SYMBOL (vmalloc_32);EXPORT_SYMBOL (vmalloc_to_page); +EXPORT_SYMBOL (vmap_pages);EXPORT_SYMBOL (mem_map);EXPORT_SYMBOL (remap_page_range);EXPORT_SYMBOL (max_mapnr);
include/linux/vmalloc.h +15 -16
@@ -1,10 +1,12 @@ -#ifndef __LINUX_VMALLOC_H -#define __LINUX_VMALLOC_H +#ifndef _LINUX_VMALLOC_H +#define _LINUX_VMALLOC_H#include <linux/spinlock.h> -#include <asm/pgtable.h> +struct page ; + +/* bits in vm_struct->flags */ #define VM_IOREMAP 0x00000001/* ioremap() and friends */ #define VM_ALLOC 0x00000002/* vmalloc() */
@@ -17,28 +19,25 @@struct vm_struct * next; }; -extern struct vm_struct *get_vm_area (unsigned long size,unsigned long flags); -extern void vfree (void * addr); +extern void *__vmalloc (unsigned long size,int gfp_mask,pgprot_t prot); -extern long vread (char *buf,char *addr,unsigned long count); -extern void vmfree_area_pages (unsigned long address,unsigned long size); +extern void *vmalloc (unsigned long size); +extern void *vmalloc_32 (unsigned long size); +extern void *vmap_pages (struct page **pages,unsigned long size, +int gfp_mask,pgprot_t prot); +extern void vfree (void * addr); + +extern struct vm_struct *get_vm_area (unsigned long size,unsigned long flags);extern int vmalloc_area_pages (unsigned long address,unsigned long size,int gfp_mask,pgprot_t prot); +extern void vmfree_area_pages (unsigned long address,unsigned long size);extern struct vm_struct *remove_kernel_area (void *addr);/* -* Various ways to allocate pages. -*/ - -extern void *vmalloc (unsigned long size); -extern void *vmalloc_32 (unsigned long size); - -/* * vmlist_lock is a read-write spinlock that protects vmlist * Used in mm/vmalloc.c (get_vm_area() and vfree()) and fs/proc/kcore.c. */ extern rwlock_t vmlist_lock; -extern struct vm_struct * vmlist; -#endif +#endif /* _LINUX_VMALLOC_H */
mm/vmalloc.c +60 -43
166 lines not on this page · raw message
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/
Quoted 5 lines · Christoph Hellwig
The old
vmalloc_area_pagesis renamed to__vmap_area_pagesandvmalloc_area_pagesis a small wrapper around it, passing in an NULL page array. Similarly__vmallocis renamed tovmap_pagesand a small wrapper is added.
I don't like the NULL page array.
I think vmalloc() should just allocate the pages and create the page array, and vmap_pages() should never check for NULL. Ok?
That way, you can also get rid of passing in the gfp_mask, since that is now entirely in the callers.
Linus
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/
Quoted 9 lines · Linus Torvalds
Quoted 4 lines
The old
vmalloc_area_pagesis renamed to__vmap_area_pagesandvmalloc_area_pagesis a small wrapper around it, passing in an NULL page array. Similarly__vmallocis renamed tovmap_pagesand a small wrapper is added.I don't like the NULL page array.
I think
vmalloc()should just allocate the pages and create the page array, andvmap_pages()should never check for NULL. Ok?
I looked into implementing your suggestion the last, but the problem is that we need to pass in a page array. To do so we could either allocate all the pages as high-order allocation or kmalloc() the array where we place the pages. Both doesn't seem very practical to me..
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/
Quoted 5 lines · Christoph Hellwig
I looked into implementing your suggestion the last, but the problem is that we need to pass in a page array. To do so we could either allocate all the pages as high-order allocation or
kmalloc()the array where we place the pages. Both doesn't seem very practical to me..
Hmm? Is that such a big deal? I think it's worth it for cleanliness, and kmalloc() should be plenty big enough (a standard 4kB kmalloc on x86 can cover 4MB worth of vmalloc() space, and anybody who wants to vmalloc more than than that had better have some good reason for it - and kmalloc() does actually work for much larger areas too).
So the kmalloc() approach sounds pretty trivial to me.
Linus
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/
Quoted 7 lines · Linus Torvalds
Hmm? Is that such a big deal? I think it's worth it for cleanliness, and
kmalloc()should be plenty big enough (a standard 4kB kmalloc on x86 can cover 4MB worth ofvmalloc()space, and anybody who wants to vmalloc more than than that had better have some good reason for it - andkmalloc()does actually work for much larger areas too).So the
kmalloc()approach sounds pretty trivial to me.
Yes, it sounds trivial. It just didn't get to my internal yuckiness barrier. I'll code it up anyway, it's not my kernel :)
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/
Quoted 7 lines · Linus Torvalds
Hmm? Is that such a big deal? I think it's worth it for cleanliness, and
kmalloc()should be plenty big enough (a standard 4kB kmalloc on x86 can cover 4MB worth ofvmalloc()space, and anybody who wants to vmalloc more than than that had better have some good reason for it - andkmalloc()does actually work for much larger areas too).So the
kmalloc()approach sounds pretty trivial to me.
It wasn't that trivial, but after the major overhaul I gave to it mm/vmalloc.c looks much nicer..
Patch will follow shortly.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/
Quoted 7 lines
Quoted 7 lines
Hmm? Is that such a big deal? I think it's worth it for cleanliness, and
kmalloc()should be plenty big enough (a standard 4kB kmalloc on x86 can cover 4MB worth ofvmalloc()space, and anybody who wants to vmalloc more than than that had better have some good reason for it - andkmalloc()does actually work for much larger areas too).So the
kmalloc()approach sounds pretty trivial to me.
Okay, here is the patch. vmalloc.c is changed from ground up, *vmalloc* and vfree stay, in addition we have vmap/vunmap for the mapping I want to add. Note that the patch adds excessive comments, so it looks like adding much more then it removes.
You can import this changeset into BK by piping this whole message to:
'| bk receive [path to repository]' or apply the patch as usual.
===================================================================
[email protected], 2002-08-05 21:05:30+02:00, [email protected]
VM: Rework vmalloc code to support mapping of arbitray pages
The vmalloc operation is split into two pieces: allocate the backing pages and map them into the kernel page tabels for virtually contingous access. (Same for vfree). A new set of interfaces, vmap & vunmap does only the second part and thus allows mapping arbitray pages into kernel virtual memory.
The vmalloc.c internals have been completly overhauled to support this, but the exported interfaces are unchanged.
arch/i386/mm/ioremap.c | 6 arch/sparc64/mm/modutil.c | 55 ++++++ drivers/char/mem.c | 5 include/linux/vmalloc.h | 57 +++---- kernel/ksyms.c | 2 mm/vmalloc.c | 364 +++++++++++++++++++++++++++++++--------------- 6 files changed, 335 insertions, 154 deletions
arch/i386/mm/ioremap.c +3 -3
@@ -159,7 +159,7 @@ area->phys_addr = phys_addr; addr = area->addr;if (remap_area_pages (VMALLOC_VMADDR (addr), phys_addr, size, flags)) { -vfree (addr); +vunmap (addr);return NULL ; }return (void *) (offset + (char *)addr);
@@ -215,13 +215,13 @@struct vm_struct *p;if (addr <= high_memory)return ; - p =remove_kernel_area ((void *) (PAGE_MASK & (unsigned long ) addr)); + p =remove_vm_area ((void *) (PAGE_MASK & (unsigned long ) addr));if (!p) {printk ("__iounmap: bad address %p\n" , addr);return ; } -vmfree_area_pages (VMALLOC_VMADDR (p->addr), p->size); +unmap_vm_area (p);if (p->flags && p->phys_addr <virt_to_phys (high_memory)) { change_page_attr(virt_to_page(__va(p->phys_addr)), p->size >> PAGE_SHIFT,
arch/sparc64/mm/modutil.c +48 -7
92 lines not on this page · raw message
drivers/char/mem.c +3 -2
@@ -210,6 +210,9 @@return 0 ; } +extern long vread (char *buf,char *addr,unsigned long count); +extern long vwrite (char *buf,char *addr,unsigned long count); +/* * This function reads the *virtual* memory as seen by the kernel. */
@@ -272,8 +275,6 @@ *ppos = p;return virtr + read; } - -extern long vwrite (char *buf,char *addr,unsigned long count);/* * This function writes to the *virtual* memory as seen by the kernel.
include/linux/vmalloc.h +30 -27
74 lines not on this page · raw message
kernel/ksyms.c +2 -0
@@ -109,6 +109,8 @@EXPORT_SYMBOL (__vmalloc);EXPORT_SYMBOL (vmalloc);EXPORT_SYMBOL (vmalloc_32); +EXPORT_SYMBOL (vmap); +EXPORT_SYMBOL (vunmap);EXPORT_SYMBOL (vmalloc_to_page);EXPORT_SYMBOL (mem_map);EXPORT_SYMBOL (remap_page_range);
mm/vmalloc.c +249 -115
533 lines not on this page · raw message
===================================================================
This BitKeeper patch contains the following changesets:
1.458
## Wrapped with gzip_uu ##
begin 664 bkpatch5312 M'XL(`+;,3CT``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email protected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`)B:<X.FLK?DWKK[^<7'U!23HZB=S!3(5:RK6N=]Y*%^ZQQ[#3_=O9 MU:>S\Z+/W>PJDQ-V3%!Z\&50"TN0X/TS,&5^[SZDU64?ETV7;B(Z*'V<A#T4 M1T0K>J#F.2(NT$.TQ/%D\6V>7S$\_<!P&'?9(0V(LQBH"K$W?JH.J/A3!P:9 M#J)*4M\5;'AWU#Y^I.5D_%4['(^#"F\)6]"4',JVU<C0@3`->)C5H5)Q]``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`OG20B<:E5X+)GF_#3<8Z!296&)BR9/K3^DW2(G* M#/<[email protected]&_P#0T5QUL'4G606AJ5.*C+EXD*XQ@+NP.,=5'&1:Y$MU.* M=94?%G?U!08G!;(\C);36:6/'MXOSZNDID'3WE05%=NS^R";L1)C48CTL:M; MO[=?,0$7M[))4MI>GKZYM%'FCN><U<C!J@U_BN3M[QX3HBQM(4N,9ELN13WO M&M2IY@CO[U+CI25.+=:Y25'X]Y2=]U*(((#4Z1LCL.UB[:0&&4^?Q_O12E M#=3_0[>BM(&(+_#0[?^=BU$WM:CU;7W.M_,'Z_]I9VO$8GFXRI,1=%;!=(8Q M*$NB.?TOEXTNC_:H.0?R.0@"KTW7$KGBTG3=:3QR94'^4\Y[UCG;*U!"PZON M!UNQAH8W8DYWDZ@@T7RVM%A'U\90Q^;)&OK"[JLD*,!<GM0A/2Y6A+*5Y2\6 I:XOI6VY_8(OV28ZW.5S^7S]OQKW;=+DXFIB.JG//E/\;@P-E(E8X```` `
end
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/