pgcl-2.6.0-test5-bk3-17
Subsystems (9)
This is a forward port of Hugh Dickins' patch to implement ABI- preserving large software PAGE_SIZE support, effectively "large VM blocksize". It's also been called "subpages". "pgcl" is an abbreviation for "page clustering", after the historical but different BSD notion.
This is meant to make memory management more efficient by reducing the number of objects to manage, as well as establishing more physical contiguity of memory by keeping it pieces larger than what individual ptes map. This very noticeably reduces the space requirements for mem_map[]. I hope to eventually demonstrate further advantages like larger fs blocksize support and reduced sglist length requirements.
This release features a rewrite of all the fault handling logic, done to incorporate a forward port of hugh's original fault handlers. This should perform much better than prior releases as they efficiently implement COW unsharing's trivial case and have bounded search overhead, though there is clearly a lot of room for further optimization.
Tested for basic userspace functionality on a 256MB Thinkpad T21 and a 32GB NUMA-Q with 32KB PAGE_SIZE. It does break when you push it, though it does run init scripts, most programs, and some benchmarks here.
Available from:
ftp://ftp.kernel.org/pub/linux/kernel/people/wli/vm/pgcl/
Incremental patches for all stages of the rewrite as well as cumulative diffs vs. 2.6.0-test5-bk3 and 2.6.0-test5 are also available there.
Errata:
(1) the CONFIG_PAGE_CLUSTER==0/PAGE_MMUCOUNT==1 case is nonfunctional
(2) some oopsen zwane found while running KDE
(3) some mysterious preempt imbalance(s)
(4) drivers and fs's are essentially totally unaudited, probably broken
(5) non-i386 are all broken
(6) CONFIG_DEBUG_HIGHMEM is nonfunctional
(7) CONFIG_DEBUG_PAGEALLOC is nonfunctional
(8) many, many more
TODO:
(1) merge to current
(2) sweep drivers/fs's
(3) optimize and rework kmap_atomic_sg() API to not impact CONFIG_NOHIGHMEM
(4) clean up potentially-removable code impacts (e.g. debug code)
(5) rework pagetable allocation
(6) rework/optimize rmap interaction
(7) rework TLB invalidations
(8) fix all bugs (as usual)
(9) eventually play with ia64's 32-bit emulation
-- wli
Quoted 4 lines · William Lee Irwin III
This is a forward port of Hugh Dickins' patch to implement ABI- preserving large software
PAGE_SIZEsupport, effectively "large VM blocksize". It's also been called "subpages". "pgcl" is an abbreviation for "page clustering", after the historical but different BSD notion.
Now also ported to 2.6.0-test11:
ftp://ftp.kernel.org/pub/linux/kernel/people/wli/vm/pgcl/pgcl-2.6.0-test11-1.gz
This also corrects some PAGE_SHIFT instances that crept into mm/mmap.c while I wasn't looking and drops sym2 driver changes.
-- wli
Quoted 4 lines · William Lee Irwin III
Quoted 4 lines
This is a forward port of Hugh Dickins' patch to implement ABI- preserving large software
PAGE_SIZEsupport, effectively "large VM blocksize". It's also been called "subpages". "pgcl" is an abbreviation for "page clustering", after the historical but different BSD notion.
Quoted 4 lines · William Lee Irwin III
Now also ported to 2.6.0-test11:
ftp://ftp.kernel.org/pub/linux/kernel/people/wli/vm/pgcl/pgcl-2.6.0-test11-1.gz This also corrects somePAGE_SHIFTinstances that crept intomm/mmap.cwhile I wasn't looking and drops sym2 driver changes.
Looks like I missed the update to the Committed_AS accounting in exec.c during the conversion. The reduced granularity on RLIMIT_STACK, stack fault alignment, and the argument size Committed_AS accounting bothers me. I should teach copy_folio() etc. to cope with unaligned pages, but there appears to be some pain wrt. how far and wide to hunt for ptes (i.e. this requires some design effort).
I do the swap bits a bit differently from the original, so the swapfile accounting needs a unit conversion.
This also arranges for pr_debug()-based logging of Committed_AS accounting, which is useful while this thing is still prototypical in nature, but will eventually get removed for "production" patch posting along with several other things of a similar nature.
-- wli
fs/exec.c +1 -1
@@ -417,7 +417,7 @@ int setup_arg_pages(struct linux_binprm#else stack_base =STACK_TOP -MAX_ARG_PAGES *PAGE_SIZE ; mm->arg_start = bprm->p + stack_base; - arg_size =STACK_TOP - (MMUPAGE_MASK & (unsigned long ) mm->arg_start); + arg_size =STACK_TOP - (PAGE_MASK & (unsigned long ) mm->arg_start);#endif bprm->p += stack_base;
include/linux/mman.h +28 -4
@@ -14,18 +14,42 @@ extern int sysctl_overcommit_memory;extern int sysctl_overcommit_ratio;extern atomic_t vm_committed_space; +#define vm_acct_memory (pages) \ +do { \ + long __pages__ = (pages); \ + pr_debug("%d: vm_acct_memory(%ld) in %s at %s:%d\n", \ + current->pid, \ + __pages__, \ + __FUNCTION__, \ + __FILE__, \ + __LINE__); \ + __vm_acct_memory(__pages__); \ +} while (0) + +#define vm_unacct_memory (pages) \ +do { \ + long __pages__ = (pages); \ + pr_debug("%d: vm_unacct_memory(%ld) in %s at %s:%d\n", \ + current->pid, \ + __pages__, \ + __FUNCTION__, \ + __FILE__, \ + __LINE__); \ + __vm_unacct_memory(__pages__); \ +} while (0) +#ifdef CONFIG_SMP -extern void vm_acct_memory (long pages); +void __vm_acct_memory (long pages);#else -static inline void vm_acct_memory (long pages) +static inline void __vm_acct_memory (long pages) {atomic_add (pages, &vm_committed_space); }#endif -static inline void vm_unacct_memory (long pages) +static inline void __vm_unacct_memory (long pages) { -vm_acct_memory (-pages); +__vm_acct_memory (-pages); }/*
include/linux/security.h +14 -2
@@ -84,6 +84,18 @@ struct nfsctl_arg;struct sched_param ;struct swap_info_struct ; +#define security_vm_enough_memory (pages) \ +({ \ + long __pages__ = (pages); \ + pr_debug("%d: vm_enough_memory(%ld) in %s at %s:%d\n", \ + current->pid, \ + __pages__, \ + __FUNCTION__, \ + __FILE__, \ + __LINE__); \ + __security_vm_enough_memory(__pages__); \ +}) +#ifdef CONFIG_SECURITY /**
@@ -1245,7 +1257,7 @@ static inline int security_syslog(int tyreturn security_ops->syslog (type); } -static inline int security_vm_enough_memory (long pages) +static inline int __security_vm_enough_memory (long pages) {return security_ops->vm_enough_memory (pages); }
@@ -1911,7 +1923,7 @@ static inline int security_syslog(int tyreturn cap_syslog (type); } -static inline int security_vm_enough_memory (long pages) +static inline int __security_vm_enough_memory (long pages) {return cap_vm_enough_memory (pages); }
mm/swap.c +2 -2
@@ -367,7 +367,7 @@ unsigned int pagevec_lookup(struct pagev static DEFINE_PER_CPU(long, committed_space) = 0; -void vm_acct_memory (long pages) +void __vm_acct_memory (long pages) {long *local;
@@ -380,7 +380,7 @@ void vm_acct_memory(long pages) }preempt_enable (); } -EXPORT_SYMBOL (vm_acct_memory); +EXPORT_SYMBOL (__vm_acct_memory);#endif #ifdef CONFIG_SMP
mm/swapfile.c +2 -2
@@ -1151,8 +1151,8 @@ asmlinkage long sys_swapoff(const char _swap_list_unlock (); goto out_dput; } -if (!security_vm_enough_memory (p->pages)) -vm_unacct_memory (p->pages); +if (!security_vm_enough_memory (PAGE_MMUCOUNT *p->pages)) +vm_unacct_memory (PAGE_MMUCOUNT *p->pages);else { err = -ENOMEM ;swap_list_unlock ();
Quoted 4 lines · William Lee Irwin III
Now also ported to 2.6.0-test11:
ftp://ftp.kernel.org/pub/linux/kernel/people/wli/vm/pgcl/pgcl-2.6.0-test11-1.gz This also corrects somePAGE_SHIFTinstances that crept intomm/mmap.cwhile I wasn't looking and drops sym2 driver changes.
Quick performance and functionality test on a 32x/64GB 700MHz P-III. Compressed dmesg attached. Not sure where the SCSI error came from.
This seems to show that reworking the rmap interaction is a pressing performance issue (however pressing kernel compiles can be considered to be, never mind all the actual bugs) since the rmap functions are burning more cpu time than bitblitting luserspace. Unfortunately there isn't really enough disk for on-disk dbench etc. so I resorted to ramfs, whose readdir() etc. methods appear not to be very well parallelized. I should probably have tried ext2 on loop on ramfs or some such nonsense.
The good news is that it does still run on the thing even after all of the massive shakeups, rewrites, and time since my last boot on it in May.
Kernel compile on ext2:
$ time make -s -j bzImage > /dev/null arch/i386/boot/setup.S: Assembler messages: arch/i386/boot/setup.S:168: Warning: value 0x77dfffff truncated to 0x77dfffff Root device is (8, 2) Boot sector 512 bytes. Setup is 2538 bytes. System is 1117 kB make -s -j bzImage > /dev/null 712.96s user 284.86s system 1140% cpu 1:27.46 total
6094329 total 3.8079
5787061 default_idle 90422.8281
85912 __page_remove_rmap 298.3056
38331 page_remove_rmap 140.9228
28287 copy_folio 88.3969
19051 page_add_rmap 79.3792
14709 __copy_to_user_ll 131.3304
7078 __copy_user_intel 40.2159 6901 __copy_from_user_ll 61.6161 5704 __page_add_rmap 22.2812 4710 __d_lookup 17.3162
4259 zap_pte_range 3.6464
4127 schedule 2.5288
3940 rmap_add_folio 14.4853
3852 kmap_atomic 8.0250
3258 kmem_cache_free 33.9375
3053 clear_page_tables 3.3476
2079 try_to_wake_up 4.0605
2012 find_get_page 31.4375
1815 current_kernel_time 22.6875
1638 kmap_atomic_sg 4.0950
1627 path_lookup 4.6222
dbench on ramfs:
$ time dbench 32 32 clients started 0 62477 296.11 MB/sec Throughput 296.11 MB/sec 32 procs dbench 32 20.60s user 989.43s system 2674% cpu 37.767 total
2328901 total 1.4552
1335226 default_idle 20862.9062
415966 .text.lock.libfs 2298.1547 168210 .text.lock.dec_and_lock 15291.8182 157310 .text.lock.dcache 291.3148 70329 dcache_readdir 146.5188 23983 atomic_dec_and_lock 347.5797 21502 __copy_to_user_ll 191.9821 18555 .text.lock.file_table 157.2458 16183 __copy_from_user_ll 144.4911 11169 d_alloc 23.2688 11055 filldir64 40.6434 6661 d_instantiate 69.3854 5826 __d_lookup 21.4191 5405 path_lookup 15.3551 4545 path_release 71.0156 3728 file_move 58.2500 3537 d_delete 18.4219 3336 fd_install 52.1250 3220 simple_prepare_write 20.1250 3081 call_rcu 48.1406 2704 vfs_readdir 21.1250
SDET 32 profile:
4605683 total 2.8778
3480214 default_idle 54378.3438
212021 page_remove_rmap 779.4890 163508 __page_remove_rmap 567.7361 98710 page_add_rmap 411.2917 87944 copy_folio 274.8250 31101 __page_add_rmap 121.4883 20883 simple_prepare_write 130.5188 19818 copy_page_range 15.6788 18555 .text.lock.dcache 34.3611 18277 zap_pte_range 15.6481 17973 rmap_add_folio 66.0772
17932 .text.lock.dec_and_lock 1630.1818
17816 __d_lookup 65.5000
17680 __down 78.9286
16517 __copy_to_user_ll 147.4732
11960 atomic_dec_and_lock 173.3333
11795 release_pages 36.8594
11256 kmap_atomic 23.4500
9696 kmem_cache_free 101.0000
9622 copy_mm 9.5456
9371 .text.lock.libfs 51.7735
SDET 128 profile:
19885416 total 12.4250
10870694 default_idle 169854.5938
1868923 page_remove_rmap 6871.0404
1722495 __page_remove_rmap 5980.8854
991381 page_add_rmap 4130.7542 619755 __down 2766.7634 403192 copy_folio 1259.9750 172220 __page_add_rmap 672.7344 164246 rmap_add_folio 603.8456 161239 __wake_up 2015.4875 123047 __d_lookup 452.3787
121684 schedule 74.5613
120902 zap_pte_range 103.5120
101418 __copy_to_user_ll 905.5179
99572 copy_page_range 78.7753
95709 simple_prepare_write 598.1812 76676 atomic_dec_and_lock 1111.2464 66245 remove_shared_vm_struct 517.5391 64159 path_lookup 182.2699 61922 release_pages 193.5062
61415 copy_mm 60.9276
61356 clear_page_tables 67.2763
I could probably use suggestions for a more mergeable method of mitigating the rmap functions' overheads than the currently available mm/rmap.c rewrites.
In more relevant news, I've started taking stabs at the rvmalloc() mess and drm, with copious help from Zwane.
-- wli
Quoted 4 lines · William Lee Irwin III
Now also ported to 2.6.0-test11:
ftp://ftp.kernel.org/pub/linux/kernel/people/wli/vm/pgcl/pgcl-2.6.0-test11-1.gz This also corrects somePAGE_SHIFTinstances that crept intomm/mmap.cwhile I wasn't looking and drops sym2 driver changes.
I wonder if this would be enough to get sysenter support going again. I've not got a sysenter-capable userspace around, so I can't really test this myself.
vs. pgcl-2.6.0-test11-5
-- wli
include/asm-i386/elf.h +2 -10
@@ -132,17 +132,11 @@ extern int dump_task_extended_fpu (struc#define VSYSCALL_ENTRY ((unsigned long) &__kernel_vsyscall)extern void __kernel_vsyscall; -/* -* Something in pgcl broke vsyscalls. Until that's tracked down, -* work around it with this: -*/ -#if 0 #define ARCH_DLINFO \ do { \ NEW_AUX_ENT(AT_SYSINFO, VSYSCALL_ENTRY); \ NEW_AUX_ENT(AT_SYSINFO_EHDR, VSYSCALL_BASE); \ } while (0) -#endif /* 0 */ /* * These macros parameterize elf_core_dump in fs/binfmt_elf.c to write out
@@ -152,7 +146,6 @@ do { \* Dumping its extra ELF program headers includes all the other information * a debugger needs to easily find how the vsyscall DSO was being used. */ -#if 0 #define ELF_CORE_EXTRA_PHDRS (VSYSCALL_EHDR->e_phnum)#define ELF_CORE_WRITE_EXTRA_PHDRS \ do { \
@@ -166,7 +159,7 @@ do { \if (phdr.p_type ==PT_LOAD ) { \BUG_ON (ofs !=0 ); \ ofs = phdr.p_offset = offset; \ - phdr.p_memsz =PAGE_ALIGN (phdr.p_memsz); \ + phdr.p_memsz =MMUPAGE_ALIGN (phdr.p_memsz); \ phdr.p_filesz = phdr.p_memsz; \ offset += phdr.p_filesz; \ } \
@@ -185,10 +178,9 @@ do { \for (i =0 ; i <VSYSCALL_EHDR ->e_phnum; ++i) { \if (vsyscall_phdrs[i].p_type ==PT_LOAD ) \DUMP_WRITE ((void *) vsyscall_phdrs[i].p_vaddr, \ -PAGE_ALIGN (vsyscall_phdrs[i].p_memsz)); \ +MMUPAGE_ALIGN (vsyscall_phdrs[i].p_memsz)); \ } \ } while (0) -#endif /* 0 */ #endif /* __KERNEL__ */
include/asm-i386/fixmap.h +15 -8
@@ -21,6 +21,11 @@#ifdef CONFIG_HIGHMEM #include <linux/threads.h> #include <asm/kmap_types.h> +#ifdef CONFIG_HIGHMEM4G +#include <asm/pgtable-2level.h> +#else +#include <asm/pgtable-3level.h> +#endif #endif /*
@@ -49,6 +54,7 @@#define PKMAP_ADDR (nr) (PKMAP_BASE + ((nr) << PAGE_SHIFT))#define LAST_PKMAP 1024#define LAST_PKMAP_MASK (LAST_PKMAP-1) +#define PKMAP_MASK (~((1 << PMD_SHIFT) - 1))/* * FIXADDR stuff is used by highmem.c for kmapping, and various
@@ -62,7 +68,7 @@* and I didn't care enough to conserve PAGE_SIZE - MMUPAGE_SIZE * worth of virtualspace. */ -#define FIXADDR_TOP (-PAGE_SIZE) +#define FIXADDR_TOP (-MMUPAGE_SIZE)#define __FIXADDR_SIZE (__end_of_permanent_fixed_addresses << MMUPAGE_SHIFT)#define FIXADDR_START (FIXADDR_TOP - __FIXADDR_SIZE)
@@ -70,13 +76,8 @@#define __virt_to_fix (x) ((FIXADDR_TOP - ((x) & MMUPAGE_MASK)) >> MMUPAGE_SHIFT)enum fixed_addresses { -#ifdef CONFIG_HIGHMEM /* reserved pte's for temporary kernel mappings */ -FIX_KMAP_BEGIN =1 , -FIX_KMAP_END =FIX_KMAP_BEGIN +((KM_TYPE_NR *NR_CPUS +1 )*PAGE_MMUCOUNT )-1 , -FIX_PKMAP_BEGIN , -FIX_PKMAP_END =FIX_PKMAP_BEGIN + (LAST_PKMAP +1 )*PAGE_MMUCOUNT -1 , -#endif +FIX_HOLE ,FIX_VSYSCALL ,#ifdef CONFIG_X86_LOCAL_APIC FIX_APIC_BASE ,/* local (CPU) APIC) -- required for SMP or not */
@@ -107,6 +108,12 @@ enum fixed_addresses {FIX_BTMAP_END = __end_of_permanent_fixed_addresses,FIX_BTMAP_BEGIN =FIX_BTMAP_END +NR_FIX_BTMAPS -1 ,FIX_WP_TEST , +#ifdef CONFIG_HIGHMEM +FIX_KMAP_BEGIN =__virt_to_fix (__fix_to_virt (FIX_WP_TEST +1 ) &PAGE_MASK ) -PAGE_MMUCOUNT +1 , +FIX_KMAP_END =FIX_KMAP_BEGIN +((KM_TYPE_NR *NR_CPUS +1 )*PAGE_MMUCOUNT )-1 , +FIX_PKMAP_BEGIN =__virt_to_fix (__fix_to_virt (FIX_KMAP_END +1 ) &PKMAP_MASK ) -PAGE_MMUCOUNT +1 , +FIX_PKMAP_END =FIX_PKMAP_BEGIN + (LAST_PKMAP +1 )*PAGE_MMUCOUNT -1 , +#endif __end_of_fixed_addresses };
@@ -129,7 +136,7 @@ extern void __set_fixmap (enum fixed_add* acting like user mode such as get_user_pages. */ #define FIXADDR_USER_START (__fix_to_virt(FIX_VSYSCALL)) -#define FIXADDR_USER_END (FIXADDR_USER_START + PAGE_SIZE) +#define FIXADDR_USER_END (FIXADDR_USER_START + MMUPAGE_SIZE)extern void __this_fixmap_does_not_exist (void );
Quoted 11 lines · William Lee Irwin III
Quoted 4 lines · William Lee Irwin III
Now also ported to 2.6.0-test11:
ftp://ftp.kernel.org/pub/linux/kernel/people/wli/vm/pgcl/pgcl-2.6.0-test11-1.gz This also corrects somePAGE_SHIFTinstances that crept intomm/mmap.cwhile I wasn't looking and drops sym2 driver changes.I wonder if this would be enough to get sysenter support going again. I've not got a sysenter-capable userspace around, so I can't really test this myself.
vs. pgcl-2.6.0-test11-5
Brilliant!
Linux arusha.mastecende.com 2.6.0-test11-pgcl #4 SMP Sun Nov 30 13:30:51 EST 2003 i686 i686 i386 GNU/Linux
tcsh S C0155C8D 0 1132 1131 (NOTLB)def85e60 00000086 def8b29c c0155c8d def85e44 def85e9c c060b940 c013010f
0001c1ec 00000000 c1143640 00008734 ff21d58e 00000019 dea41fd0 0000000e
fff24fc0 00000008 7fffffff c1657d7c de7f9000 c0131035 dc21bfcc c1119338
Call Trace:
[<c0155c8d>] private_folio_page+0x1d/0x2c0
[<c013010f>] __mod_timer+0x10f/0x360
[<c0131035>] schedule_timeout+0xb5/0xc0
[<c032cc53>] read_chan+0x2b3/0xc50
[<c011d1f1>] do_page_fault+0x5b1/0x871
[<c0121830>] default_wake_function+0x0/0x20
[<c032dfc1>] set_termios+0x111/0x180
[<c0121830>] default_wake_function+0x0/0x20
[<c0326c3b>] tty_read+0x15b/0x1a0
[<c0326ae0>] tty_read+0x0/0x1a0
[<c0326ae0>] tty_read+0x0/0x1a0
[<c016ac9c>] vfs_read+0xac/0xf0
[<c016aead>] sys_read+0x2d/0x50
[<c0109719>] sysenter_past_esp+0x52/0x79Quoted 7 lines · Zwane Mwaikambo
Brilliant!
Linux arusha.mastecende.com 2.6.0-test11-pgcl #4 SMP Sun Nov 30 13:30:51 EST 2003 i686 i686 i386 GNU/Linux
tcsh S C0155C8D 0 1132 1131 (NOTLB)def85e6000000086def8b29cc0155c8ddef85e44def85e9cc060b940c013010f0001c1ec 00000000 c1143640 00008734 ff21d58e 00000019 dea41fd0 0000000e fff24fc0 00000008 7fffffff c1657d7c de7f9000 c0131035 dc21bfcc c1119338 Call Trace:
[...]
Quoted 9 lines
[<c0121830>] default_wake_function+0x0/0x20 [<c032dfc1>] set_termios+0x111/0x180 [<c0121830>] default_wake_function+0x0/0x20 [<c0326c3b>] tty_read+0x15b/0x1a0 [<c0326ae0>] tty_read+0x0/0x1a0 [<c0326ae0>] tty_read+0x0/0x1a0 [<c016ac9c>] vfs_read+0xac/0xf0 [<c016aead>] sys_read+0x2d/0x50 [<c0109719>] sysenter_past_esp+0x52/0x79
I'll call this pgcl-2.6.0-test11-6, then.
-- wli
Quoted 4 lines · William Lee Irwin III
I wonder if this would be enough to get sysenter support going again. I've not got a sysenter-capable userspace around, so I can't really test this myself.
vs. pgcl-2.6.0-test11-5
Stack decoding fixes, shutting up some compiler warnings, and dumping PAGE_SIZE and MMUPAGE_SIZE into /proc/meminfo (for lack of a better place).
The printk()'s down there should eventually get ripped out anyway for minimal impact and a quieter boot, but until then...
-- wli
arch/i386/kernel/process.c +2 -2
@@ -648,12 +648,12 @@ unsigned long get_wchan(struct task_strureturn 0 ; stack_page = (unsigned long )p->thread_info; esp = p->thread.esp; -if (!stack_page || esp < stack_page || esp >8188 +stack_page) +if (!stack_page || esp < stack_page || esp >THREAD_SIZE +stack_page-4 )return 0 ;/* include/asm-i386/system.h:switch_to() pushes ebp last. */ ebp = *(unsigned long *) esp; do { - if (ebp < stack_page || ebp > 8184+stack_page) + if (ebp < stack_page || ebp > THREAD_SIZE + stack_page - 8) return 0; eip = *(unsigned long *) (ebp+4); if (eip < first_sched || eip >= last_sched)
arch/i386/mm/discontig.c +3 -3
@@ -268,10 +268,10 @@ unsigned long __init setup_memory(void)for (nid =0 ; nid < numnodes; nid++)find_max_pfn_node (nid);printk ("vmallocspace = [0x%lx, 0x%lx)\n" , -VMALLOC_START ,VMALLOC_END ); + (unsigned long )VMALLOC_START , (unsigned long )VMALLOC_END );printk ("fixmapspace = [0x%lx, 0x%lx)\n" , -FIXADDR_START ,FIXADDR_TOP ); -printk ("MAXMEM = 0x%lx\n" ,MAXMEM ); + (unsigned long )FIXADDR_START , (unsigned long )FIXADDR_TOP ); +printk ("MAXMEM = 0x%lx\n" , (unsigned long )MAXMEM );for (nid =0 ; nid < numnodes; ++nid)printk ("node %d at pfns [0x%lx, 0x%lx)\n" , nid, node_start_pfn[nid], node_end_pfn[nid]);
arch/i386/mm/init.c +6 -6
@@ -514,24 +514,24 @@ void __init mem_init(void) );printk ("MAXMEM=0x%lx\n" ,MAXMEM );printk ("vmalloc: start = 0x%lx, end = 0x%lx\n" , -VMALLOC_START ,VMALLOC_END ); + (unsigned long )VMALLOC_START , (unsigned long )VMALLOC_END );printk ("fixaddr: start = 0x%lx, end = 0x%lx\n" , -FIXADDR_START ,FIXADDR_TOP ); + (unsigned long )FIXADDR_START , (unsigned long )FIXADDR_TOP );#ifdef CONFIG_HIGHMEM -printk ("FIX_KMAP_END == %lx\n" ,__fix_to_virt (FIX_KMAP_END )); +printk ("FIX_KMAP_END == %lx\n" , (unsigned long )__fix_to_virt (FIX_KMAP_END ));if (__fix_to_virt (FIX_KMAP_END ) %PAGE_SIZE )printk (KERN_CRIT "kmap_atomic() area misaligned!\n" ); -printk ("FIX_KMAP_BEGIN == %lx\n" ,__fix_to_virt (FIX_KMAP_BEGIN )); +printk ("FIX_KMAP_BEGIN == %lx\n" , (unsigned long )__fix_to_virt (FIX_KMAP_BEGIN ));if ((__fix_to_virt (FIX_KMAP_BEGIN ) +MMUPAGE_SIZE ) %PAGE_SIZE )printk (KERN_CRIT "kmap_atomic() area misaligned!\n" ); -printk ("FIX_PKMAP_END == %lx\n" ,__fix_to_virt (FIX_PKMAP_END )); +printk ("FIX_PKMAP_END == %lx\n" , (unsigned long )__fix_to_virt (FIX_PKMAP_END ));if (__fix_to_virt (FIX_PKMAP_END ) %PAGE_SIZE )printk (KERN_CRIT "kmap() area misaligned!\n" ); -printk ("FIX_PKMAP_BEGIN == %lx\n" ,__fix_to_virt (FIX_PKMAP_BEGIN )); +printk ("FIX_PKMAP_BEGIN == %lx\n" , (unsigned long )__fix_to_virt (FIX_PKMAP_BEGIN ));if ((__fix_to_virt (FIX_PKMAP_BEGIN ) +MMUPAGE_SIZE ) %PAGE_SIZE )printk (KERN_CRIT "kmap() area misaligned!\n" );#endif
fs/proc/proc_misc.c +6 -2
@@ -206,7 +206,9 @@ static int meminfo_read_proc(char *page,"PageTables: %8lu kB\n" "VmallocTotal: %8lu kB\n" "VmallocUsed: %8lu kB\n" -"VmallocChunk: %8lu kB\n" , +"VmallocChunk: %8lu kB\n" +"PAGE_SIZE: %8lu kB\n" +"MMUPAGE_SIZE: %8d kB\n" ,K (i.totalram),K (i.freeram), K(i.bufferram),
@@ -228,7 +230,9 @@ static int meminfo_read_proc(char *page,K (ps.nr_page_table_pages), vmtot, vmi.used, - vmi.largest_chunk + vmi.largest_chunk, +K (PAGE_SIZE ), +K (MMUPAGE_SIZE ) ); len +=hugetlb_report_meminfo (page + len);
include/asm-i386/processor.h +10 -2
@@ -495,8 +495,16 @@ extern unsigned long thread_saved_pc(strvoid show_trace (struct task_struct *task,unsigned long *stack);unsigned long get_wchan (struct task_struct *p); -#define KSTK_EIP (tsk) (((unsigned long *)(4096+(unsigned long)(tsk)->thread_info))[1019]) -#define KSTK_ESP (tsk) (((unsigned long *)(4096+(unsigned long)(tsk)->thread_info))[1022]) + +#define THREAD_SIZE_LONGS (THREAD_SIZE/sizeof(unsigned long)) +#define task_pt_regs (task) \ +({ \ + unsigned long *__ptr__ = (unsigned long *)(task)->thread_info; \ + (struct pt_regs *)(&__ptr__[THREAD_SIZE_LONGS-1]); \ +}) + +#define KSTK_EIP (task) (task_pt_regs(task)->eip) +#define KSTK_ESP (task) (task_pt_regs(task)->esp)struct microcode_header {unsigned int hdrver;
Quoted 4 lines · William Lee Irwin III
Stack decoding fixes, shutting up some compiler warnings, and dumping
PAGE_SIZEandMMUPAGE_SIZEinto/proc/meminfo(for lack of a better place). The printk()'s down there should eventually get ripped out anyway for minimal impact and a quieter boot, but until then...
This fixes some bad address calculations in arch/i386/mm/boot_ioremap.c
-- wli
arch/i386/mm/boot_ioremap.c +10 -10
@@ -31,7 +31,7 @@#define BOOT_PTE_PTRS (PTRS_PER_PTE*2)#define boot_pte_index (address) \ - (((address) >> PAGE_SHIFT) & (BOOT_PTE_PTRS - 1)) + (((address) >> MMUPAGE_SHIFT) & (BOOT_PTE_PTRS - 1))static inline boot_pte_t *boot_vaddr_to_pte (void *address) {
@@ -52,17 +52,17 @@ static void __boot_ioremap(unsigned longchar *vaddr = virtual_source; pte =boot_vaddr_to_pte (virtual_source); -for (i=0 ; i < nrpages; i++, phys_addr +=PAGE_SIZE , pte++) { -set_pte (pte,pfn_pte (phys_addr>>PAGE_SHIFT ,PAGE_KERNEL )); -__flush_tlb_one (&vaddr[i*PAGE_SIZE ]); +for (i=0 ; i < nrpages; i++, phys_addr +=MMUPAGE_SIZE , pte++) { +set_pte (pte,pfn_pte (phys_addr>>MMUPAGE_SHIFT ,PAGE_KERNEL )); +__flush_tlb_one (&vaddr[i*MMUPAGE_SIZE ]); } }/* the virtual space we're going to remap comes from this array */ #define BOOT_IOREMAP_PAGES 4 -#define BOOT_IOREMAP_SIZE (BOOT_IOREMAP_PAGES*PAGE_SIZE) +#define BOOT_IOREMAP_SIZE (BOOT_IOREMAP_PAGES*MMUPAGE_SIZE) __initdata char boot_ioremap_space[BOOT_IOREMAP_SIZE] - __attribute__ ((aligned (PAGE_SIZE))); + __attribute__ ((aligned (MMUPAGE_SIZE)));/* * This only applies to things which need to ioremap before paging_init()
@@ -83,11 +83,11 @@ __init void* boot_ioremap(unsigned long last_addr = phys_addr + size -1 ;/* page align the requested address */ - offset = phys_addr & ~PAGE_MASK ; - phys_addr &=PAGE_MASK ; - size =PAGE_ALIGN (last_addr) - phys_addr; + offset = phys_addr & ~MMUPAGE_MASK ; + phys_addr &=MMUPAGE_MASK ; + size =MMUPAGE_ALIGN (last_addr) - phys_addr; - nrpages = size >>PAGE_SHIFT ; + nrpages = size >>MMUPAGE_SHIFT ;if (nrpages >BOOT_IOREMAP_PAGES )return NULL ;
Quoted 4 lines · William Lee Irwin III
Quoted 4 lines
I wonder if this would be enough to get sysenter support going again. I've not got a sysenter-capable userspace around, so I can't really test this myself.
vs. pgcl-2.6.0-test11-5
Quoted 4 lines · William Lee Irwin III
Stack decoding fixes, shutting up some compiler warnings, and dumping
PAGE_SIZEandMMUPAGE_SIZEinto/proc/meminfo(for lack of a better place). The printk()'s down there should eventually get ripped out anyway for minimal impact and a quieter boot, but until then...
Woops, those page sizes were a bit off. Come to think of it, so is aio_setup_ring()...
-- wli
fs/aio.c +2 -2
@@ -187,8 +187,8 @@ static int aio_setup_ring(struct kioctxstruct io_event *__event; \unsigned long pfn; \ pfn = (info)->ring_pages[pos/AIO_EVENTS_PER_PAGE ]; \ - __event =kmap_atomic (pfn_to_page (pfn), km); \ - __event += (pfn %PAGE_MMUCOUNT ) *MMUPAGE_SIZE ; \ + __event = (void *)((char *)kmap_atomic (pfn_to_page (pfn), km) \ + +MMUPAGE_SIZE * (pfn %PAGE_MMUCOUNT )); \ __event += pos %AIO_EVENTS_PER_PAGE ; \ __event; \ })
fs/proc/proc_misc.c +2 -2
@@ -231,8 +231,8 @@ static int meminfo_read_proc(char *page, vmtot, vmi.used, vmi.largest_chunk, -K (PAGE_SIZE ), -K (MMUPAGE_SIZE ) +PAGE_SIZE >>10 , +MMUPAGE_SIZE >>10 ); len +=hugetlb_report_meminfo (page + len);
Quoted 6 lines · William Lee Irwin III
Quoted 4 lines · William Lee Irwin III
Quoted 4 lines
I wonder if this would be enough to get sysenter support going again. I've not got a sysenter-capable userspace around, so I can't really test this myself.
vs. pgcl-2.6.0-test11-5
Quoted 8 lines
Quoted 4 lines · William Lee Irwin III
Stack decoding fixes, shutting up some compiler warnings, and dumping
PAGE_SIZEandMMUPAGE_SIZEinto/proc/meminfo(for lack of a better place). The printk()'s down there should eventually get ripped out anyway for minimal impact and a quieter boot, but until then...Woops, those page sizes were a bit off. Come to think of it, so is
aio_setup_ring()...
And here is a sync point, it gets DRI memory allocation/mapping code working. There are a few things which need to be ironed out wrt the radeon driver, but this is some good progress.
Tested with the radeon DRI driver.
Index: linux-2.6.0-test11/drivers/char/drm/drmP.h
===================================================================
RCS file: /build/cvsroot/linux-2.6.0-test11/drivers/char/drm/drmP.h,v retrieving revision 1.1.1.1
drivers/char/drm/drmP.h +1 -1
@@ -203,7 +203,7 @@ static inline struct page * vmalloc_to_p#define DRM_RPR_ARG (vma) vma,#endif -#define VM_OFFSET (vma) ((vma)->vm_pgoff << PAGE_SHIFT) +#define VM_OFFSET (vma) ((vma)->vm_pgoff << MMUPAGE_SHIFT)/*@}*/
Index: linux-2.6.0-test11/drivers/char/drm/drm_bufs.h
===================================================================
RCS file: /build/cvsroot/linux-2.6.0-test11/drivers/char/drm/drm_bufs.h,v retrieving revision 1.1.1.1
drivers/char/drm/drm_bufs.h +4 -4
@@ -120,7 +120,7 @@ int DRM(addmap)( struct inode *inode, st }DRM_DEBUG ("offset = 0x%08lx, size = 0x%08lx, type = %d\n" , map->offset, map->size, map->type ); -if ( (map->offset & (~PAGE_MASK )) || (map->size & (~PAGE_MASK )) ) { +if ((map->offset & ~MMUPAGE_MASK ) || (map->size & ~MMUPAGE_MASK )) {DRM (free)( map,sizeof (*map),DRM_MEM_MAPS );return -EINVAL ; }
@@ -392,7 +392,7 @@ int DRM(addbufs_agp)( struct inode *inod size =1 << order; alignment = (request.flags & _DRM_PAGE_ALIGN) - ?PAGE_ALIGN (size) : size; + ?MMUPAGE_ALIGN (size) : size; page_order = order -PAGE_SHIFT >0 ? order -PAGE_SHIFT :0 ; total =PAGE_SIZE << page_order;
@@ -572,7 +572,7 @@ int DRM(addbufs_pci)( struct inode *inodif ( dev->queue_count )return -EBUSY ;/* Not while in use */ alignment = (request.flags & _DRM_PAGE_ALIGN) - ?PAGE_ALIGN (size) : size; + ?MMUPAGE_ALIGN (size) : size; page_order = order -PAGE_SHIFT >0 ? order -PAGE_SHIFT :0 ; total =PAGE_SIZE << page_order;
@@ -800,7 +800,7 @@ int DRM(addbufs_sg)( struct inode *inode size =1 << order; alignment = (request.flags & _DRM_PAGE_ALIGN) - ?PAGE_ALIGN (size) : size; + ?MMUPAGE_ALIGN (size) : size; page_order = order -PAGE_SHIFT >0 ? order -PAGE_SHIFT :0 ; total =PAGE_SIZE << page_order;
Index: linux-2.6.0-test11/drivers/char/drm/drm_ioctl.h
===================================================================
RCS file: /build/cvsroot/linux-2.6.0-test11/drivers/char/drm/drm_ioctl.h,v retrieving revision 1.1.1.1
diff -u -p -B -r1.1.1.1 drm_ioctl.h
Index: linux-2.6.0-test11/drivers/char/drm/drm_memory.h
===================================================================
RCS file: /build/cvsroot/linux-2.6.0-test11/drivers/char/drm/drm_memory.h,v retrieving revision 1.1.1.1
drivers/char/drm/drm_memory.h +3 -2
@@ -279,7 +281,7 @@ unsigned long DRM(alloc_pages)(int orderfor (addr = address, sz = bytes; sz >0 ; addr +=PAGE_SIZE , sz -=PAGE_SIZE ) { -SetPageReserved (virt_to_page (addr)); +SetPageReserved (pfn_to_page (vmalloc_to_pfn ((void *)addr))); }return address;
@@ -307,7 +309,8 @@ void DRM(free_pages)(unsigned long addrefor (addr = address, sz = bytes; sz >0 ; addr +=PAGE_SIZE , sz -=PAGE_SIZE ) { -ClearPageReserved (virt_to_page (addr)); + +ClearPageReserved (pfn_to_page (vmalloc_to_pfn ((void *)addr))); }free_pages (address, order);
Index: linux-2.6.0-test11/drivers/char/drm/drm_scatter.h
===================================================================
RCS file: /build/cvsroot/linux-2.6.0-test11/drivers/char/drm/drm_scatter.h,v retrieving revision 1.1.1.1
drivers/char/drm/drm_scatter.h +1 -1
@@ -137,7 +137,7 @@ int DRM(sg_alloc)( struct inode *inode,DRM_DEBUG ("sg alloc virtual = %p\n" , entry->virtual );for ( i = entry->handle, j =0 ; j < pages ; i +=PAGE_SIZE , j++ ) { - entry->pagelist[j] =vmalloc_to_page ((void *)i); + entry->pagelist[j] =pfn_to_page (vmalloc_to_pfn ((void *)i));if (!entry->pagelist[j]) goto failed;SetPageReserved (entry->pagelist[j]);
Index: linux-2.6.0-test11/drivers/char/drm/drm_vm.h
===================================================================
RCS file: /build/cvsroot/linux-2.6.0-test11/drivers/char/drm/drm_vm.h,v retrieving revision 1.1.1.1
drivers/char/drm/drm_vm.h +1 -1
@@ -166,7 +166,7 @@ struct page *DRM(vm_shm_nopage)(struct v offset = address - vma->vm_start; i = (unsigned long )map->handle + offset; - page =vmalloc_to_page ((void *)i); + page =pfn_to_page (vmalloc_to_pfn ((void *)i));if (!page)return NOPAGE_OOM ;get_page (page);
Quoted 2 lines · William Lee Irwin III
Quoted 2 lines
Woops, those page sizes were a bit off. Come to think of it, so is
aio_setup_ring()...
Quoted 4 lines · Zwane Mwaikambo
And here is a sync point, it gets DRI memory allocation/mapping code working. There are a few things which need to be ironed out wrt the radeon driver, but this is some good progress.
Tested with the radeon DRI driver.
Terrific! I'll plop that and the various minimal fixes I've got atop -9 and ship it as -10 shortly.
-- wli
Review enum fixed_addresses { · line 117
Quoted 11 lines · William Lee Irwin III
(quoted diff) +6
-0@@ -107,6 +108,12 @@ enum fixed_addresses { FIX_BTMAP_END = __end_of_permanent_fixed_addresses, FIX_BTMAP_BEGIN = FIX_BTMAP_END + NR_FIX_BTMAPS - 1, FIX_WP_TEST, +#ifdef CONFIG_HIGHMEM + FIX_KMAP_BEGIN = __virt_to_fix(__fix_to_virt(FIX_WP_TEST+1) & PAGE_MASK) - PAGE_MMUCOUNT + 1, + FIX_KMAP_END = FIX_KMAP_BEGIN+((KM_TYPE_NR*NR_CPUS+1)*PAGE_MMUCOUNT)-1, + FIX_PKMAP_BEGIN = __virt_to_fix(__fix_to_virt(FIX_KMAP_END+1) & PKMAP_MASK) - PAGE_MMUCOUNT + 1, + FIX_PKMAP_END = FIX_PKMAP_BEGIN + (LAST_PKMAP+1)*PAGE_MMUCOUNT - 1, +#endif __end_of_fixed_addresses
BZZZT. kmap_atomic_to_pfn() et al all check vaddr < FIXADDR_START and FIXADDR_START == __fix_to_virt(__end_of_permanent_fixed_addresses); hence, plopping these guys down out by the FIX_BTMAP_*'s is bogus. Worse yet, vmallocspace will overlap the bastards since we have VMALLOC_END == FIXADDR_START - 2*MMUPAGE_SIZE.
Conservative and unnecessarily invasive fix (debug code and all) below.
I wonder if at some point I'll get buried under all the pr_debug()'s. I guess it might give ppl an idea of exactly how much I've had to debug over the course of all this, assuming anyone bothers looking. At any rate, this does have the distinct advantage of running userspace.
-- wli
arch/i386/mm/highmem.c +28 -8
@@ -45,6 +45,9 @@ void *kmap_atomic(struct page *page, enu idx = type +KM_TYPE_NR *smp_processor_id (); vaddr =__fix_to_virt (FIX_KMAP_END ) +PAGE_SIZE *idx; +pr_debug ("kmap_atomic(%d) has pfn 0x%lx, idx %d, vaddr 0x%lx\n" , + type,page_to_pfn (page), idx, vaddr); +BUG_ON (vaddr %PAGE_SIZE );BUG_ON (vaddr >__fix_to_virt (FIX_KMAP_BEGIN ));BUG_ON (vaddr <__fix_to_virt (FIX_KMAP_END ));
@@ -54,6 +57,9 @@ void *kmap_atomic(struct page *page, enu pgd =pgd_offset_k (addr); pmd =pmd_offset (pgd, addr); +/* barf on non-present pagetables */ +BUG_ON (pmd_none (*pmd)); +/* barf on highmem-allocated pagetables */ BUG_ON ((pmd_val (*pmd) >>MMUPAGE_SHIFT ) >= max_low_pfn);
@@ -66,8 +72,13 @@ void *kmap_atomic(struct page *page, enuBUG_ON (addr < vaddr);BUG_ON (addr - vaddr >=PAGE_SIZE );BUG_ON (!pfn_valid (pfn + k)); -if (pte_pfn (pte[k]) == pfn + k) +pr_debug ("%s: mapping pfn 0x%lx at vaddr 0x%lx\n" , + __FUNCTION__, pfn, vaddr); +if (pte_pfn (pte[k]) == pfn + k) { +pr_debug ("%s: skipping already-set kmap_atomic() pte\n" , + __FUNCTION__);continue ; + }set_pte (&pte[k],pfn_pte (pfn + k, kmap_prot));__flush_tlb_one (addr);
@@ -86,7 +97,7 @@ void kunmap_atomic(void *kvaddr, enum kmpmd_t *pmd;pte_t *pte; -if (vaddr <FIXADDR_START ) {// FIXME +if (vaddr <__fix_to_virt (__end_of_fixed_addresses)) {// FIXME dec_preempt_count ();return ; }
@@ -106,7 +117,9 @@ void kunmap_atomic(void *kvaddr, enum kmfor (k =0 ; k <PAGE_MMUCOUNT ; ++k, vaddr +=MMUPAGE_SIZE ) {pte_clear (&pte[k]);__flush_tlb_one (vaddr); - } + } else + pr_debug("%s: skipping already-set kmap_atomic() pte\n", + __FUNCTION__);dec_preempt_count (); }
@@ -119,17 +132,18 @@ unsigned long kmap_atomic_to_pfn(void *ppmd_t *pmd;pte_t *pte; -if (vaddr <FIXADDR_START ) +/* +* Not for vmallocspace!!! +*/ +BUG_ON (vaddr >=VMALLOC_START && vaddr < __VMALLOC_END); +if (vaddr <VMALLOC_START )return __pa (vaddr)/MMUPAGE_SIZE ; pgd =pgd_offset_k (vaddr); pmd =pmd_offset (pgd, vaddr); +BUG_ON (pmd_none (*pmd)); pte =pte_offset_kernel (pmd, vaddr); -/* -* unsigned long idx = virt_to_fix(vaddr); -* pte = &kmap_pte[idx*PAGE_MMUCOUNT]; -*/ return pte_pfn (*pte); }
@@ -145,11 +159,15 @@ void kmap_atomic_sg(pte_t *ptes[], pte_ainc_preempt_count (); idx = type +KM_TYPE_NR *smp_processor_id (); base = vaddr =__fix_to_virt (FIX_KMAP_END ) +PAGE_SIZE *idx; +pr_debug ("kmap_atomic_sg(%d) has idx %d, vaddr 0x%lx\n" , + type, idx, vaddr); +BUG_ON (vaddr %PAGE_SIZE );BUG_ON (vaddr >__fix_to_virt (FIX_KMAP_BEGIN ));BUG_ON (vaddr <__fix_to_virt (FIX_KMAP_END )); pgd =pgd_offset_k (vaddr); pmd =pmd_offset (pgd, vaddr); +BUG_ON (pmd_none (*pmd)); pte =pte_offset_kernel (pmd, vaddr);for (k =0 ; k <PAGE_MMUCOUNT ; ++k, vaddr +=MMUPAGE_SIZE ) {unsigned long pfn = paddrs[k]/MMUPAGE_SIZE ;
@@ -163,6 +181,8 @@ void kmap_atomic_sg(pte_t *ptes[], pte_aBUG_ON ((u32 )ptes[k] < base);BUG_ON ((u32 )ptes[k] - base >=PAGE_SIZE ); +pr_debug ("%s: mapping pfn 0x%lx at vaddr 0x%lx\n" , + __FUNCTION__, pfn, vaddr);if (pte_pfn (pte[k]) != pfn) {set_pte (&pte[k],pfn_pte (pfn, kmap_prot));__flush_tlb_one (vaddr);
arch/i386/mm/init.c +0 -8
@@ -212,11 +212,6 @@ EXPORT_SYMBOL(kmap_prot);#define kmap_init () do { kmap_prot = PAGE_KERNEL; } while (0) -void __init permanent_kmaps_init (pgd_t *pgd_base) -{ -page_table_range_init (PKMAP_BASE ,PKMAP_BASE +PAGE_SIZE *LAST_PKMAP , pgd_base); -} -void __init one_highpage_init (struct page *page,unsigned long pfn,int bad_ppro) {if (page_is_ram (pfn) && !(bad_ppro &&page_kills_ppro (pfn))) {
@@ -243,7 +238,6 @@ extern void set_highmem_pages_init(int);#else #define kmap_init () do { } while (0) -#define permanent_kmaps_init (pgd_base) do { } while (0)#define set_highmem_pages_init (bad_ppro) do { } while (0)#endif /* CONFIG_HIGHMEM */
@@ -288,8 +282,6 @@ static void __init pagetable_init (void) vaddr =__fix_to_virt (__end_of_fixed_addresses -1 ) &PMD_MASK ;page_table_range_init (vaddr,0 , pgd_base); -permanent_kmaps_init (pgd_base); -#ifdef CONFIG_X86_PAE /* * Add low memory identity-mappings - SMP needs it when
include/asm-generic/rmap.h +19 -3
@@ -57,6 +57,8 @@ static inline void pgtable_add_rmap(stru page->mapping = (void *)mm; page->index = address & ~(VIRT_AREA_MAPPED_PER_PTE_PAGE -1 );inc_page_state (nr_page_table_pages); +pr_debug ("%s: installing pte page at pfn 0x%lx at uvaddr 0x%lx\n" , + __FUNCTION__,page_to_pfn (page), page->index); }static inline void pgtable_remove_rmap (struct page * page)
@@ -75,6 +77,8 @@ static inline void pgtable_remove_rmap(sBUG_ON (atomic_read (&page->count) <=0 ); } +pr_debug ("%s: removing pte page at pfn 0x%lx at uvaddr 0x%lx\n" , + __FUNCTION__,page_to_pfn (page), page->index); page->mapping =NULL ; page->index =0 ;dec_page_state (nr_page_table_pages);
@@ -109,24 +113,36 @@ static inline unsigned long ptep_to_addrunsigned long kvaddr = (unsigned long )ptep;unsigned long swpage_voff = kvaddr/sizeof (pte_t); +pr_debug ("entered ptep_to_address(%p)\n" , ptep); +pr_debug ("swpage_voff = 0x%lx\n" , swpage_voff); +if (1 ) {pgd_t *pgd;pmd_t *pmd;pte_t *pte; -unsigned long pfn; +unsigned long vaddr, pfn;struct page *page; pgd =pgd_offset_k (kvaddr); +pr_debug ("ptep_to_address() saw pgd mapping ptep at %p\n" , pgd); pmd =pmd_offset (pgd, kvaddr); +pr_debug ("ptep_to_address() saw pmd mapping ptep at %p\n" , pmd); pte =pte_offset_kernel (pmd, kvaddr); +pr_debug ("ptep_to_address() saw pte mapping ptep at %p\n" , pte); pfn =pte_pfn (*pte); +pr_debug ("ptep_to_address() saw ptep held in pfn 0x%lx\n" , + pfn); +if (pfn !=kmap_atomic_to_pfn (ptep)) +pr_debug ("pfn doesn't match kmap_atomic_to_pfn()!\n" ); page =pfn_to_page (pfn); -return page->index +PMD_SIZE *(pfn %PAGE_MMUCOUNT ) + vaddr = page->index +PMD_SIZE *(pfn %PAGE_MMUCOUNT ) +MMUPAGE_SIZE *(swpage_voff %PTRS_PER_PTE ); +pr_debug ("ptep_to_address() returning 0x%lx\n" , vaddr); +return vaddr; }else {struct page *page =pfn_to_page (kmap_atomic_to_pfn (ptep)); -WARN_ON (kvaddr > (unsigned long )(-PAGE_SIZE )); +WARN_ON (kvaddr > (unsigned long )(-MMUPAGE_SIZE )); swpage_voff %=MMUPAGES_MAPPED_PER_PTE_PAGE ;/* WARN_ON(swpage_voff != pfn - page_to_pfn(page)); */
include/asm-i386/fixmap.h +9 -7
@@ -69,11 +69,12 @@* worth of virtualspace. */ #define FIXADDR_TOP (-MMUPAGE_SIZE) -#define __FIXADDR_SIZE (__end_of_permanent_fixed_addresses << MMUPAGE_SHIFT) +#define __FIXADDR_SIZE (__end_of_fixed_addresses << MMUPAGE_SHIFT)#define FIXADDR_START (FIXADDR_TOP - __FIXADDR_SIZE)#define __fix_to_virt (x) (FIXADDR_TOP - ((x) << MMUPAGE_SHIFT))#define __virt_to_fix (x) ((FIXADDR_TOP - ((x) & MMUPAGE_MASK)) >> MMUPAGE_SHIFT) +#define __fixmap_align (x,a) __virt_to_fix(__fix_to_virt(x) & (a))enum fixed_addresses {/* reserved pte's for temporary kernel mappings */
@@ -102,18 +103,19 @@ enum fixed_addresses { FIX_ACPI_BEGIN, FIX_ACPI_END = FIX_ACPI_BEGIN + FIX_ACPI_PAGES - 1,#endif +#ifdef CONFIG_HIGHMEM +FIX_HIGHMEM_HOLE , +FIX_KMAP_BEGIN =__fixmap_align (FIX_HIGHMEM_HOLE +1 ,PAGE_MASK ) +1 , +FIX_KMAP_END =FIX_KMAP_BEGIN +KM_TYPE_NR *NR_CPUS *PAGE_MMUCOUNT -1 , +FIX_PKMAP_BEGIN =__fixmap_align (FIX_KMAP_END +1 ,PKMAP_MASK ) +1 , +FIX_PKMAP_END =FIX_PKMAP_BEGIN +LAST_PKMAP *PAGE_MMUCOUNT -1 , +#endif __end_of_permanent_fixed_addresses,/* temporary boot-time mappings, used before ioremap() is functional */ #define NR_FIX_BTMAPS 16FIX_BTMAP_END = __end_of_permanent_fixed_addresses,FIX_BTMAP_BEGIN =FIX_BTMAP_END +NR_FIX_BTMAPS -1 ,FIX_WP_TEST , -#ifdef CONFIG_HIGHMEM -FIX_KMAP_BEGIN =__virt_to_fix (__fix_to_virt (FIX_WP_TEST +1 ) &PAGE_MASK ) -PAGE_MMUCOUNT +1 , -FIX_KMAP_END =FIX_KMAP_BEGIN +((KM_TYPE_NR *NR_CPUS +1 )*PAGE_MMUCOUNT )-1 , -FIX_PKMAP_BEGIN =__virt_to_fix (__fix_to_virt (FIX_KMAP_END +1 ) &PKMAP_MASK ) -PAGE_MMUCOUNT +1 , -FIX_PKMAP_END =FIX_PKMAP_BEGIN + (LAST_PKMAP +1 )*PAGE_MMUCOUNT -1 , -#endif __end_of_fixed_addresses };
include/asm-i386/pgtable.h +2 -1
@@ -81,7 +81,8 @@ void paging_init(void);* The vmalloc() routines leaves a hole of 4kB between each vmalloced * area for the same reason. ;) */ -#define VMALLOC_END (FIXADDR_START-2*MMUPAGE_SIZE) +#define __VMALLOC_END __fix_to_virt(__end_of_fixed_addresses) +#define VMALLOC_END ((__VMALLOC_END-2*MMUPAGE_SIZE) & PMD_MASK)#define __VMALLOC_START (VMALLOC_END - VMALLOC_RESERVE - 2*MMUPAGE_SIZE)#define VMALLOC_START \ (high_memory \