* 2.5.73-mm2 @ 2003-06-28 3:21 ` Andrew Morton 0 siblings, 0 replies; 49+ messages in thread From: Andrew Morton @ 2003-06-28 3:21 UTC (permalink / raw) To: linux-kernel, linux-mm ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/ Just bits and pieces. Changes since 2.5.73-mm1: linus.patch Latest Linus tree -show_stack-fix.patch -pci-1.patch -pci-2.patch -pci-3.patch -pci-4.patch -pci-5.patch -alsa-pnp-fix.patch -setscheduler-fix.patch -ide_setting_sem-fix.patch -misc6.patch -AT_SECURE-auxv-entry.patch -common-kernel-DSO-name.patch -get_unmapped_area-speedup.patch -d_invalidate-fix.patch -nfs_unlink-d_count-fix.patch -hpfs-d_count-fix.patch -smbfs-oops-workaround.patch -enable-cardbus-bursting.patch -n_tty-column-counting-fix.patch -numa-normalised-node-load.patch -enable-local-apic-on-p4.patch -knfsd-umount-fix.patch -getrlimit-ifdef-fix.patch -amd64-monotonic-clock.patch Merged +kgdb-ga-docco-fixes.patch kgdb documentation fixes +pppoe-revert.patch Back out broken pppoe changes until it gets fixed up. +move_vma-VM_LOCKED-fix.patch Fix the mremap use-after-free fix +ipcsem-speedup.patch Speed up sysv semaphore operations +feral-fix.patch Fix the linux_isp qlogic driver for non-odular builds. -cfq-2.patch Dropped. I'm firming up the IO scheduler rework for a merge and the -mm CFQ inplementation is way out of date. +blk-batching-cleanups.patch Minor touchups +truncate-pagefault-race-fix-fix.patch Tighten up the MP synchronisation -security-vm_enough_memory.patch +security_vm_enough_memory.patch New version +nbd-remove-blksize-bits.patch +nbd-kobject-oops-fix.patch +nbd-paranioa-cleanups.patch +nbd-locking-fixes.patch More NBD work -nr_running-speedup.patch Dropped. It seemed to have no net benefit. +lowmem_page_address-cleanup.patch Simplify lowmem_page_address() +numa-memory-reporting-fix.patch Fix NUMA memory reporting (needs more work) +syslog-efault-reporting.patch check copy_*_user return values +acpismp-fix.patch Fix `acpismp=force' +div64-cleanup.patch Consolidate and fix div64 code +init_timer-debug-trap.patch Debug code to catch people running init_timer() against a running timer. +dvd-ram-rw-fix.patch ide-scsi RW mount fix +mixcomwd-update.patch +arc-rimi-race-fix.patch minor fixups +slab-drain-all-objects-fix.patch Make slab free up all the objects when destroying caches. +ext3-remove-version.patch Remove the version information from ext3. +cdrom-eject-hang-fix.patch Fix mount-time hangs caused by CDROM eject commands. All 138 patches: linus.patch cset-1.1348.16.4-to-1.1516.txt.gz mm.patch add -mmN to EXTRAVERSION kgdb-ga.patch kgdb stub for ia32 (George Anzinger's one) kgdb-use-ggdb.patch kgdb-ga-docco-fixes.patch kgdb doc. edits/corrections HZ-100.patch handle-no-readpage-2.patch check for presence of readpage() in the readahead code pppoe-revert.patch PPPOE reversion config_spinline.patch uninline spinlocks for profiling accuracy. ppc64-fixes-2.patch Maek ppc64 compile ppc64-bat-initialisation-fix.patch ppc64: BAT initialisation fix ppc64-pci-update.patch ppc64-reloc_hide.patch ppc64-semaphore-reimplementation.patch ppc64: use the ia32 semaphore implementation sym-do-160.patch make the SYM driver do 160 MB/sec x86_64-fixes.patch x86_64 fixes irqreturn-snd-via-fix.patch via sound irqreturn fix config-PAGE_OFFSET.patch Configurable kenrel/user memory split lru_cache_add-check.patch lru_cache_add debug check delay-ksoftirqd-fallback.patch Try harded in IRQ context before falling back to ksoftirqd fb-image-depth-fix.patch fbdev image depth fix move_vma-VM_LOCKED-fix.patch move_vma() make_pages_present() fix ds-09-vicam-usercopy-fix.patch vicam usercopy fix buffer-debug.patch buffer.c debugging reiserfs-unmapped-buffer-fix.patch Fix reiserfs BUG e100-use-after-free-fix.patch 3-unmap-page-debugging.patch page unmappng debug patch VM_RESERVED-check.patch VM_RESERVED check ipcsem-speedup.patch ipc semaphore optimization rcu-stats.patch RCU statistics reporting mtrr-hang-fix.patch Fix mtrr-related hang reslabify-pgds-and-pmds.patch re-slabify i386 pgd's and pmd's linux-isp.patch isp-update-1.patch isp-remove-pci_detect.patch feral-fix.patch linux-isp fix list_del-debug.patch list_del debug check airo-schedule-fix.patch airo.c: don't sleep in atomic regions resurrect-batch_requests.patch bring back the batch_requests function kblockd.patch Create `kblockd' workqueue cfq-infrastructure.patch elevator-completion-api.patch elevator completion API as-iosched.patch anticipatory I/O scheduler AS: pgbench improvement AS: discrete read fifo batches AS sync/async batches AS: hash removal fix AS jumbo patch (for SCSI and TCQ) AS: fix stupid thinko AS: no batch-antic-limit AS: autotune write batches AS: divide by zero fix AS: more HZ != 1000 fixes AS: update_write_batch tuning AS locking AS HZ fixes as-double-free-and-debug.patch AS: fix a leak + more debugging as-fix-seek-estimation.patch AS: maybe repair performance drop of random read O_DIRECT as-fix-seeky-loads.patch AS: fix IBM's seek load unplug-use-kblockd.patch Use kblockd for running request queues per-queue-nr_requests.patch per queue nr_requests blk-invert-watermarks.patch blk_congestion_wait threshold cleanup blk-as-hint.patch blk-as-hint get_request_wait-oom-fix.patch handle OOM in get_request_wait(). blk-fair-batches.patch blk-fair-batches blk-fair-batches-2.patch blk fair batches #2 generic-io-contexts.patch generic io contexts blk-request-batching.patch block request batching get_io_context-fix.patch get_io_context fixes blk-allocation-commentary.patch block allocation comments blk-batching-throttle-fix.patch blk batch requests fix blk-batching-cleanups.patch block batching cleanups print-build-options-on-oops.patch print a few config options on oops mmap-prefault.patch prefault of executable mmaps bio-debug-trap.patch BIO debugging patch sound-irq-hack.patch show_task-free-stack-fix.patch show_task() fix and cleanup put_task_struct-debug.patch ia32-mknod64.patch mknod64 for ia32 ext2-64-bit-special-inodes.patch ext2: support for 64-bit device nodes ext3-64-bit-special-inodes.patch ext3: support for 64-bit device nodes 64-bit-dev_t-kdev_t.patch 64-bit dev_t and kdev_t oops-dump-preceding-code.patch i386 oops output: dump preceding code lockmeter.patch invalidate_mmap_range.patch Interface to invalidate regions of mmaps aio-mm-refcounting-fix.patch fix /proc mm_struct refcounting bug aio-01-retry.patch AIO: Core retry infrastructure io_submit_one-EINVAL-fix.patch Fix aio process hang on EINVAL aio-02-lockpage_wq.patch AIO: Async page wait aio-03-fs_read.patch AIO: Filesystem aio read aio-04-buffer_wq.patch AIO: Async buffer wait aio-05-fs_write.patch AIO: Filesystem aio write aio-05-fs_write-fix.patch aio-06-bread_wq.patch AIO: Async block read aio-06-bread_wq-fix.patch aio-07-ext2getblk_wq.patch AIO: Async get block for ext2 O_SYNC-speedup-2.patch speed up O_SYNC writes aio-09-o_sync.patch aio O_SYNC aio-10-BUG-fix.patch AIO: fix a BUG aio-11-workqueue-flush.patch AIO: flush workqueues before destroying ioctx'es aio-12-readahead.patch AIO: readahead fixes lock_buffer_wq-fix.patch lock_buffer_wq fix unuse_mm-locked.patch AIO: hold the context lock across unuse_mm aio-take-task_lock.patch From: Suparna Bhattacharya <suparna@in.ibm.com> Subject: Re: 2.5.72-mm1 - Under heavy testing with AIO,.. vmstat seems to blow the kernel vfsmount_lock.patch From: Maneesh Soni <maneesh@in.ibm.com> Subject: [patch 1/2] vfsmount_lock sched-hot-balancing-fix.patch fix for CPU scheduler load distribution truncate-pagefault-race-fix.patch Fix vmtruncate race and distributed filesystem race truncate-pagefault-race-fix-fix.patch Make sure truncate fix has no race sleepometer.patch sleep instrumentation time-goes-backwards.patch demonstrate do_gettimeofday() going backwards skip-apic-ids-on-boot.patch skip apicids on boot printk-oops-mangle-fix.patch disentangle printk's whilst oopsing on SMP 20-odirect_enable.patch 21-odirect_cruft.patch 22-read_proc.patch 23-write_proc.patch 24-commit_proc.patch 25-odirect.patch nfs-O_DIRECT-always-enabled.patch Force CONFIG_NFS_DIRECTIO seqcount-locking.patch i_size atomic access: infrastructure i_size-atomic-access.patch i_size atomic access aha152x-oops-fix.patch aha152X oops fixes security_vm_enough_memory.patch Security hook for vm_enough_memory nbd-cleanups.patch NBD: cosmetic cleanups nbd-enhanced-diagnostics.patch nbd: enhanced diagnostics support nbd-remove-blksize-bits.patch nbd: remove unneeded blksize_bits field nbd-kobject-oops-fix.patch nbd: initialise the embedded kobject nbd-paranioa-cleanups.patch nbd: cleanup PARANOIA usage & code nbd-locking-fixes.patch nbd: fix locking issues with ioctl UI pcmcia-event-20030623-1.patch pcmcia-event-20030623-2.patch pcmcia-event-20030623-3.patch pcmcia-event-20030623-4.patch pcmcia-event-20030623-5.patch pcmcia-event-20030623-6.patch sym2-bus_addr-fix.patch sym53c8xx_2 bus_addr fix lost-tick-speedstep-fix.patch Fix lost tick detection for speedstep sym2-remove-broken-bios-check.patch remove a bogus check in sym2 driver rename-timer.patch timer cleanups lowmem_page_address-cleanup.patch cleanup and generalise lowmem_page_address numa-memory-reporting-fix.patch fix NUMA memory reporting ... again syslog-efault-reporting.patch Fix syslog(2) EFAULT reporting acpismp-fix.patch ACPI_HT_ONLY acpismp=force div64-cleanup.patch Kill div64.h dupes and parenthesize do_div() parameters init_timer-debug-trap.patch init_timer debugging dvd-ram-rw-fix.patch 2.5.73 can't mount DVD-RAM via ide-scsi mixcomwd-update.patch Remove check_region and MOD_*_USE_COUNT from mixcomwd.c arc-rimi-race-fix.patch Remove racy check_mem_region() call from arc-rimi.c slab-drain-all-objects-fix.patch kmem_cache_destroy() forgets to drain all objects ext3-remove-version.patch ext3: remove the version number cdrom-eject-hang-fix.patch cdrom eject scribbles on the request flags ^ permalink raw reply [flat|nested] 49+ messages in thread
* 2.5.73-mm2 @ 2003-06-28 3:21 ` Andrew Morton 0 siblings, 0 replies; 49+ messages in thread From: Andrew Morton @ 2003-06-28 3:21 UTC (permalink / raw) To: linux-kernel, linux-mm ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/ Just bits and pieces. Changes since 2.5.73-mm1: linus.patch Latest Linus tree -show_stack-fix.patch -pci-1.patch -pci-2.patch -pci-3.patch -pci-4.patch -pci-5.patch -alsa-pnp-fix.patch -setscheduler-fix.patch -ide_setting_sem-fix.patch -misc6.patch -AT_SECURE-auxv-entry.patch -common-kernel-DSO-name.patch -get_unmapped_area-speedup.patch -d_invalidate-fix.patch -nfs_unlink-d_count-fix.patch -hpfs-d_count-fix.patch -smbfs-oops-workaround.patch -enable-cardbus-bursting.patch -n_tty-column-counting-fix.patch -numa-normalised-node-load.patch -enable-local-apic-on-p4.patch -knfsd-umount-fix.patch -getrlimit-ifdef-fix.patch -amd64-monotonic-clock.patch Merged +kgdb-ga-docco-fixes.patch kgdb documentation fixes +pppoe-revert.patch Back out broken pppoe changes until it gets fixed up. +move_vma-VM_LOCKED-fix.patch Fix the mremap use-after-free fix +ipcsem-speedup.patch Speed up sysv semaphore operations +feral-fix.patch Fix the linux_isp qlogic driver for non-odular builds. -cfq-2.patch Dropped. I'm firming up the IO scheduler rework for a merge and the -mm CFQ inplementation is way out of date. +blk-batching-cleanups.patch Minor touchups +truncate-pagefault-race-fix-fix.patch Tighten up the MP synchronisation -security-vm_enough_memory.patch +security_vm_enough_memory.patch New version +nbd-remove-blksize-bits.patch +nbd-kobject-oops-fix.patch +nbd-paranioa-cleanups.patch +nbd-locking-fixes.patch More NBD work -nr_running-speedup.patch Dropped. It seemed to have no net benefit. +lowmem_page_address-cleanup.patch Simplify lowmem_page_address() +numa-memory-reporting-fix.patch Fix NUMA memory reporting (needs more work) +syslog-efault-reporting.patch check copy_*_user return values +acpismp-fix.patch Fix `acpismp=force' +div64-cleanup.patch Consolidate and fix div64 code +init_timer-debug-trap.patch Debug code to catch people running init_timer() against a running timer. +dvd-ram-rw-fix.patch ide-scsi RW mount fix +mixcomwd-update.patch +arc-rimi-race-fix.patch minor fixups +slab-drain-all-objects-fix.patch Make slab free up all the objects when destroying caches. +ext3-remove-version.patch Remove the version information from ext3. +cdrom-eject-hang-fix.patch Fix mount-time hangs caused by CDROM eject commands. All 138 patches: linus.patch cset-1.1348.16.4-to-1.1516.txt.gz mm.patch add -mmN to EXTRAVERSION kgdb-ga.patch kgdb stub for ia32 (George Anzinger's one) kgdb-use-ggdb.patch kgdb-ga-docco-fixes.patch kgdb doc. edits/corrections HZ-100.patch handle-no-readpage-2.patch check for presence of readpage() in the readahead code pppoe-revert.patch PPPOE reversion config_spinline.patch uninline spinlocks for profiling accuracy. ppc64-fixes-2.patch Maek ppc64 compile ppc64-bat-initialisation-fix.patch ppc64: BAT initialisation fix ppc64-pci-update.patch ppc64-reloc_hide.patch ppc64-semaphore-reimplementation.patch ppc64: use the ia32 semaphore implementation sym-do-160.patch make the SYM driver do 160 MB/sec x86_64-fixes.patch x86_64 fixes irqreturn-snd-via-fix.patch via sound irqreturn fix config-PAGE_OFFSET.patch Configurable kenrel/user memory split lru_cache_add-check.patch lru_cache_add debug check delay-ksoftirqd-fallback.patch Try harded in IRQ context before falling back to ksoftirqd fb-image-depth-fix.patch fbdev image depth fix move_vma-VM_LOCKED-fix.patch move_vma() make_pages_present() fix ds-09-vicam-usercopy-fix.patch vicam usercopy fix buffer-debug.patch buffer.c debugging reiserfs-unmapped-buffer-fix.patch Fix reiserfs BUG e100-use-after-free-fix.patch 3-unmap-page-debugging.patch page unmappng debug patch VM_RESERVED-check.patch VM_RESERVED check ipcsem-speedup.patch ipc semaphore optimization rcu-stats.patch RCU statistics reporting mtrr-hang-fix.patch Fix mtrr-related hang reslabify-pgds-and-pmds.patch re-slabify i386 pgd's and pmd's linux-isp.patch isp-update-1.patch isp-remove-pci_detect.patch feral-fix.patch linux-isp fix list_del-debug.patch list_del debug check airo-schedule-fix.patch airo.c: don't sleep in atomic regions resurrect-batch_requests.patch bring back the batch_requests function kblockd.patch Create `kblockd' workqueue cfq-infrastructure.patch elevator-completion-api.patch elevator completion API as-iosched.patch anticipatory I/O scheduler AS: pgbench improvement AS: discrete read fifo batches AS sync/async batches AS: hash removal fix AS jumbo patch (for SCSI and TCQ) AS: fix stupid thinko AS: no batch-antic-limit AS: autotune write batches AS: divide by zero fix AS: more HZ != 1000 fixes AS: update_write_batch tuning AS locking AS HZ fixes as-double-free-and-debug.patch AS: fix a leak + more debugging as-fix-seek-estimation.patch AS: maybe repair performance drop of random read O_DIRECT as-fix-seeky-loads.patch AS: fix IBM's seek load unplug-use-kblockd.patch Use kblockd for running request queues per-queue-nr_requests.patch per queue nr_requests blk-invert-watermarks.patch blk_congestion_wait threshold cleanup blk-as-hint.patch blk-as-hint get_request_wait-oom-fix.patch handle OOM in get_request_wait(). blk-fair-batches.patch blk-fair-batches blk-fair-batches-2.patch blk fair batches #2 generic-io-contexts.patch generic io contexts blk-request-batching.patch block request batching get_io_context-fix.patch get_io_context fixes blk-allocation-commentary.patch block allocation comments blk-batching-throttle-fix.patch blk batch requests fix blk-batching-cleanups.patch block batching cleanups print-build-options-on-oops.patch print a few config options on oops mmap-prefault.patch prefault of executable mmaps bio-debug-trap.patch BIO debugging patch sound-irq-hack.patch show_task-free-stack-fix.patch show_task() fix and cleanup put_task_struct-debug.patch ia32-mknod64.patch mknod64 for ia32 ext2-64-bit-special-inodes.patch ext2: support for 64-bit device nodes ext3-64-bit-special-inodes.patch ext3: support for 64-bit device nodes 64-bit-dev_t-kdev_t.patch 64-bit dev_t and kdev_t oops-dump-preceding-code.patch i386 oops output: dump preceding code lockmeter.patch invalidate_mmap_range.patch Interface to invalidate regions of mmaps aio-mm-refcounting-fix.patch fix /proc mm_struct refcounting bug aio-01-retry.patch AIO: Core retry infrastructure io_submit_one-EINVAL-fix.patch Fix aio process hang on EINVAL aio-02-lockpage_wq.patch AIO: Async page wait aio-03-fs_read.patch AIO: Filesystem aio read aio-04-buffer_wq.patch AIO: Async buffer wait aio-05-fs_write.patch AIO: Filesystem aio write aio-05-fs_write-fix.patch aio-06-bread_wq.patch AIO: Async block read aio-06-bread_wq-fix.patch aio-07-ext2getblk_wq.patch AIO: Async get block for ext2 O_SYNC-speedup-2.patch speed up O_SYNC writes aio-09-o_sync.patch aio O_SYNC aio-10-BUG-fix.patch AIO: fix a BUG aio-11-workqueue-flush.patch AIO: flush workqueues before destroying ioctx'es aio-12-readahead.patch AIO: readahead fixes lock_buffer_wq-fix.patch lock_buffer_wq fix unuse_mm-locked.patch AIO: hold the context lock across unuse_mm aio-take-task_lock.patch From: Suparna Bhattacharya <suparna@in.ibm.com> Subject: Re: 2.5.72-mm1 - Under heavy testing with AIO,.. vmstat seems to blow the kernel vfsmount_lock.patch From: Maneesh Soni <maneesh@in.ibm.com> Subject: [patch 1/2] vfsmount_lock sched-hot-balancing-fix.patch fix for CPU scheduler load distribution truncate-pagefault-race-fix.patch Fix vmtruncate race and distributed filesystem race truncate-pagefault-race-fix-fix.patch Make sure truncate fix has no race sleepometer.patch sleep instrumentation time-goes-backwards.patch demonstrate do_gettimeofday() going backwards skip-apic-ids-on-boot.patch skip apicids on boot printk-oops-mangle-fix.patch disentangle printk's whilst oopsing on SMP 20-odirect_enable.patch 21-odirect_cruft.patch 22-read_proc.patch 23-write_proc.patch 24-commit_proc.patch 25-odirect.patch nfs-O_DIRECT-always-enabled.patch Force CONFIG_NFS_DIRECTIO seqcount-locking.patch i_size atomic access: infrastructure i_size-atomic-access.patch i_size atomic access aha152x-oops-fix.patch aha152X oops fixes security_vm_enough_memory.patch Security hook for vm_enough_memory nbd-cleanups.patch NBD: cosmetic cleanups nbd-enhanced-diagnostics.patch nbd: enhanced diagnostics support nbd-remove-blksize-bits.patch nbd: remove unneeded blksize_bits field nbd-kobject-oops-fix.patch nbd: initialise the embedded kobject nbd-paranioa-cleanups.patch nbd: cleanup PARANOIA usage & code nbd-locking-fixes.patch nbd: fix locking issues with ioctl UI pcmcia-event-20030623-1.patch pcmcia-event-20030623-2.patch pcmcia-event-20030623-3.patch pcmcia-event-20030623-4.patch pcmcia-event-20030623-5.patch pcmcia-event-20030623-6.patch sym2-bus_addr-fix.patch sym53c8xx_2 bus_addr fix lost-tick-speedstep-fix.patch Fix lost tick detection for speedstep sym2-remove-broken-bios-check.patch remove a bogus check in sym2 driver rename-timer.patch timer cleanups lowmem_page_address-cleanup.patch cleanup and generalise lowmem_page_address numa-memory-reporting-fix.patch fix NUMA memory reporting ... again syslog-efault-reporting.patch Fix syslog(2) EFAULT reporting acpismp-fix.patch ACPI_HT_ONLY acpismp=force div64-cleanup.patch Kill div64.h dupes and parenthesize do_div() parameters init_timer-debug-trap.patch init_timer debugging dvd-ram-rw-fix.patch 2.5.73 can't mount DVD-RAM via ide-scsi mixcomwd-update.patch Remove check_region and MOD_*_USE_COUNT from mixcomwd.c arc-rimi-race-fix.patch Remove racy check_mem_region() call from arc-rimi.c slab-drain-all-objects-fix.patch kmem_cache_destroy() forgets to drain all objects ext3-remove-version.patch ext3: remove the version number cdrom-eject-hang-fix.patch cdrom eject scribbles on the request flags -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-28 3:21 ` 2.5.73-mm2 Andrew Morton @ 2003-06-28 8:56 ` William Lee Irwin III -1 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-06-28 8:56 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-mm On Fri, Jun 27, 2003 at 08:21:30PM -0700, Andrew Morton wrote: > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/ > Just bits and pieces. This could almost be sent through rusty, but I think in general BKL removal patches should go through whatever ringer -mm et al provide. -- wli Remove spurious BKL acquisitions in /proc/. The BKL is not required to access nr_threads for reporting, and get_locks_status() takes it internally, wrapping all operations with it. diff -prauN wli-2.5.73-3/fs/proc/proc_misc.c wli-2.5.73-4/fs/proc/proc_misc.c --- wli-2.5.73-3/fs/proc/proc_misc.c 2003-06-23 10:29:54.000000000 -0700 +++ wli-2.5.73-4/fs/proc/proc_misc.c 2003-06-23 10:32:25.000000000 -0700 @@ -497,11 +497,10 @@ static int ds1286_read_proc(char *page, static int locks_read_proc(char *page, char **start, off_t off, int count, int *eof, void *data) { - int len; - lock_kernel(); - len = get_locks_status(page, start, off, count); - unlock_kernel(); - if (len < count) *eof = 1; + int len = get_locks_status(page, start, off, count); + + if (len < count) + *eof = 1; return len; } diff -prauN wli-2.5.73-3/fs/proc/root.c wli-2.5.73-4/fs/proc/root.c --- wli-2.5.73-3/fs/proc/root.c 2003-06-22 11:33:07.000000000 -0700 +++ wli-2.5.73-4/fs/proc/root.c 2003-06-23 10:32:25.000000000 -0700 @@ -81,11 +81,13 @@ void __init proc_root_init(void) static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry) { - if (dir->i_ino == PROC_ROOT_INO) { /* check for safety... */ - lock_kernel(); + /* + * nr_threads is actually protected by the tasklist_lock; + * however, it's conventional to do reads, especially for + * reporting, without any locking whatsoever. + */ + if (dir->i_ino == PROC_ROOT_INO) /* check for safety... */ dir->i_nlink = proc_root.nlink + nr_threads; - unlock_kernel(); - } if (!proc_lookup(dir, dentry)) { return NULL; ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-06-28 8:56 ` William Lee Irwin III 0 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-06-28 8:56 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-mm On Fri, Jun 27, 2003 at 08:21:30PM -0700, Andrew Morton wrote: > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/ > Just bits and pieces. This could almost be sent through rusty, but I think in general BKL removal patches should go through whatever ringer -mm et al provide. -- wli Remove spurious BKL acquisitions in /proc/. The BKL is not required to access nr_threads for reporting, and get_locks_status() takes it internally, wrapping all operations with it. diff -prauN wli-2.5.73-3/fs/proc/proc_misc.c wli-2.5.73-4/fs/proc/proc_misc.c --- wli-2.5.73-3/fs/proc/proc_misc.c 2003-06-23 10:29:54.000000000 -0700 +++ wli-2.5.73-4/fs/proc/proc_misc.c 2003-06-23 10:32:25.000000000 -0700 @@ -497,11 +497,10 @@ static int ds1286_read_proc(char *page, static int locks_read_proc(char *page, char **start, off_t off, int count, int *eof, void *data) { - int len; - lock_kernel(); - len = get_locks_status(page, start, off, count); - unlock_kernel(); - if (len < count) *eof = 1; + int len = get_locks_status(page, start, off, count); + + if (len < count) + *eof = 1; return len; } diff -prauN wli-2.5.73-3/fs/proc/root.c wli-2.5.73-4/fs/proc/root.c --- wli-2.5.73-3/fs/proc/root.c 2003-06-22 11:33:07.000000000 -0700 +++ wli-2.5.73-4/fs/proc/root.c 2003-06-23 10:32:25.000000000 -0700 @@ -81,11 +81,13 @@ void __init proc_root_init(void) static struct dentry *proc_root_lookup(struct inode * dir, struct dentry * dentry) { - if (dir->i_ino == PROC_ROOT_INO) { /* check for safety... */ - lock_kernel(); + /* + * nr_threads is actually protected by the tasklist_lock; + * however, it's conventional to do reads, especially for + * reporting, without any locking whatsoever. + */ + if (dir->i_ino == PROC_ROOT_INO) /* check for safety... */ dir->i_nlink = proc_root.nlink + nr_threads; - unlock_kernel(); - } if (!proc_lookup(dir, dentry)) { return NULL; -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-28 3:21 ` 2.5.73-mm2 Andrew Morton @ 2003-06-28 15:54 ` William Lee Irwin III -1 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-06-28 15:54 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-mm On Fri, Jun 27, 2003 at 08:21:30PM -0700, Andrew Morton wrote: > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/ > Just bits and pieces. Here's highpmd. This allocates L2 pagetables from highmem, decreasing the per-process lowmem overhead on CONFIG_HIGHMEM64G from 20KB to 8KB. Some attempts were made to update non-i386 architectures to the new API's, though they're entirely untested. It's been tested for a while in -wli on i386 machines, both lowmem and highmem boxen. -- wli diff -prauN mm2-2.5.73-1/arch/i386/Kconfig mm2-2.5.73-2/arch/i386/Kconfig --- mm2-2.5.73-1/arch/i386/Kconfig 2003-06-28 03:09:46.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/Kconfig 2003-06-28 03:11:37.000000000 -0700 @@ -765,6 +765,15 @@ config HIGHPTE low memory. Setting this option will put user-space page table entries in high memory. +config HIGHPMD + bool "Allocate 2nd-level pagetables from highmem" + depends on HIGHMEM64G + help + The VM uses one pmd entry for each pagetable page of physical + memory allocated. For systems with extreme amounts of highmem, + this cannot be tolerated. Setting this option will put + userspace 2nd-level pagetables in highmem. + config MATH_EMULATION bool "Math emulation" ---help--- diff -prauN mm2-2.5.73-1/arch/i386/kernel/vm86.c mm2-2.5.73-2/arch/i386/kernel/vm86.c --- mm2-2.5.73-1/arch/i386/kernel/vm86.c 2003-06-22 11:32:33.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/kernel/vm86.c 2003-06-28 03:11:37.000000000 -0700 @@ -144,12 +144,14 @@ static void mark_screen_rdonly(struct ta pgd_clear(pgd); goto out; } - pmd = pmd_offset(pgd, 0xA0000); - if (pmd_none(*pmd)) + pmd = pmd_offset_map(pgd, 0xA0000); + if (pmd_none(*pmd)) { + pmd_unmap(pmd); goto out; - if (pmd_bad(*pmd)) { + } else if (pmd_bad(*pmd)) { pmd_ERROR(*pmd); pmd_clear(pmd); + pmd_unmap(pmd); goto out; } pte = mapped = pte_offset_map(pmd, 0xA0000); @@ -159,6 +161,7 @@ static void mark_screen_rdonly(struct ta pte++; } pte_unmap(mapped); + pmd_unmap(pmd); out: spin_unlock(&tsk->mm->page_table_lock); preempt_enable(); diff -prauN mm2-2.5.73-1/arch/i386/mm/fault.c mm2-2.5.73-2/arch/i386/mm/fault.c --- mm2-2.5.73-1/arch/i386/mm/fault.c 2003-06-28 03:09:46.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/fault.c 2003-06-28 03:28:31.000000000 -0700 @@ -253,6 +253,7 @@ no_context: printk(" printing eip:\n"); printk("%08lx\n", regs->eip); asm("movl %%cr3,%0":"=r" (page)); +#ifndef CONFIG_HIGHPMD /* Oh boy. Error reporting is going to blow major goats. */ page = ((unsigned long *) __va(page))[address >> 22]; printk(KERN_ALERT "*pde = %08lx\n", page); /* @@ -268,7 +269,14 @@ no_context: page = ((unsigned long *) __va(page))[address >> PAGE_SHIFT]; printk(KERN_ALERT "*pte = %08lx\n", page); } -#endif +#endif /* !CONFIG_HIGHPTE */ +#else /* CONFIG_HIGHPMD */ + printk(KERN_ALERT "%%cr3 = 0x%lx\n", page); + /* Mask off flag bits. It should end up 32B-aligned. */ + page &= ~(PTRS_PER_PGD*sizeof(pgd_t) - 1); + printk(KERN_ALERT "*pdpte = 0x%Lx\n", + pgd_val(((pgd_t *)__va(page))[address >> PGDIR_SHIFT])); +#endif /* CONFIG_HIGHPMD */ die("Oops", regs, error_code); bust_spinlocks(0); do_exit(SIGKILL); @@ -336,8 +344,8 @@ vmalloc_fault: * and redundant with the set_pmd() on non-PAE. */ - pmd = pmd_offset(pgd, address); - pmd_k = pmd_offset(pgd_k, address); + pmd = pmd_offset_kernel(pgd, address); + pmd_k = pmd_offset_kernel(pgd_k, address); if (!pmd_present(*pmd_k)) goto no_context; set_pmd(pmd, *pmd_k); diff -prauN mm2-2.5.73-1/arch/i386/mm/hugetlbpage.c mm2-2.5.73-2/arch/i386/mm/hugetlbpage.c --- mm2-2.5.73-1/arch/i386/mm/hugetlbpage.c 2003-06-22 11:33:17.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/hugetlbpage.c 2003-06-28 03:11:37.000000000 -0700 @@ -87,8 +87,8 @@ static pte_t *huge_pte_alloc(struct mm_s pmd_t *pmd = NULL; pgd = pgd_offset(mm, addr); - pmd = pmd_alloc(mm, pgd, addr); - return (pte_t *) pmd; + pmd = pmd_alloc_map(mm, pgd, addr); + return (pte_t *)pmd; } static pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr) @@ -97,8 +97,8 @@ static pte_t *huge_pte_offset(struct mm_ pmd_t *pmd = NULL; pgd = pgd_offset(mm, addr); - pmd = pmd_offset(pgd, addr); - return (pte_t *) pmd; + pmd = pmd_offset_map(pgd, addr); + return (pte_t *)pmd; } static void set_huge_pte(struct mm_struct *mm, struct vm_area_struct *vma, struct page *page, pte_t * page_table, int write_access) @@ -145,6 +145,8 @@ int copy_hugetlb_page_range(struct mm_st ptepage = pte_page(entry); get_page(ptepage); set_pte(dst_pte, entry); + pmd_unmap(dst_pte); + pmd_unmap_nested(src_pte); dst->rss += (HPAGE_SIZE / PAGE_SIZE); addr += HPAGE_SIZE; } @@ -182,6 +184,7 @@ follow_hugetlb_page(struct mm_struct *mm get_page(page); pages[i] = page; + pmd_unmap(pte); } if (vmas) @@ -271,6 +274,7 @@ follow_huge_pmd(struct mm_struct *mm, un page += ((address & ~HPAGE_MASK) >> PAGE_SHIFT); get_page(page); } + pmd_unmap(pmd); return page; } #endif @@ -314,6 +318,7 @@ void unmap_hugepage_range(struct vm_area page = pte_page(*pte); huge_page_release(page); pte_clear(pte); + pmd_unmap(pte); } mm->rss -= (end - start) >> PAGE_SHIFT; flush_tlb_range(vma, start, end); @@ -358,16 +363,19 @@ int hugetlb_prefault(struct address_spac page = alloc_hugetlb_page(); if (!page) { ret = -ENOMEM; + pmd_unmap(pte); goto out; } ret = add_to_page_cache(page, mapping, idx, GFP_ATOMIC); unlock_page(page); if (ret) { free_huge_page(page); + pmd_unmap(pte); goto out; } } set_huge_pte(mm, vma, page, pte, vma->vm_flags & VM_WRITE); + pmd_unmap(pte); } out: spin_unlock(&mm->page_table_lock); diff -prauN mm2-2.5.73-1/arch/i386/mm/init.c mm2-2.5.73-2/arch/i386/mm/init.c --- mm2-2.5.73-1/arch/i386/mm/init.c 2003-06-28 03:09:46.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/init.c 2003-06-28 03:22:00.000000000 -0700 @@ -59,10 +59,10 @@ static pmd_t * __init one_md_table_init( #ifdef CONFIG_X86_PAE pmd_table = (pmd_t *) alloc_bootmem_low_pages(PAGE_SIZE); set_pgd(pgd, __pgd(__pa(pmd_table) | _PAGE_PRESENT)); - if (pmd_table != pmd_offset(pgd, 0)) + if (pmd_table != pmd_offset_kernel(pgd, 0)) BUG(); #else - pmd_table = pmd_offset(pgd, 0); + pmd_table = pmd_offset_kernel(pgd, 0); #endif return pmd_table; @@ -113,7 +113,7 @@ static void __init page_table_range_init if (pgd_none(*pgd)) one_md_table_init(pgd); - pmd = pmd_offset(pgd, vaddr); + pmd = pmd_offset_kernel(pgd, vaddr); for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); pmd++, pmd_idx++) { if (pmd_none(*pmd)) one_page_table_init(pmd); @@ -194,7 +194,7 @@ pte_t *kmap_pte; pgprot_t kmap_prot; #define kmap_get_fixmap_pte(vaddr) \ - pte_offset_kernel(pmd_offset(pgd_offset_k(vaddr), (vaddr)), (vaddr)) + pte_offset_kernel(pmd_offset_kernel(pgd_offset_k(vaddr), (vaddr)), (vaddr)) void __init kmap_init(void) { @@ -218,7 +218,7 @@ void __init permanent_kmaps_init(pgd_t * page_table_range_init(vaddr, vaddr + PAGE_SIZE*LAST_PKMAP, pgd_base); pgd = swapper_pg_dir + pgd_index(vaddr); - pmd = pmd_offset(pgd, vaddr); + pmd = pmd_offset_kernel(pgd, vaddr); pte = pte_offset_kernel(pmd, vaddr); pkmap_page_table = pte; } @@ -513,20 +513,9 @@ void __init mem_init(void) } kmem_cache_t *pgd_cache; -kmem_cache_t *pmd_cache; void __init pgtable_cache_init(void) { - if (PTRS_PER_PMD > 1) { - pmd_cache = kmem_cache_create("pmd", - PTRS_PER_PMD*sizeof(pmd_t), - 0, - SLAB_HWCACHE_ALIGN | SLAB_MUST_HWCACHE_ALIGN, - pmd_ctor, - NULL); - if (!pmd_cache) - panic("pgtable_cache_init(): cannot create pmd cache"); - } pgd_cache = kmem_cache_create("pgd", PTRS_PER_PGD*sizeof(pgd_t), 0, diff -prauN mm2-2.5.73-1/arch/i386/mm/ioremap.c mm2-2.5.73-2/arch/i386/mm/ioremap.c --- mm2-2.5.73-1/arch/i386/mm/ioremap.c 2003-06-22 11:32:38.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/ioremap.c 2003-06-28 03:11:37.000000000 -0700 @@ -82,7 +82,7 @@ static int remap_area_pages(unsigned lon spin_lock(&init_mm.page_table_lock); do { pmd_t *pmd; - pmd = pmd_alloc(&init_mm, dir, address); + pmd = pmd_alloc_kernel(&init_mm, dir, address); error = -ENOMEM; if (!pmd) break; diff -prauN mm2-2.5.73-1/arch/i386/mm/pageattr.c mm2-2.5.73-2/arch/i386/mm/pageattr.c --- mm2-2.5.73-1/arch/i386/mm/pageattr.c 2003-06-28 03:09:46.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/pageattr.c 2003-06-28 03:12:16.000000000 -0700 @@ -23,7 +23,7 @@ static inline pte_t *lookup_address(unsi pmd_t *pmd; if (pgd_none(*pgd)) return NULL; - pmd = pmd_offset(pgd, address); + pmd = pmd_offset_kernel(pgd, address); if (pmd_none(*pmd)) return NULL; if (pmd_large(*pmd)) @@ -79,7 +79,7 @@ static void set_pmd_pte(pte_t *kpte, uns pgd_t *pgd; pmd_t *pmd; pgd = (pgd_t *)page_address(page) + pgd_index(address); - pmd = pmd_offset(pgd, address); + pmd = pmd_offset_kernel(pgd, address); set_pte_atomic((pte_t *)pmd, pte); } spin_unlock_irqrestore(&pgd_lock, flags); @@ -92,7 +92,7 @@ static void set_pmd_pte(pte_t *kpte, uns static inline void revert_page(struct page *kpte_page, unsigned long address) { pte_t *linear = (pte_t *) - pmd_offset(pgd_offset(&init_mm, address), address); + pmd_offset_kernel(pgd_offset_k(address), address); set_pmd_pte(linear, address, pfn_pte((__pa(address) & LARGE_PAGE_MASK) >> PAGE_SHIFT, PAGE_KERNEL_LARGE)); diff -prauN mm2-2.5.73-1/arch/i386/mm/pgtable.c mm2-2.5.73-2/arch/i386/mm/pgtable.c --- mm2-2.5.73-1/arch/i386/mm/pgtable.c 2003-06-28 03:09:46.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/pgtable.c 2003-06-28 08:20:29.000000000 -0700 @@ -70,7 +70,7 @@ static void set_pte_pfn(unsigned long va BUG(); return; } - pmd = pmd_offset(pgd, vaddr); + pmd = pmd_offset_kernel(pgd, vaddr); if (pmd_none(*pmd)) { BUG(); return; @@ -110,7 +110,7 @@ void set_pmd_pfn(unsigned long vaddr, un printk ("set_pmd_pfn: pgd_none\n"); return; /* BUG(); */ } - pmd = pmd_offset(pgd, vaddr); + pmd = pmd_offset_kernel(pgd, vaddr); set_pmd(pmd, pfn_pmd(pfn, flags)); /* * It's enough to flush this one mapping. @@ -152,11 +152,6 @@ struct page *pte_alloc_one(struct mm_str return pte; } -void pmd_ctor(void *pmd, kmem_cache_t *cache, unsigned long flags) -{ - memset(pmd, 0, PTRS_PER_PMD*sizeof(pmd_t)); -} - /* * List of all pgd's needed for non-PAE so it can invalidate entries * in both cached and uncached pgd's; not needed for PAE since the @@ -203,6 +198,12 @@ void pgd_dtor(void *pgd, kmem_cache_t *c spin_unlock_irqrestore(&pgd_lock, flags); } +#ifdef CONFIG_HIGHPMD +#define GFP_PMD (__GFP_REPEAT|__GFP_HIGHMEM|GFP_KERNEL) +#else +#define GFP_PMD (__GFP_REPEAT|GFP_KERNEL) +#endif + pgd_t *pgd_alloc(struct mm_struct *mm) { int i; @@ -212,16 +213,17 @@ pgd_t *pgd_alloc(struct mm_struct *mm) return pgd; for (i = 0; i < USER_PTRS_PER_PGD; ++i) { - pmd_t *pmd = kmem_cache_alloc(pmd_cache, GFP_KERNEL); + struct page *pmd = alloc_page(GFP_PMD); if (!pmd) goto out_oom; - set_pgd(&pgd[i], __pgd(1 + __pa((u64)((u32)pmd)))); + clear_highpage(pmd); + set_pgd(&pgd[i], __pgd(1ULL | (u64)page_to_pfn(pmd) << PAGE_SHIFT)); } return pgd; out_oom: for (i--; i >= 0; i--) - kmem_cache_free(pmd_cache, (void *)__va(pgd_val(pgd[i])-1)); + __free_page(pgd_page(pgd[i])); kmem_cache_free(pgd_cache, pgd); return NULL; } @@ -233,7 +235,7 @@ void pgd_free(pgd_t *pgd) /* in the PAE case user pgd entries are overwritten before usage */ if (PTRS_PER_PMD > 1) for (i = 0; i < USER_PTRS_PER_PGD; ++i) - kmem_cache_free(pmd_cache, (void *)__va(pgd_val(pgd[i])-1)); + __free_page(pgd_page(pgd[i])); /* in the non-PAE case, clear_page_tables() clears user pgd entries */ kmem_cache_free(pgd_cache, pgd); } diff -prauN mm2-2.5.73-1/arch/sparc/mm/srmmu.c mm2-2.5.73-2/arch/sparc/mm/srmmu.c --- mm2-2.5.73-1/arch/sparc/mm/srmmu.c 2003-06-22 11:32:56.000000000 -0700 +++ mm2-2.5.73-2/arch/sparc/mm/srmmu.c 2003-06-28 03:11:37.000000000 -0700 @@ -2180,7 +2180,7 @@ void __init ld_mmu_srmmu(void) BTFIXUPSET_CALL(pte_pfn, srmmu_pte_pfn, BTFIXUPCALL_NORM); BTFIXUPSET_CALL(pmd_page, srmmu_pmd_page, BTFIXUPCALL_NORM); - BTFIXUPSET_CALL(pgd_page, srmmu_pgd_page, BTFIXUPCALL_NORM); + BTFIXUPSET_CALL(__pgd_page, srmmu_pgd_page, BTFIXUPCALL_NORM); BTFIXUPSET_SETHI(none_mask, 0xF0000000); diff -prauN mm2-2.5.73-1/arch/sparc/mm/sun4c.c mm2-2.5.73-2/arch/sparc/mm/sun4c.c --- mm2-2.5.73-1/arch/sparc/mm/sun4c.c 2003-06-22 11:33:06.000000000 -0700 +++ mm2-2.5.73-2/arch/sparc/mm/sun4c.c 2003-06-28 03:11:37.000000000 -0700 @@ -2252,5 +2252,5 @@ void __init ld_mmu_sun4c(void) /* These should _never_ get called with two level tables. */ BTFIXUPSET_CALL(pgd_set, sun4c_pgd_set, BTFIXUPCALL_NOP); - BTFIXUPSET_CALL(pgd_page, sun4c_pgd_page, BTFIXUPCALL_RETO0); + BTFIXUPSET_CALL(__pgd_page, sun4c_pgd_page, BTFIXUPCALL_RETO0); } diff -prauN mm2-2.5.73-1/drivers/char/drm/drm_memory.h mm2-2.5.73-2/drivers/char/drm/drm_memory.h --- mm2-2.5.73-1/drivers/char/drm/drm_memory.h 2003-06-22 11:32:35.000000000 -0700 +++ mm2-2.5.73-2/drivers/char/drm/drm_memory.h 2003-06-28 03:11:37.000000000 -0700 @@ -123,7 +123,7 @@ static inline unsigned long drm_follow_page (void *vaddr) { pgd_t *pgd = pgd_offset_k((unsigned long) vaddr); - pmd_t *pmd = pmd_offset(pgd, (unsigned long) vaddr); + pmd_t *pmd = pmd_offset_kernel(pgd, (unsigned long)vaddr); pte_t *ptep = pte_offset_kernel(pmd, (unsigned long) vaddr); return pte_pfn(*ptep) << PAGE_SHIFT; } diff -prauN mm2-2.5.73-1/fs/exec.c mm2-2.5.73-2/fs/exec.c --- mm2-2.5.73-1/fs/exec.c 2003-06-28 03:09:53.000000000 -0700 +++ mm2-2.5.73-2/fs/exec.c 2003-06-28 03:11:37.000000000 -0700 @@ -304,10 +304,10 @@ void put_dirty_page(struct task_struct * if (!pte_chain) goto out_sig; spin_lock(&tsk->mm->page_table_lock); - pmd = pmd_alloc(tsk->mm, pgd, address); + pmd = pmd_alloc_map(tsk->mm, pgd, address); if (!pmd) goto out; - pte = pte_alloc_map(tsk->mm, pmd, address); + pte = pte_alloc_map(tsk->mm, &pmd, address); if (!pte) goto out; if (!pte_none(*pte)) { @@ -319,6 +319,7 @@ void put_dirty_page(struct task_struct * set_pte(pte, pte_mkdirty(pte_mkwrite(mk_pte(page, prot)))); pte_chain = page_add_rmap(page, pte, pte_chain); pte_unmap(pte); + pmd_unmap(pmd); tsk->mm->rss++; spin_unlock(&tsk->mm->page_table_lock); @@ -326,6 +327,8 @@ void put_dirty_page(struct task_struct * pte_chain_free(pte_chain); return; out: + if (pmd) + pmd_unmap(pmd); spin_unlock(&tsk->mm->page_table_lock); out_sig: __free_page(page); diff -prauN mm2-2.5.73-1/include/asm-alpha/pgtable.h mm2-2.5.73-2/include/asm-alpha/pgtable.h --- mm2-2.5.73-1/include/asm-alpha/pgtable.h 2003-06-22 11:32:38.000000000 -0700 +++ mm2-2.5.73-2/include/asm-alpha/pgtable.h 2003-06-28 08:20:41.000000000 -0700 @@ -229,9 +229,11 @@ pmd_page_kernel(pmd_t pmd) #define pmd_page(pmd) (mem_map + ((pmd_val(pmd) & _PFN_MASK) >> 32)) #endif -extern inline unsigned long pgd_page(pgd_t pgd) +extern inline unsigned long __pgd_page(pgd_t pgd) { return PAGE_OFFSET + ((pgd_val(pgd) & _PFN_MASK) >> (32-PAGE_SHIFT)); } +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) + extern inline int pte_none(pte_t pte) { return !pte_val(pte); } extern inline int pte_present(pte_t pte) { return pte_val(pte) & _PAGE_VALID; } extern inline void pte_clear(pte_t *ptep) { pte_val(*ptep) = 0; } @@ -280,7 +282,7 @@ extern inline pte_t pte_mkyoung(pte_t pt /* Find an entry in the second-level page table.. */ extern inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address) { - return (pmd_t *) pgd_page(*dir) + ((address >> PMD_SHIFT) & (PTRS_PER_PAGE - 1)); + return (pmd_t *)__pgd_page(*dir) + ((address >> PMD_SHIFT) & (PTRS_PER_PAGE - 1)); } /* Find an entry in the third-level page table.. */ diff -prauN mm2-2.5.73-1/include/asm-arm/pgtable.h mm2-2.5.73-2/include/asm-arm/pgtable.h --- mm2-2.5.73-1/include/asm-arm/pgtable.h 2003-06-22 11:32:38.000000000 -0700 +++ mm2-2.5.73-2/include/asm-arm/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -125,6 +125,11 @@ extern struct page *empty_zero_page; /* Find an entry in the second-level page table.. */ #define pmd_offset(dir, addr) ((pmd_t *)(dir)) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) /* Find an entry in the third-level page table.. */ #define __pte_index(addr) (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) diff -prauN mm2-2.5.73-1/include/asm-arm26/pgtable.h mm2-2.5.73-2/include/asm-arm26/pgtable.h --- mm2-2.5.73-1/include/asm-arm26/pgtable.h 2003-06-22 11:32:32.000000000 -0700 +++ mm2-2.5.73-2/include/asm-arm26/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -189,6 +189,12 @@ extern struct page *empty_zero_page; #define pte_unmap(pte) do { } while (0) #define pte_unmap_nested(pte) do { } while (0) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pgd, addr) do { } while (0) +#define pmd_unmap_nested(pgd, addr) do { } while (0) + #define _PAGE_PRESENT 0x01 #define _PAGE_READONLY 0x02 diff -prauN mm2-2.5.73-1/include/asm-h8300/pgtable.h mm2-2.5.73-2/include/asm-h8300/pgtable.h --- mm2-2.5.73-1/include/asm-h8300/pgtable.h 2003-06-22 11:32:42.000000000 -0700 +++ mm2-2.5.73-2/include/asm-h8300/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -15,6 +15,11 @@ typedef pte_t *pte_addr_t; #define pgd_clear(pgdp) #define kern_addr_valid(addr) (1) #define pmd_offset(a, b) ((void *)0) +#define pmd_offset_kernel(a,b) pmd_offset(a,b) +#define pmd_offset_map(a,b) pmd_offset(a,b) +#define pmd_offset_map_nested(a,b) pmd_offset(a,b) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) #define PAGE_NONE __pgprot(0) /* these mean nothing to NO_MM */ #define PAGE_SHARED __pgprot(0) /* these mean nothing to NO_MM */ diff -prauN mm2-2.5.73-1/include/asm-i386/kmap_types.h mm2-2.5.73-2/include/asm-i386/kmap_types.h --- mm2-2.5.73-1/include/asm-i386/kmap_types.h 2003-06-22 11:33:01.000000000 -0700 +++ mm2-2.5.73-2/include/asm-i386/kmap_types.h 2003-06-28 03:11:37.000000000 -0700 @@ -17,14 +17,16 @@ D(3) KM_USER0, D(4) KM_USER1, D(5) KM_BIO_SRC_IRQ, D(6) KM_BIO_DST_IRQ, -D(7) KM_PTE0, -D(8) KM_PTE1, -D(9) KM_PTE2, -D(10) KM_IRQ0, -D(11) KM_IRQ1, -D(12) KM_SOFTIRQ0, -D(13) KM_SOFTIRQ1, -D(14) KM_TYPE_NR +D(7) KM_PMD0, +D(8) KM_PMD1, +D(9) KM_PTE0, +D(10) KM_PTE1, +D(11) KM_PTE2, +D(12) KM_IRQ0, +D(13) KM_IRQ1, +D(14) KM_SOFTIRQ0, +D(15) KM_SOFTIRQ1, +D(16) KM_TYPE_NR }; #undef D diff -prauN mm2-2.5.73-1/include/asm-i386/pgalloc.h mm2-2.5.73-2/include/asm-i386/pgalloc.h --- mm2-2.5.73-1/include/asm-i386/pgalloc.h 2003-06-22 11:32:31.000000000 -0700 +++ mm2-2.5.73-2/include/asm-i386/pgalloc.h 2003-06-28 08:06:24.000000000 -0700 @@ -46,6 +46,7 @@ static inline void pte_free(struct page */ #define pmd_alloc_one(mm, addr) ({ BUG(); ((pmd_t *)2); }) +#define pmd_alloc_one_kernel(mm, addr) ({ BUG(); ((pmd_t *)2); }) #define pmd_free(x) do { } while (0) #define __pmd_free_tlb(tlb,x) do { } while (0) #define pgd_populate(mm, pmd, pte) BUG() diff -prauN mm2-2.5.73-1/include/asm-i386/pgtable-2level.h mm2-2.5.73-2/include/asm-i386/pgtable-2level.h --- mm2-2.5.73-1/include/asm-i386/pgtable-2level.h 2003-06-22 11:32:55.000000000 -0700 +++ mm2-2.5.73-2/include/asm-i386/pgtable-2level.h 2003-06-28 03:11:37.000000000 -0700 @@ -48,13 +48,15 @@ static inline int pgd_present(pgd_t pgd) #define set_pmd(pmdptr, pmdval) (*(pmdptr) = pmdval) #define set_pgd(pgdptr, pgdval) (*(pgdptr) = pgdval) -#define pgd_page(pgd) \ -((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +#define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT) + +#define pmd_offset_map(pgd, addr) ({ (pmd_t *)(pgd); }) +#define pmd_offset_map_nested(pgd, addr) pmd_offset_map(pgd, addr) +#define pmd_offset_kernel(pgd, addr) pmd_offset_map(pgd, addr) + +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) -static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address) -{ - return (pmd_t *) dir; -} #define ptep_get_and_clear(xp) __pte(xchg(&(xp)->pte_low, 0)) #define pte_same(a, b) ((a).pte_low == (b).pte_low) #define pte_page(x) pfn_to_page(pte_pfn(x)) diff -prauN mm2-2.5.73-1/include/asm-i386/pgtable-3level.h mm2-2.5.73-2/include/asm-i386/pgtable-3level.h --- mm2-2.5.73-1/include/asm-i386/pgtable-3level.h 2003-06-28 03:09:54.000000000 -0700 +++ mm2-2.5.73-2/include/asm-i386/pgtable-3level.h 2003-06-28 08:21:14.000000000 -0700 @@ -64,12 +64,32 @@ static inline void set_pte(pte_t *ptep, */ static inline void pgd_clear (pgd_t * pgd) { } -#define pgd_page(pgd) \ -((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +static inline unsigned long pgd_pfn(pgd_t pgd) +{ + return pgd_val(pgd) >> PAGE_SHIFT; +} + +#define pgd_page(pgd) pfn_to_page(pgd_pfn(pgd)) + +#define pmd_offset_kernel(pgd, addr) \ + ((pmd_t *)__va(pgd_val(*(pgd)) & PAGE_MASK) + pmd_index(addr)) /* Find an entry in the second-level page table.. */ -#define pmd_offset(dir, address) ((pmd_t *) pgd_page(*(dir)) + \ - pmd_index(address)) +#ifdef CONFIG_HIGHPMD +#define __pmd_offset(pgd, addr, type) \ + ((pmd_t *)kmap_atomic(pgd_page(*(pgd)), type) + pmd_index(addr)) +#define __pmd_unmap(pmd, type) kunmap_atomic(pmd, type) +#else +#define __pmd_offset(pgd, addr, type) \ + ((pmd_t *)__va(pgd_val(*(pgd)) & PAGE_MASK) + pmd_index(addr)) +#define __pmd_unmap(pmd, type) do { } while (0) +#endif + +#define pmd_offset_map(pgd, addr) __pmd_offset(pgd, addr, KM_PMD0) +#define pmd_offset_map_nested(pgd, addr) __pmd_offset(pgd, addr, KM_PMD1) + +#define pmd_unmap(pmd) __pmd_unmap(pmd, KM_PMD0); +#define pmd_unmap_nested(pmd) __pmd_unmap(pmd, KM_PMD1); static inline pte_t ptep_get_and_clear(pte_t *ptep) { diff -prauN mm2-2.5.73-1/include/asm-i386/pgtable.h mm2-2.5.73-2/include/asm-i386/pgtable.h --- mm2-2.5.73-1/include/asm-i386/pgtable.h 2003-06-28 03:09:54.000000000 -0700 +++ mm2-2.5.73-2/include/asm-i386/pgtable.h 2003-06-28 08:18:44.000000000 -0700 @@ -33,11 +33,9 @@ extern unsigned long empty_zero_page[1024]; extern pgd_t swapper_pg_dir[1024]; extern kmem_cache_t *pgd_cache; -extern kmem_cache_t *pmd_cache; extern spinlock_t pgd_lock; extern struct list_head pgd_list; -void pmd_ctor(void *, kmem_cache_t *, unsigned long); void pgd_ctor(void *, kmem_cache_t *, unsigned long); void pgd_dtor(void *, kmem_cache_t *, unsigned long); void pgtable_cache_init(void); diff -prauN mm2-2.5.73-1/include/asm-ia64/pgtable.h mm2-2.5.73-2/include/asm-ia64/pgtable.h --- mm2-2.5.73-1/include/asm-ia64/pgtable.h 2003-06-22 11:32:39.000000000 -0700 +++ mm2-2.5.73-2/include/asm-ia64/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -257,7 +257,8 @@ ia64_phys_addr_valid (unsigned long addr #define pgd_bad(pgd) (!ia64_phys_addr_valid(pgd_val(pgd))) #define pgd_present(pgd) (pgd_val(pgd) != 0UL) #define pgd_clear(pgdp) (pgd_val(*(pgdp)) = 0UL) -#define pgd_page(pgd) ((unsigned long) __va(pgd_val(pgd) & _PFN_MASK)) +#define __pgd_page(pgd) ((unsigned long)__va(pgd_val(pgd) & _PFN_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) /* * The following have defined behavior only work if pte_present() is true. @@ -326,7 +327,13 @@ pgd_offset (struct mm_struct *mm, unsign /* Find an entry in the second-level page table.. */ #define pmd_offset(dir,addr) \ - ((pmd_t *) pgd_page(*(dir)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))) + ((pmd_t *)__pgd_page(*(dir)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))) + +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) /* * Find an entry in the third-level page table. This looks more complicated than it diff -prauN mm2-2.5.73-1/include/asm-m68k/motorola_pgtable.h mm2-2.5.73-2/include/asm-m68k/motorola_pgtable.h --- mm2-2.5.73-1/include/asm-m68k/motorola_pgtable.h 2003-06-22 11:32:57.000000000 -0700 +++ mm2-2.5.73-2/include/asm-m68k/motorola_pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -115,6 +115,7 @@ extern inline void pgd_set(pgd_t * pgdp, #define __pte_page(pte) ((unsigned long)__va(pte_val(pte) & PAGE_MASK)) #define __pmd_page(pmd) ((unsigned long)__va(pmd_val(pmd) & _TABLE_MASK)) #define __pgd_page(pgd) ((unsigned long)__va(pgd_val(pgd) & _TABLE_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) #define pte_none(pte) (!pte_val(pte)) diff -prauN mm2-2.5.73-1/include/asm-m68knommu/pgtable.h mm2-2.5.73-2/include/asm-m68knommu/pgtable.h --- mm2-2.5.73-1/include/asm-m68knommu/pgtable.h 2003-06-22 11:32:56.000000000 -0700 +++ mm2-2.5.73-2/include/asm-m68knommu/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -21,7 +21,12 @@ typedef pte_t *pte_addr_t; #define pgd_bad(pgd) (0) #define pgd_clear(pgdp) #define kern_addr_valid(addr) (1) -#define pmd_offset(a, b) ((void *)0) +#define pmd_offset(a, b) ((void *)0) +#define pmd_offset_kernel(a, b) pmd_offset(a, b) +#define pmd_offset_map(a, b) pmd_offset(a, b) +#define pmd_offset_map_nested(a, b) pmd_offset(a, b) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) #define PAGE_NONE __pgprot(0) #define PAGE_SHARED __pgprot(0) diff -prauN mm2-2.5.73-1/include/asm-mips64/pgtable.h mm2-2.5.73-2/include/asm-mips64/pgtable.h --- mm2-2.5.73-1/include/asm-mips64/pgtable.h 2003-06-28 03:09:55.000000000 -0700 +++ mm2-2.5.73-2/include/asm-mips64/pgtable.h 2003-06-28 03:16:19.000000000 -0700 @@ -155,11 +155,13 @@ extern pmd_t empty_bad_pmd_table[2*PAGE_ #define pmd_page(pmd) (pfn_to_page(pmd_phys(pmd) >> PAGE_SHIFT)) #define pmd_page_kernel(pmd) pmd_val(pmd) -static inline unsigned long pgd_page(pgd_t pgd) +static inline unsigned long __pgd_page(pgd_t pgd) { return pgd_val(pgd); } +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) + static inline int pte_none(pte_t pte) { return !(pte_val(pte) & ~_PAGE_GLOBAL); @@ -397,7 +399,7 @@ static inline pte_t pte_modify(pte_t pte /* Find an entry in the second-level page table.. */ static inline pmd_t *pmd_offset(pgd_t * dir, unsigned long address) { - return (pmd_t *) pgd_page(*dir) + + return (pmd_t *)__pgd_page(*dir) + ((address >> PMD_SHIFT) & (PTRS_PER_PMD - 1)); } diff -prauN mm2-2.5.73-1/include/asm-parisc/pgtable.h mm2-2.5.73-2/include/asm-parisc/pgtable.h --- mm2-2.5.73-1/include/asm-parisc/pgtable.h 2003-06-22 11:33:15.000000000 -0700 +++ mm2-2.5.73-2/include/asm-parisc/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -242,7 +242,8 @@ extern unsigned long *empty_zero_page; #ifdef __LP64__ -#define pgd_page(pgd) ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +#define __pgd_page(pgd) ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) /* For 64 bit we have three level tables */ @@ -339,11 +340,17 @@ extern inline pte_t pte_modify(pte_t pte #ifdef __LP64__ #define pmd_offset(dir,address) \ -((pmd_t *) pgd_page(*(dir)) + (((address)>>PMD_SHIFT) & (PTRS_PER_PMD-1))) +((pmd_t *)__pgd_page(*(dir)) + (((address)>>PMD_SHIFT) & (PTRS_PER_PMD-1))) #else #define pmd_offset(dir,addr) ((pmd_t *) dir) #endif +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) + /* Find an entry in the third-level page table.. */ #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1)) #define pte_offset_kernel(pmd, address) \ diff -prauN mm2-2.5.73-1/include/asm-ppc/pgtable.h mm2-2.5.73-2/include/asm-ppc/pgtable.h --- mm2-2.5.73-1/include/asm-ppc/pgtable.h 2003-06-22 11:32:37.000000000 -0700 +++ mm2-2.5.73-2/include/asm-ppc/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -370,8 +370,9 @@ static inline int pgd_bad(pgd_t pgd) { static inline int pgd_present(pgd_t pgd) { return 1; } #define pgd_clear(xp) do { } while (0) -#define pgd_page(pgd) \ +#define __pgd_page(pgd) \ ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) /* * The following only work if pte_present() is true. diff -prauN mm2-2.5.73-1/include/asm-ppc64/pgtable.h mm2-2.5.73-2/include/asm-ppc64/pgtable.h --- mm2-2.5.73-1/include/asm-ppc64/pgtable.h 2003-06-22 11:33:18.000000000 -0700 +++ mm2-2.5.73-2/include/asm-ppc64/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -190,7 +190,8 @@ extern unsigned long empty_zero_page[PAG #define pgd_bad(pgd) ((pgd_val(pgd)) == 0) #define pgd_present(pgd) (pgd_val(pgd) != 0UL) #define pgd_clear(pgdp) (pgd_val(*(pgdp)) = 0UL) -#define pgd_page(pgd) (__bpn_to_ba(pgd_val(pgd))) +#define __pgd_page(pgd) (__bpn_to_ba(pgd_val(pgd))) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) /* * Find an entry in a page-table-directory. We combine the address region @@ -203,12 +204,18 @@ extern unsigned long empty_zero_page[PAG /* Find an entry in the second-level page table.. */ #define pmd_offset(dir,addr) \ - ((pmd_t *) pgd_page(*(dir)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))) + ((pmd_t *)__pgd_page(*(dir)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))) /* Find an entry in the third-level page table.. */ #define pte_offset_kernel(dir,addr) \ ((pte_t *) pmd_page_kernel(*(dir)) + (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) + #define pte_offset_map(dir,addr) pte_offset_kernel((dir), (addr)) #define pte_offset_map_nested(dir,addr) pte_offset_kernel((dir), (addr)) #define pte_unmap(pte) do { } while(0) diff -prauN mm2-2.5.73-1/include/asm-s390/pgtable.h mm2-2.5.73-2/include/asm-s390/pgtable.h --- mm2-2.5.73-1/include/asm-s390/pgtable.h 2003-06-22 11:33:07.000000000 -0700 +++ mm2-2.5.73-2/include/asm-s390/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -613,6 +613,7 @@ static inline pte_t mk_pte_phys(unsigned /* to find an entry in a page-table-directory */ #define pgd_index(address) ((address >> PGDIR_SHIFT) & (PTRS_PER_PGD-1)) #define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address)) +#define pgd_page(pgd) virt_to_page(pgd_page_kernel(pgd)) /* to find an entry in a kernel page-table-directory */ #define pgd_offset_k(address) pgd_offset(&init_mm, address) @@ -634,6 +635,12 @@ extern inline pmd_t * pmd_offset(pgd_t * #endif /* __s390x__ */ +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) + /* Find an entry in the third-level page table.. */ #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1)) #define pte_offset_kernel(pmd, address) \ diff -prauN mm2-2.5.73-1/include/asm-sh/pgtable-2level.h mm2-2.5.73-2/include/asm-sh/pgtable-2level.h --- mm2-2.5.73-1/include/asm-sh/pgtable-2level.h 2003-06-22 11:33:32.000000000 -0700 +++ mm2-2.5.73-2/include/asm-sh/pgtable-2level.h 2003-06-28 03:11:37.000000000 -0700 @@ -48,8 +48,9 @@ static inline void pgd_clear (pgd_t * pg #define set_pmd(pmdptr, pmdval) (*(pmdptr) = pmdval) #define set_pgd(pgdptr, pgdval) (*(pgdptr) = pgdval) -#define pgd_page(pgd) \ +#define __pgd_page(pgd) \ ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address) { diff -prauN mm2-2.5.73-1/include/asm-sparc/pgtable.h mm2-2.5.73-2/include/asm-sparc/pgtable.h --- mm2-2.5.73-1/include/asm-sparc/pgtable.h 2003-06-22 11:32:56.000000000 -0700 +++ mm2-2.5.73-2/include/asm-sparc/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -202,10 +202,11 @@ extern unsigned long empty_zero_page; /* */ BTFIXUPDEF_CALL_CONST(struct page *, pmd_page, pmd_t) -BTFIXUPDEF_CALL_CONST(unsigned long, pgd_page, pgd_t) +BTFIXUPDEF_CALL_CONST(unsigned long, __pgd_page, pgd_t) #define pmd_page(pmd) BTFIXUP_CALL(pmd_page)(pmd) -#define pgd_page(pgd) BTFIXUP_CALL(pgd_page)(pgd) +#define __pgd_page(pgd) BTFIXUP_CALL(__pgd_page)(pgd) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) BTFIXUPDEF_SETHI(none_mask) BTFIXUPDEF_CALL_CONST(int, pte_present, pte_t) @@ -352,6 +353,11 @@ extern __inline__ pte_t pte_modify(pte_t /* Find an entry in the second-level page table.. */ BTFIXUPDEF_CALL(pmd_t *, pmd_offset, pgd_t *, unsigned long) #define pmd_offset(dir,addr) BTFIXUP_CALL(pmd_offset)(dir,addr) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) /* Find an entry in the third-level page table.. */ BTFIXUPDEF_CALL(pte_t *, pte_offset_kernel, pmd_t *, unsigned long) diff -prauN mm2-2.5.73-1/include/asm-sparc64/pgtable.h mm2-2.5.73-2/include/asm-sparc64/pgtable.h --- mm2-2.5.73-1/include/asm-sparc64/pgtable.h 2003-06-22 11:32:31.000000000 -0700 +++ mm2-2.5.73-2/include/asm-sparc64/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -228,7 +228,8 @@ static inline pte_t pte_modify(pte_t ori (pgd_val(*(pgdp)) = (__pa((unsigned long) (pmdp)) >> 11UL)) #define __pmd_page(pmd) ((unsigned long) __va((pmd_val(pmd)<<11UL))) #define pmd_page(pmd) virt_to_page((void *)__pmd_page(pmd)) -#define pgd_page(pgd) ((unsigned long) __va((pgd_val(pgd)<<11UL))) +#define __pgd_page(pgd) ((unsigned long) __va((pgd_val(pgd)<<11UL))) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) #define pte_none(pte) (!pte_val(pte)) #define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT) #define pte_clear(pte) (pte_val(*(pte)) = 0UL) @@ -270,8 +271,13 @@ static inline pte_t pte_modify(pte_t ori #define pgd_offset_k(address) pgd_offset(&init_mm, address) /* Find an entry in the second-level page table.. */ -#define pmd_offset(dir, address) ((pmd_t *) pgd_page(*(dir)) + \ +#define pmd_offset(dir, address) ((pmd_t *)__pgd_page(*(dir)) + \ ((address >> PMD_SHIFT) & (REAL_PTRS_PER_PMD-1))) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) /* Find an entry in the third-level page table.. */ #define pte_index(dir, address) ((pte_t *) __pmd_page(*(dir)) + \ diff -prauN mm2-2.5.73-1/include/asm-v850/pgtable.h mm2-2.5.73-2/include/asm-v850/pgtable.h --- mm2-2.5.73-1/include/asm-v850/pgtable.h 2003-06-22 11:32:58.000000000 -0700 +++ mm2-2.5.73-2/include/asm-v850/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -13,6 +13,11 @@ typedef pte_t *pte_addr_t; #define pgd_clear(pgdp) ((void)0) #define pmd_offset(a, b) ((void *)0) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) #define kern_addr_valid(addr) (1) diff -prauN mm2-2.5.73-1/include/asm-x86_64/pgtable.h mm2-2.5.73-2/include/asm-x86_64/pgtable.h --- mm2-2.5.73-1/include/asm-x86_64/pgtable.h 2003-06-28 03:09:57.000000000 -0700 +++ mm2-2.5.73-2/include/asm-x86_64/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -98,8 +98,9 @@ static inline void set_pml4(pml4_t *dst, pml4_val(*dst) = pml4_val(val); } -#define pgd_page(pgd) \ +#define __pgd_page(pgd) \ ((unsigned long) __va(pgd_val(pgd) & PHYSICAL_PAGE_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) #define ptep_get_and_clear(xp) __pte(xchg(&(xp)->pte, 0)) #define pte_same(a, b) ((a).pte == (b).pte) @@ -332,8 +333,13 @@ static inline pgd_t *current_pgd_offset_ #define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)) #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1)) -#define pmd_offset(dir, address) ((pmd_t *) pgd_page(*(dir)) + \ +#define pmd_offset(dir, address) ((pmd_t *)__pgd_page(*(dir)) + \ pmd_index(address)) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) #define pmd_none(x) (!pmd_val(x)) #define pmd_present(x) (pmd_val(x) & _PAGE_PRESENT) #define pmd_clear(xp) do { set_pmd(xp, __pmd(0)); } while (0) diff -prauN mm2-2.5.73-1/include/linux/mm.h mm2-2.5.73-2/include/linux/mm.h --- mm2-2.5.73-1/include/linux/mm.h 2003-06-28 03:09:57.000000000 -0700 +++ mm2-2.5.73-2/include/linux/mm.h 2003-06-28 03:11:37.000000000 -0700 @@ -426,8 +426,9 @@ extern void invalidate_mmap_range(struct loff_t const holelen); extern int vmtruncate(struct inode * inode, loff_t offset); extern pmd_t *FASTCALL(__pmd_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)); +pmd_t *FASTCALL(__pmd_alloc_kernel(struct mm_struct *mm, pgd_t *pmd, unsigned long address)); extern pte_t *FASTCALL(pte_alloc_kernel(struct mm_struct *mm, pmd_t *pmd, unsigned long address)); -extern pte_t *FASTCALL(pte_alloc_map(struct mm_struct *mm, pmd_t *pmd, unsigned long address)); +pte_t *FASTCALL(pte_alloc_map(struct mm_struct *mm, pmd_t **pmd, unsigned long address)); extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot); extern int handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma, unsigned long address, int write_access); extern int make_pages_present(unsigned long addr, unsigned long end); @@ -488,12 +489,11 @@ static inline int set_page_dirty(struct * inlining and the symmetry break with pte_alloc_map() that does all * of this out-of-line. */ -static inline pmd_t *pmd_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) -{ - if (pgd_none(*pgd)) - return __pmd_alloc(mm, pgd, address); - return pmd_offset(pgd, address); -} +#define pmd_alloc_map(mm, pgd, addr) \ + (pgd_none(*(pgd))? __pmd_alloc(mm,pgd,addr): pmd_offset_map(pgd,addr)) + +#define pmd_alloc_kernel(mm, pgd, addr) \ + (pgd_none(*(pgd))? __pmd_alloc_kernel(mm,pgd,addr): pmd_offset_kernel(pgd,addr)) extern void free_area_init(unsigned long * zones_size); extern void free_area_init_node(int nid, pg_data_t *pgdat, struct page *pmap, diff -prauN mm2-2.5.73-1/mm/fremap.c mm2-2.5.73-2/mm/fremap.c --- mm2-2.5.73-1/mm/fremap.c 2003-06-22 11:32:31.000000000 -0700 +++ mm2-2.5.73-2/mm/fremap.c 2003-06-28 03:11:37.000000000 -0700 @@ -67,11 +67,11 @@ int install_page(struct mm_struct *mm, s pgd = pgd_offset(mm, addr); spin_lock(&mm->page_table_lock); - pmd = pmd_alloc(mm, pgd, addr); + pmd = pmd_alloc_map(mm, pgd, addr); if (!pmd) goto err_unlock; - pte = pte_alloc_map(mm, pmd, addr); + pte = pte_alloc_map(mm, &pmd, addr); if (!pte) goto err_unlock; @@ -82,6 +82,7 @@ int install_page(struct mm_struct *mm, s set_pte(pte, mk_pte(page, prot)); pte_chain = page_add_rmap(page, pte, pte_chain); pte_unmap(pte); + pmd_unmap(pmd); if (flush) flush_tlb_page(vma, addr); update_mmu_cache(vma, addr, *pte); diff -prauN mm2-2.5.73-1/mm/memory.c mm2-2.5.73-2/mm/memory.c --- mm2-2.5.73-1/mm/memory.c 2003-06-28 03:09:58.000000000 -0700 +++ mm2-2.5.73-2/mm/memory.c 2003-06-28 08:35:17.000000000 -0700 @@ -103,6 +103,7 @@ static inline void free_one_pmd(struct m static inline void free_one_pgd(struct mmu_gather *tlb, pgd_t * dir) { pmd_t * pmd, * md, * emd; + struct page *page; if (pgd_none(*dir)) return; @@ -111,7 +112,8 @@ static inline void free_one_pgd(struct m pgd_clear(dir); return; } - pmd = pmd_offset(dir, 0); + page = pgd_page(*dir); + pmd = pmd_offset_map(dir, 0); pgd_clear(dir); /* * Beware if changing the loop below. It once used int j, @@ -128,7 +130,8 @@ static inline void free_one_pgd(struct m */ for (md = pmd, emd = pmd + PTRS_PER_PMD; md < emd; md++) free_one_pmd(tlb,md); - pmd_free_tlb(tlb, pmd); + pmd_unmap(pmd); + pmd_free_tlb(tlb, page); } /* @@ -148,30 +151,40 @@ void clear_page_tables(struct mmu_gather } while (--nr); } -pte_t * pte_alloc_map(struct mm_struct *mm, pmd_t *pmd, unsigned long address) +/* + * error return happens with pmd unmapped + */ +pte_t *pte_alloc_map(struct mm_struct *mm, pmd_t **pmd, unsigned long address) { - if (!pmd_present(*pmd)) { + if (!pmd_present(**pmd)) { + pgd_t *pgd; struct page *new; + pmd_unmap(*pmd); spin_unlock(&mm->page_table_lock); new = pte_alloc_one(mm, address); spin_lock(&mm->page_table_lock); - if (!new) + if (!new) { + *pmd = NULL; return NULL; + } + + pgd = pgd_offset(mm, address); + *pmd = pmd_offset_map(pgd, address); /* * Because we dropped the lock, we should re-check the * entry, as somebody else could have populated it.. */ - if (pmd_present(*pmd)) { + if (pmd_present(**pmd)) { pte_free(new); goto out; } pgtable_add_rmap(new, mm, address); - pmd_populate(mm, pmd, new); + pmd_populate(mm, *pmd, new); } out: - return pte_offset_map(pmd, address); + return pte_offset_map(*pmd, address); } pte_t * pte_alloc_kernel(struct mm_struct *mm, pmd_t *pmd, unsigned long address) @@ -256,11 +269,10 @@ skip_copy_pmd_range: address = (address continue; } - src_pmd = pmd_offset(src_pgd, address); - dst_pmd = pmd_alloc(dst, dst_pgd, address); + dst_pmd = pmd_alloc_map(dst, dst_pgd, address); if (!dst_pmd) goto nomem; - + src_pmd = pmd_offset_map_nested(src_pgd, address); do { pte_t * src_pte, * dst_pte; @@ -273,15 +285,20 @@ skip_copy_pmd_range: address = (address pmd_clear(src_pmd); skip_copy_pte_range: address = (address + PMD_SIZE) & PMD_MASK; - if (address >= end) + if (address >= end) { + pmd_unmap(dst_pmd); + pmd_unmap_nested(src_pmd); goto out; + } goto cont_copy_pmd_range; } - dst_pte = pte_alloc_map(dst, dst_pmd, address); + pmd_unmap_nested(src_pmd); + dst_pte = pte_alloc_map(dst, &dst_pmd, address); if (!dst_pte) goto nomem; spin_lock(&src->page_table_lock); + src_pmd = pmd_offset_map_nested(src_pgd, address); src_pte = pte_offset_map_nested(src_pmd, address); do { pte_t pte = *src_pte; @@ -348,6 +365,8 @@ skip_copy_pte_range: */ pte_unmap_nested(src_pte); pte_unmap(dst_pte); + pmd_unmap_nested(src_pmd); + pmd_unmap(dst_pmd); spin_unlock(&src->page_table_lock); spin_unlock(&dst->page_table_lock); pte_chain = pte_chain_alloc(GFP_KERNEL); @@ -355,12 +374,16 @@ skip_copy_pte_range: if (!pte_chain) goto nomem; spin_lock(&src->page_table_lock); + dst_pmd = pmd_offset_map(dst_pgd, address); + src_pmd = pmd_offset_map_nested(src_pgd, address); dst_pte = pte_offset_map(dst_pmd, address); src_pte = pte_offset_map_nested(src_pmd, address); cont_copy_pte_range_noset: address += PAGE_SIZE; if (address >= end) { + pmd_unmap(dst_pmd); + pmd_unmap_nested(src_pmd); pte_unmap_nested(src_pte); pte_unmap(dst_pte); goto out_unlock; @@ -376,6 +399,8 @@ cont_copy_pmd_range: src_pmd++; dst_pmd++; } while ((unsigned long)src_pmd & PMD_TABLE_MASK); + pmd_unmap_nested(src_pmd-1); + pmd_unmap(dst_pmd-1); } out_unlock: spin_unlock(&src->page_table_lock); @@ -451,7 +476,7 @@ zap_pmd_range(struct mmu_gather *tlb, pg pgd_clear(dir); return; } - pmd = pmd_offset(dir, address); + pmd = pmd_offset_map(dir, address); end = address + size; if (end > ((address + PGDIR_SIZE) & PGDIR_MASK)) end = ((address + PGDIR_SIZE) & PGDIR_MASK); @@ -460,6 +485,7 @@ zap_pmd_range(struct mmu_gather *tlb, pg address = (address + PMD_SIZE) & PMD_MASK; pmd++; } while (address < end); + pmd_unmap(pmd - 1); } void unmap_page_range(struct mmu_gather *tlb, struct vm_area_struct *vma, @@ -641,20 +667,27 @@ follow_page(struct mm_struct *mm, unsign if (pgd_none(*pgd) || pgd_bad(*pgd)) goto out; - pmd = pmd_offset(pgd, address); + pmd = pmd_offset_map(pgd, address); if (pmd_none(*pmd)) - goto out; - if (pmd_huge(*pmd)) - return follow_huge_pmd(mm, address, pmd, write); - if (pmd_bad(*pmd)) - goto out; + goto out_unmap; + if (pmd_bad(*pmd)) { + pmd_ERROR(*pmd); + pmd_clear(pmd); + goto out_unmap; + } + if (pmd_huge(*pmd)) { + struct page *page = follow_huge_pmd(mm, address, pmd, write); + pmd_unmap(pmd); + return page; + } ptep = pte_offset_map(pmd, address); if (!ptep) - goto out; + goto out_unmap; pte = *ptep; pte_unmap(ptep); + pmd_unmap(pmd); if (pte_present(pte)) { if (!write || (pte_write(pte) && pte_dirty(pte))) { pfn = pte_pfn(pte); @@ -665,6 +698,9 @@ follow_page(struct mm_struct *mm, unsign out: return NULL; +out_unmap: + pmd_unmap(pmd); + goto out; } /* @@ -723,7 +759,7 @@ int get_user_pages(struct task_struct *t pgd = pgd_offset_k(pg); if (!pgd) return i ? : -EFAULT; - pmd = pmd_offset(pgd, pg); + pmd = pmd_offset_kernel(pgd, pg); if (!pmd) return i ? : -EFAULT; pte = pte_offset_kernel(pmd, pg); @@ -815,8 +851,8 @@ static void zeromap_pte_range(pte_t * pt } while (address && (address < end)); } -static inline int zeromap_pmd_range(struct mm_struct *mm, pmd_t * pmd, unsigned long address, - unsigned long size, pgprot_t prot) +static inline int zeromap_pmd_range(struct mm_struct *mm, pmd_t **pmd, + unsigned long address, unsigned long size, pgprot_t prot) { unsigned long end; @@ -825,13 +861,13 @@ static inline int zeromap_pmd_range(stru if (end > PGDIR_SIZE) end = PGDIR_SIZE; do { - pte_t * pte = pte_alloc_map(mm, pmd, address); + pte_t *pte = pte_alloc_map(mm, pmd, address); if (!pte) return -ENOMEM; zeromap_pte_range(pte, address, end - address, prot); pte_unmap(pte); address = (address + PMD_SIZE) & PMD_MASK; - pmd++; + (*pmd)++; } while (address && (address < end)); return 0; } @@ -851,13 +887,14 @@ int zeromap_page_range(struct vm_area_st spin_lock(&mm->page_table_lock); do { - pmd_t *pmd = pmd_alloc(mm, dir, address); + pmd_t *pmd = pmd_alloc_map(mm, dir, address); error = -ENOMEM; if (!pmd) break; - error = zeromap_pmd_range(mm, pmd, address, end - address, prot); + error = zeromap_pmd_range(mm, &pmd, address, end - address, prot); if (error) break; + pmd_unmap(pmd - 1); address = (address + PGDIR_SIZE) & PGDIR_MASK; dir++; } while (address && (address < end)); @@ -892,8 +929,9 @@ static inline void remap_pte_range(pte_t } while (address && (address < end)); } -static inline int remap_pmd_range(struct mm_struct *mm, pmd_t * pmd, unsigned long address, unsigned long size, - unsigned long phys_addr, pgprot_t prot) +static inline int remap_pmd_range(struct mm_struct *mm, pmd_t **pmd, + unsigned long address, unsigned long size, + unsigned long phys_addr, pgprot_t prot) { unsigned long base, end; @@ -904,13 +942,13 @@ static inline int remap_pmd_range(struct end = PGDIR_SIZE; phys_addr -= address; do { - pte_t * pte = pte_alloc_map(mm, pmd, base + address); + pte_t *pte = pte_alloc_map(mm, pmd, base + address); if (!pte) return -ENOMEM; remap_pte_range(pte, base + address, end - address, address + phys_addr, prot); pte_unmap(pte); address = (address + PMD_SIZE) & PMD_MASK; - pmd++; + (*pmd)++; } while (address && (address < end)); return 0; } @@ -932,13 +970,14 @@ int remap_page_range(struct vm_area_stru spin_lock(&mm->page_table_lock); do { - pmd_t *pmd = pmd_alloc(mm, dir, from); + pmd_t *pmd = pmd_alloc_map(mm, dir, from); error = -ENOMEM; if (!pmd) break; - error = remap_pmd_range(mm, pmd, from, end - from, phys_addr + from, prot); + error = remap_pmd_range(mm, &pmd, from, end - from, phys_addr + from, prot); if (error) break; + pmd_unmap(pmd - 1); from = (from + PGDIR_SIZE) & PGDIR_MASK; dir++; } while (from && (from < end)); @@ -1008,6 +1047,7 @@ static int do_wp_page(struct mm_struct * * data, but for the moment just pretend this is OOM. */ pte_unmap(page_table); + pmd_unmap(pmd); printk(KERN_ERR "do_wp_page: bogus page at address %08lx\n", address); goto oom; @@ -1022,11 +1062,13 @@ static int do_wp_page(struct mm_struct * establish_pte(vma, address, page_table, pte_mkyoung(pte_mkdirty(pte_mkwrite(pte)))); pte_unmap(page_table); + pmd_unmap(pmd); ret = VM_FAULT_MINOR; goto out; } } pte_unmap(page_table); + pmd_unmap(pmd); /* * Ok, we need to copy. Oh, well.. @@ -1046,6 +1088,7 @@ static int do_wp_page(struct mm_struct * * Re-check the pte - we dropped the lock */ spin_lock(&mm->page_table_lock); + pmd = pmd_offset_map(pgd_offset(mm, address), address); page_table = pte_offset_map(pmd, address); if (pte_same(*page_table, pte)) { if (PageReserved(old_page)) @@ -1059,6 +1102,7 @@ static int do_wp_page(struct mm_struct * new_page = old_page; } pte_unmap(page_table); + pmd_unmap(pmd); page_cache_release(new_page); page_cache_release(old_page); ret = VM_FAULT_MINOR; @@ -1227,6 +1271,7 @@ static int do_swap_page(struct mm_struct struct pte_chain *pte_chain = NULL; pte_unmap(page_table); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); page = lookup_swap_cache(entry); if (!page) { @@ -1238,12 +1283,14 @@ static int do_swap_page(struct mm_struct * we released the page table lock. */ spin_lock(&mm->page_table_lock); + pmd = pmd_offset_map(pgd_offset(mm, address), address); page_table = pte_offset_map(pmd, address); if (pte_same(*page_table, orig_pte)) ret = VM_FAULT_OOM; else ret = VM_FAULT_MINOR; pte_unmap(page_table); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); goto out; } @@ -1266,9 +1313,11 @@ static int do_swap_page(struct mm_struct * released the page table lock. */ spin_lock(&mm->page_table_lock); + pmd = pmd_offset_map(pgd_offset(mm, address), address); page_table = pte_offset_map(pmd, address); if (!pte_same(*page_table, orig_pte)) { pte_unmap(page_table); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); unlock_page(page); page_cache_release(page); @@ -1294,6 +1343,7 @@ static int do_swap_page(struct mm_struct /* No need to invalidate - it was non-present before */ update_mmu_cache(vma, address, pte); + pmd_unmap(pmd); pte_unmap(page_table); spin_unlock(&mm->page_table_lock); out: @@ -1319,11 +1369,13 @@ do_anonymous_page(struct mm_struct *mm, pte_chain = pte_chain_alloc(GFP_ATOMIC); if (!pte_chain) { pte_unmap(page_table); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); pte_chain = pte_chain_alloc(GFP_KERNEL); if (!pte_chain) goto no_mem; spin_lock(&mm->page_table_lock); + pmd = pmd_offset_map(pgd_offset(mm, addr), addr); page_table = pte_offset_map(pmd, addr); } @@ -1334,6 +1386,7 @@ do_anonymous_page(struct mm_struct *mm, if (write_access) { /* Allocate our own private page. */ pte_unmap(page_table); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); page = alloc_page(GFP_HIGHUSER); @@ -1342,9 +1395,11 @@ do_anonymous_page(struct mm_struct *mm, clear_user_highpage(page, addr); spin_lock(&mm->page_table_lock); + pmd = pmd_offset_map(pgd_offset(mm, addr), addr); page_table = pte_offset_map(pmd, addr); if (!pte_none(*page_table)) { + pmd_unmap(pmd); pte_unmap(page_table); page_cache_release(page); spin_unlock(&mm->page_table_lock); @@ -1360,6 +1415,7 @@ do_anonymous_page(struct mm_struct *mm, set_pte(page_table, entry); /* ignores ZERO_PAGE */ pte_chain = page_add_rmap(page, page_table, pte_chain); + pmd_unmap(pmd); pte_unmap(page_table); /* No need to invalidate - it was non-present before */ @@ -1402,6 +1458,7 @@ do_no_page(struct mm_struct *mm, struct return do_anonymous_page(mm, vma, page_table, pmd, write_access, address); pte_unmap(page_table); + pmd_unmap(pmd); mapping = vma->vm_file->f_dentry->d_inode->i_mapping; sequence = atomic_read(&mapping->truncate_count); @@ -1446,6 +1503,7 @@ retry: page_cache_release(new_page); goto retry; } + pmd = pmd_offset_map(pgd_offset(mm, address), address); page_table = pte_offset_map(pmd, address); /* @@ -1468,9 +1526,11 @@ retry: set_pte(page_table, entry); pte_chain = page_add_rmap(new_page, page_table, pte_chain); pte_unmap(page_table); + pmd_unmap(pmd); } else { /* One of our sibling threads was faster, back out. */ pte_unmap(page_table); + pmd_unmap(pmd); page_cache_release(new_page); spin_unlock(&mm->page_table_lock); ret = VM_FAULT_MINOR; @@ -1514,6 +1574,7 @@ static int do_file_page(struct mm_struct pgoff = pte_to_pgoff(*pte); pte_unmap(pte); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); err = vma->vm_ops->populate(vma, address & PAGE_MASK, PAGE_SIZE, vma->vm_page_prot, pgoff, 0); @@ -1574,6 +1635,7 @@ static inline int handle_pte_fault(struc entry = pte_mkyoung(entry); establish_pte(vma, address, pte, entry); pte_unmap(pte); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); return VM_FAULT_MINOR; } @@ -1600,10 +1662,10 @@ int handle_mm_fault(struct mm_struct *mm * and the SMP-safe atomic PTE updates. */ spin_lock(&mm->page_table_lock); - pmd = pmd_alloc(mm, pgd, address); + pmd = pmd_alloc_map(mm, pgd, address); if (pmd) { - pte_t * pte = pte_alloc_map(mm, pmd, address); + pte_t *pte = pte_alloc_map(mm, &pmd, address); if (pte) return handle_pte_fault(mm, vma, address, write_access, pte, pmd); } @@ -1640,7 +1702,30 @@ pmd_t *__pmd_alloc(struct mm_struct *mm, } pgd_populate(mm, pgd, new); out: - return pmd_offset(pgd, address); + return pmd_offset_map(pgd, address); +} + +pmd_t *__pmd_alloc_kernel(struct mm_struct *mm, pgd_t *pgd, unsigned long address) +{ + pmd_t *new; + + spin_unlock(&mm->page_table_lock); + new = pmd_alloc_one_kernel(mm, address); + spin_lock(&mm->page_table_lock); + if (!new) + return NULL; + + /* + * Because we dropped the lock, we should re-check the + * entry, as somebody else could have populated it.. + */ + if (pgd_present(*pgd)) { + pmd_free(new); + goto out; + } + pgd_populate(mm, pgd, new); +out: + return pmd_offset_kernel(pgd, address); } int make_pages_present(unsigned long addr, unsigned long end) @@ -1672,7 +1757,7 @@ struct page * vmalloc_to_page(void * vma pte_t *ptep, pte; if (!pgd_none(*pgd)) { - pmd = pmd_offset(pgd, addr); + pmd = pmd_offset_map(pgd, addr); if (!pmd_none(*pmd)) { preempt_disable(); ptep = pte_offset_map(pmd, addr); @@ -1682,6 +1767,7 @@ struct page * vmalloc_to_page(void * vma pte_unmap(ptep); preempt_enable(); } + pmd_unmap(pmd); } return page; } diff -prauN mm2-2.5.73-1/mm/mprotect.c mm2-2.5.73-2/mm/mprotect.c --- mm2-2.5.73-1/mm/mprotect.c 2003-06-28 03:09:58.000000000 -0700 +++ mm2-2.5.73-2/mm/mprotect.c 2003-06-28 03:11:37.000000000 -0700 @@ -73,7 +73,7 @@ change_pmd_range(pgd_t *pgd, unsigned lo pgd_clear(pgd); return; } - pmd = pmd_offset(pgd, address); + pmd = pmd_offset_map(pgd, address); address &= ~PGDIR_MASK; end = address + size; if (end > PGDIR_SIZE) @@ -83,6 +83,7 @@ change_pmd_range(pgd_t *pgd, unsigned lo address = (address + PMD_SIZE) & PMD_MASK; pmd++; } while (address && (address < end)); + pmd_unmap(pmd - 1); } static void diff -prauN mm2-2.5.73-1/mm/mremap.c mm2-2.5.73-2/mm/mremap.c --- mm2-2.5.73-1/mm/mremap.c 2003-06-28 03:09:58.000000000 -0700 +++ mm2-2.5.73-2/mm/mremap.c 2003-06-28 03:11:37.000000000 -0700 @@ -38,7 +38,7 @@ static pte_t *get_one_pte_map_nested(str goto end; } - pmd = pmd_offset(pgd, addr); + pmd = pmd_offset_map_nested(pgd, addr); if (pmd_none(*pmd)) goto end; if (pmd_bad(*pmd)) { @@ -53,6 +53,7 @@ static pte_t *get_one_pte_map_nested(str pte = NULL; } end: + pmd_unmap_nested(pmd); return pte; } @@ -61,12 +62,15 @@ static inline int page_table_present(str { pgd_t *pgd; pmd_t *pmd; + int ret; pgd = pgd_offset(mm, addr); if (pgd_none(*pgd)) return 0; - pmd = pmd_offset(pgd, addr); - return pmd_present(*pmd); + pmd = pmd_offset_map(pgd, addr); + ret = pmd_present(*pmd); + pmd_unmap(pmd); + return ret != 0; } #else #define page_table_present(mm, addr) (1) @@ -77,9 +81,10 @@ static inline pte_t *alloc_one_pte_map(s pmd_t *pmd; pte_t *pte = NULL; - pmd = pmd_alloc(mm, pgd_offset(mm, addr), addr); + pmd = pmd_alloc_map(mm, pgd_offset(mm, addr), addr); if (pmd) - pte = pte_alloc_map(mm, pmd, addr); + pte = pte_alloc_map(mm, &pmd, addr); + pmd_unmap(pmd); return pte; } diff -prauN mm2-2.5.73-1/mm/msync.c mm2-2.5.73-2/mm/msync.c --- mm2-2.5.73-1/mm/msync.c 2003-06-22 11:32:42.000000000 -0700 +++ mm2-2.5.73-2/mm/msync.c 2003-06-28 03:11:37.000000000 -0700 @@ -82,7 +82,7 @@ static inline int filemap_sync_pmd_range pgd_clear(pgd); return 0; } - pmd = pmd_offset(pgd, address); + pmd = pmd_offset_map(pgd, address); if ((address & PGDIR_MASK) != (end & PGDIR_MASK)) end = (address & PGDIR_MASK) + PGDIR_SIZE; error = 0; @@ -91,6 +91,7 @@ static inline int filemap_sync_pmd_range address = (address + PMD_SIZE) & PMD_MASK; pmd++; } while (address && (address < end)); + pmd_unmap(pmd - 1); return error; } diff -prauN mm2-2.5.73-1/mm/slab.c mm2-2.5.73-2/mm/slab.c --- mm2-2.5.73-1/mm/slab.c 2003-06-28 03:09:58.000000000 -0700 +++ mm2-2.5.73-2/mm/slab.c 2003-06-28 03:27:00.000000000 -0700 @@ -2717,7 +2717,7 @@ void ptrinfo(unsigned long addr) printk("No pgd.\n"); break; } - pmd = pmd_offset(pgd, addr); + pmd = pmd_offset_kernel(pgd, addr); if (pmd_none(*pmd)) { printk("No pmd.\n"); break; diff -prauN mm2-2.5.73-1/mm/swapfile.c mm2-2.5.73-2/mm/swapfile.c --- mm2-2.5.73-1/mm/swapfile.c 2003-06-28 03:09:58.000000000 -0700 +++ mm2-2.5.73-2/mm/swapfile.c 2003-06-28 03:11:37.000000000 -0700 @@ -448,7 +448,7 @@ static int unuse_pgd(struct vm_area_stru pgd_clear(dir); return 0; } - pmd = pmd_offset(dir, address); + pmd = pmd_offset_map(dir, address); offset = address & PGDIR_MASK; address &= ~PGDIR_MASK; end = address + size; @@ -463,6 +463,7 @@ static int unuse_pgd(struct vm_area_stru address = (address + PMD_SIZE) & PMD_MASK; pmd++; } while (address && (address < end)); + pmd_unmap(pmd - 1); return 0; } diff -prauN mm2-2.5.73-1/mm/vmalloc.c mm2-2.5.73-2/mm/vmalloc.c --- mm2-2.5.73-1/mm/vmalloc.c 2003-06-22 11:32:56.000000000 -0700 +++ mm2-2.5.73-2/mm/vmalloc.c 2003-06-28 03:11:37.000000000 -0700 @@ -70,7 +70,7 @@ static void unmap_area_pmd(pgd_t *dir, u return; } - pmd = pmd_offset(dir, address); + pmd = pmd_offset_kernel(dir, address); address &= ~PGDIR_MASK; end = address + size; if (end > PGDIR_SIZE) @@ -159,7 +159,7 @@ int map_vm_area(struct vm_struct *area, dir = pgd_offset_k(address); spin_lock(&init_mm.page_table_lock); do { - pmd_t *pmd = pmd_alloc(&init_mm, dir, address); + pmd_t *pmd = pmd_alloc_kernel(&init_mm, dir, address); if (!pmd) { err = -ENOMEM; break; ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-06-28 15:54 ` William Lee Irwin III 0 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-06-28 15:54 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-mm On Fri, Jun 27, 2003 at 08:21:30PM -0700, Andrew Morton wrote: > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/ > Just bits and pieces. Here's highpmd. This allocates L2 pagetables from highmem, decreasing the per-process lowmem overhead on CONFIG_HIGHMEM64G from 20KB to 8KB. Some attempts were made to update non-i386 architectures to the new API's, though they're entirely untested. It's been tested for a while in -wli on i386 machines, both lowmem and highmem boxen. -- wli diff -prauN mm2-2.5.73-1/arch/i386/Kconfig mm2-2.5.73-2/arch/i386/Kconfig --- mm2-2.5.73-1/arch/i386/Kconfig 2003-06-28 03:09:46.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/Kconfig 2003-06-28 03:11:37.000000000 -0700 @@ -765,6 +765,15 @@ config HIGHPTE low memory. Setting this option will put user-space page table entries in high memory. +config HIGHPMD + bool "Allocate 2nd-level pagetables from highmem" + depends on HIGHMEM64G + help + The VM uses one pmd entry for each pagetable page of physical + memory allocated. For systems with extreme amounts of highmem, + this cannot be tolerated. Setting this option will put + userspace 2nd-level pagetables in highmem. + config MATH_EMULATION bool "Math emulation" ---help--- diff -prauN mm2-2.5.73-1/arch/i386/kernel/vm86.c mm2-2.5.73-2/arch/i386/kernel/vm86.c --- mm2-2.5.73-1/arch/i386/kernel/vm86.c 2003-06-22 11:32:33.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/kernel/vm86.c 2003-06-28 03:11:37.000000000 -0700 @@ -144,12 +144,14 @@ static void mark_screen_rdonly(struct ta pgd_clear(pgd); goto out; } - pmd = pmd_offset(pgd, 0xA0000); - if (pmd_none(*pmd)) + pmd = pmd_offset_map(pgd, 0xA0000); + if (pmd_none(*pmd)) { + pmd_unmap(pmd); goto out; - if (pmd_bad(*pmd)) { + } else if (pmd_bad(*pmd)) { pmd_ERROR(*pmd); pmd_clear(pmd); + pmd_unmap(pmd); goto out; } pte = mapped = pte_offset_map(pmd, 0xA0000); @@ -159,6 +161,7 @@ static void mark_screen_rdonly(struct ta pte++; } pte_unmap(mapped); + pmd_unmap(pmd); out: spin_unlock(&tsk->mm->page_table_lock); preempt_enable(); diff -prauN mm2-2.5.73-1/arch/i386/mm/fault.c mm2-2.5.73-2/arch/i386/mm/fault.c --- mm2-2.5.73-1/arch/i386/mm/fault.c 2003-06-28 03:09:46.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/fault.c 2003-06-28 03:28:31.000000000 -0700 @@ -253,6 +253,7 @@ no_context: printk(" printing eip:\n"); printk("%08lx\n", regs->eip); asm("movl %%cr3,%0":"=r" (page)); +#ifndef CONFIG_HIGHPMD /* Oh boy. Error reporting is going to blow major goats. */ page = ((unsigned long *) __va(page))[address >> 22]; printk(KERN_ALERT "*pde = %08lx\n", page); /* @@ -268,7 +269,14 @@ no_context: page = ((unsigned long *) __va(page))[address >> PAGE_SHIFT]; printk(KERN_ALERT "*pte = %08lx\n", page); } -#endif +#endif /* !CONFIG_HIGHPTE */ +#else /* CONFIG_HIGHPMD */ + printk(KERN_ALERT "%%cr3 = 0x%lx\n", page); + /* Mask off flag bits. It should end up 32B-aligned. */ + page &= ~(PTRS_PER_PGD*sizeof(pgd_t) - 1); + printk(KERN_ALERT "*pdpte = 0x%Lx\n", + pgd_val(((pgd_t *)__va(page))[address >> PGDIR_SHIFT])); +#endif /* CONFIG_HIGHPMD */ die("Oops", regs, error_code); bust_spinlocks(0); do_exit(SIGKILL); @@ -336,8 +344,8 @@ vmalloc_fault: * and redundant with the set_pmd() on non-PAE. */ - pmd = pmd_offset(pgd, address); - pmd_k = pmd_offset(pgd_k, address); + pmd = pmd_offset_kernel(pgd, address); + pmd_k = pmd_offset_kernel(pgd_k, address); if (!pmd_present(*pmd_k)) goto no_context; set_pmd(pmd, *pmd_k); diff -prauN mm2-2.5.73-1/arch/i386/mm/hugetlbpage.c mm2-2.5.73-2/arch/i386/mm/hugetlbpage.c --- mm2-2.5.73-1/arch/i386/mm/hugetlbpage.c 2003-06-22 11:33:17.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/hugetlbpage.c 2003-06-28 03:11:37.000000000 -0700 @@ -87,8 +87,8 @@ static pte_t *huge_pte_alloc(struct mm_s pmd_t *pmd = NULL; pgd = pgd_offset(mm, addr); - pmd = pmd_alloc(mm, pgd, addr); - return (pte_t *) pmd; + pmd = pmd_alloc_map(mm, pgd, addr); + return (pte_t *)pmd; } static pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr) @@ -97,8 +97,8 @@ static pte_t *huge_pte_offset(struct mm_ pmd_t *pmd = NULL; pgd = pgd_offset(mm, addr); - pmd = pmd_offset(pgd, addr); - return (pte_t *) pmd; + pmd = pmd_offset_map(pgd, addr); + return (pte_t *)pmd; } static void set_huge_pte(struct mm_struct *mm, struct vm_area_struct *vma, struct page *page, pte_t * page_table, int write_access) @@ -145,6 +145,8 @@ int copy_hugetlb_page_range(struct mm_st ptepage = pte_page(entry); get_page(ptepage); set_pte(dst_pte, entry); + pmd_unmap(dst_pte); + pmd_unmap_nested(src_pte); dst->rss += (HPAGE_SIZE / PAGE_SIZE); addr += HPAGE_SIZE; } @@ -182,6 +184,7 @@ follow_hugetlb_page(struct mm_struct *mm get_page(page); pages[i] = page; + pmd_unmap(pte); } if (vmas) @@ -271,6 +274,7 @@ follow_huge_pmd(struct mm_struct *mm, un page += ((address & ~HPAGE_MASK) >> PAGE_SHIFT); get_page(page); } + pmd_unmap(pmd); return page; } #endif @@ -314,6 +318,7 @@ void unmap_hugepage_range(struct vm_area page = pte_page(*pte); huge_page_release(page); pte_clear(pte); + pmd_unmap(pte); } mm->rss -= (end - start) >> PAGE_SHIFT; flush_tlb_range(vma, start, end); @@ -358,16 +363,19 @@ int hugetlb_prefault(struct address_spac page = alloc_hugetlb_page(); if (!page) { ret = -ENOMEM; + pmd_unmap(pte); goto out; } ret = add_to_page_cache(page, mapping, idx, GFP_ATOMIC); unlock_page(page); if (ret) { free_huge_page(page); + pmd_unmap(pte); goto out; } } set_huge_pte(mm, vma, page, pte, vma->vm_flags & VM_WRITE); + pmd_unmap(pte); } out: spin_unlock(&mm->page_table_lock); diff -prauN mm2-2.5.73-1/arch/i386/mm/init.c mm2-2.5.73-2/arch/i386/mm/init.c --- mm2-2.5.73-1/arch/i386/mm/init.c 2003-06-28 03:09:46.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/init.c 2003-06-28 03:22:00.000000000 -0700 @@ -59,10 +59,10 @@ static pmd_t * __init one_md_table_init( #ifdef CONFIG_X86_PAE pmd_table = (pmd_t *) alloc_bootmem_low_pages(PAGE_SIZE); set_pgd(pgd, __pgd(__pa(pmd_table) | _PAGE_PRESENT)); - if (pmd_table != pmd_offset(pgd, 0)) + if (pmd_table != pmd_offset_kernel(pgd, 0)) BUG(); #else - pmd_table = pmd_offset(pgd, 0); + pmd_table = pmd_offset_kernel(pgd, 0); #endif return pmd_table; @@ -113,7 +113,7 @@ static void __init page_table_range_init if (pgd_none(*pgd)) one_md_table_init(pgd); - pmd = pmd_offset(pgd, vaddr); + pmd = pmd_offset_kernel(pgd, vaddr); for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); pmd++, pmd_idx++) { if (pmd_none(*pmd)) one_page_table_init(pmd); @@ -194,7 +194,7 @@ pte_t *kmap_pte; pgprot_t kmap_prot; #define kmap_get_fixmap_pte(vaddr) \ - pte_offset_kernel(pmd_offset(pgd_offset_k(vaddr), (vaddr)), (vaddr)) + pte_offset_kernel(pmd_offset_kernel(pgd_offset_k(vaddr), (vaddr)), (vaddr)) void __init kmap_init(void) { @@ -218,7 +218,7 @@ void __init permanent_kmaps_init(pgd_t * page_table_range_init(vaddr, vaddr + PAGE_SIZE*LAST_PKMAP, pgd_base); pgd = swapper_pg_dir + pgd_index(vaddr); - pmd = pmd_offset(pgd, vaddr); + pmd = pmd_offset_kernel(pgd, vaddr); pte = pte_offset_kernel(pmd, vaddr); pkmap_page_table = pte; } @@ -513,20 +513,9 @@ void __init mem_init(void) } kmem_cache_t *pgd_cache; -kmem_cache_t *pmd_cache; void __init pgtable_cache_init(void) { - if (PTRS_PER_PMD > 1) { - pmd_cache = kmem_cache_create("pmd", - PTRS_PER_PMD*sizeof(pmd_t), - 0, - SLAB_HWCACHE_ALIGN | SLAB_MUST_HWCACHE_ALIGN, - pmd_ctor, - NULL); - if (!pmd_cache) - panic("pgtable_cache_init(): cannot create pmd cache"); - } pgd_cache = kmem_cache_create("pgd", PTRS_PER_PGD*sizeof(pgd_t), 0, diff -prauN mm2-2.5.73-1/arch/i386/mm/ioremap.c mm2-2.5.73-2/arch/i386/mm/ioremap.c --- mm2-2.5.73-1/arch/i386/mm/ioremap.c 2003-06-22 11:32:38.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/ioremap.c 2003-06-28 03:11:37.000000000 -0700 @@ -82,7 +82,7 @@ static int remap_area_pages(unsigned lon spin_lock(&init_mm.page_table_lock); do { pmd_t *pmd; - pmd = pmd_alloc(&init_mm, dir, address); + pmd = pmd_alloc_kernel(&init_mm, dir, address); error = -ENOMEM; if (!pmd) break; diff -prauN mm2-2.5.73-1/arch/i386/mm/pageattr.c mm2-2.5.73-2/arch/i386/mm/pageattr.c --- mm2-2.5.73-1/arch/i386/mm/pageattr.c 2003-06-28 03:09:46.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/pageattr.c 2003-06-28 03:12:16.000000000 -0700 @@ -23,7 +23,7 @@ static inline pte_t *lookup_address(unsi pmd_t *pmd; if (pgd_none(*pgd)) return NULL; - pmd = pmd_offset(pgd, address); + pmd = pmd_offset_kernel(pgd, address); if (pmd_none(*pmd)) return NULL; if (pmd_large(*pmd)) @@ -79,7 +79,7 @@ static void set_pmd_pte(pte_t *kpte, uns pgd_t *pgd; pmd_t *pmd; pgd = (pgd_t *)page_address(page) + pgd_index(address); - pmd = pmd_offset(pgd, address); + pmd = pmd_offset_kernel(pgd, address); set_pte_atomic((pte_t *)pmd, pte); } spin_unlock_irqrestore(&pgd_lock, flags); @@ -92,7 +92,7 @@ static void set_pmd_pte(pte_t *kpte, uns static inline void revert_page(struct page *kpte_page, unsigned long address) { pte_t *linear = (pte_t *) - pmd_offset(pgd_offset(&init_mm, address), address); + pmd_offset_kernel(pgd_offset_k(address), address); set_pmd_pte(linear, address, pfn_pte((__pa(address) & LARGE_PAGE_MASK) >> PAGE_SHIFT, PAGE_KERNEL_LARGE)); diff -prauN mm2-2.5.73-1/arch/i386/mm/pgtable.c mm2-2.5.73-2/arch/i386/mm/pgtable.c --- mm2-2.5.73-1/arch/i386/mm/pgtable.c 2003-06-28 03:09:46.000000000 -0700 +++ mm2-2.5.73-2/arch/i386/mm/pgtable.c 2003-06-28 08:20:29.000000000 -0700 @@ -70,7 +70,7 @@ static void set_pte_pfn(unsigned long va BUG(); return; } - pmd = pmd_offset(pgd, vaddr); + pmd = pmd_offset_kernel(pgd, vaddr); if (pmd_none(*pmd)) { BUG(); return; @@ -110,7 +110,7 @@ void set_pmd_pfn(unsigned long vaddr, un printk ("set_pmd_pfn: pgd_none\n"); return; /* BUG(); */ } - pmd = pmd_offset(pgd, vaddr); + pmd = pmd_offset_kernel(pgd, vaddr); set_pmd(pmd, pfn_pmd(pfn, flags)); /* * It's enough to flush this one mapping. @@ -152,11 +152,6 @@ struct page *pte_alloc_one(struct mm_str return pte; } -void pmd_ctor(void *pmd, kmem_cache_t *cache, unsigned long flags) -{ - memset(pmd, 0, PTRS_PER_PMD*sizeof(pmd_t)); -} - /* * List of all pgd's needed for non-PAE so it can invalidate entries * in both cached and uncached pgd's; not needed for PAE since the @@ -203,6 +198,12 @@ void pgd_dtor(void *pgd, kmem_cache_t *c spin_unlock_irqrestore(&pgd_lock, flags); } +#ifdef CONFIG_HIGHPMD +#define GFP_PMD (__GFP_REPEAT|__GFP_HIGHMEM|GFP_KERNEL) +#else +#define GFP_PMD (__GFP_REPEAT|GFP_KERNEL) +#endif + pgd_t *pgd_alloc(struct mm_struct *mm) { int i; @@ -212,16 +213,17 @@ pgd_t *pgd_alloc(struct mm_struct *mm) return pgd; for (i = 0; i < USER_PTRS_PER_PGD; ++i) { - pmd_t *pmd = kmem_cache_alloc(pmd_cache, GFP_KERNEL); + struct page *pmd = alloc_page(GFP_PMD); if (!pmd) goto out_oom; - set_pgd(&pgd[i], __pgd(1 + __pa((u64)((u32)pmd)))); + clear_highpage(pmd); + set_pgd(&pgd[i], __pgd(1ULL | (u64)page_to_pfn(pmd) << PAGE_SHIFT)); } return pgd; out_oom: for (i--; i >= 0; i--) - kmem_cache_free(pmd_cache, (void *)__va(pgd_val(pgd[i])-1)); + __free_page(pgd_page(pgd[i])); kmem_cache_free(pgd_cache, pgd); return NULL; } @@ -233,7 +235,7 @@ void pgd_free(pgd_t *pgd) /* in the PAE case user pgd entries are overwritten before usage */ if (PTRS_PER_PMD > 1) for (i = 0; i < USER_PTRS_PER_PGD; ++i) - kmem_cache_free(pmd_cache, (void *)__va(pgd_val(pgd[i])-1)); + __free_page(pgd_page(pgd[i])); /* in the non-PAE case, clear_page_tables() clears user pgd entries */ kmem_cache_free(pgd_cache, pgd); } diff -prauN mm2-2.5.73-1/arch/sparc/mm/srmmu.c mm2-2.5.73-2/arch/sparc/mm/srmmu.c --- mm2-2.5.73-1/arch/sparc/mm/srmmu.c 2003-06-22 11:32:56.000000000 -0700 +++ mm2-2.5.73-2/arch/sparc/mm/srmmu.c 2003-06-28 03:11:37.000000000 -0700 @@ -2180,7 +2180,7 @@ void __init ld_mmu_srmmu(void) BTFIXUPSET_CALL(pte_pfn, srmmu_pte_pfn, BTFIXUPCALL_NORM); BTFIXUPSET_CALL(pmd_page, srmmu_pmd_page, BTFIXUPCALL_NORM); - BTFIXUPSET_CALL(pgd_page, srmmu_pgd_page, BTFIXUPCALL_NORM); + BTFIXUPSET_CALL(__pgd_page, srmmu_pgd_page, BTFIXUPCALL_NORM); BTFIXUPSET_SETHI(none_mask, 0xF0000000); diff -prauN mm2-2.5.73-1/arch/sparc/mm/sun4c.c mm2-2.5.73-2/arch/sparc/mm/sun4c.c --- mm2-2.5.73-1/arch/sparc/mm/sun4c.c 2003-06-22 11:33:06.000000000 -0700 +++ mm2-2.5.73-2/arch/sparc/mm/sun4c.c 2003-06-28 03:11:37.000000000 -0700 @@ -2252,5 +2252,5 @@ void __init ld_mmu_sun4c(void) /* These should _never_ get called with two level tables. */ BTFIXUPSET_CALL(pgd_set, sun4c_pgd_set, BTFIXUPCALL_NOP); - BTFIXUPSET_CALL(pgd_page, sun4c_pgd_page, BTFIXUPCALL_RETO0); + BTFIXUPSET_CALL(__pgd_page, sun4c_pgd_page, BTFIXUPCALL_RETO0); } diff -prauN mm2-2.5.73-1/drivers/char/drm/drm_memory.h mm2-2.5.73-2/drivers/char/drm/drm_memory.h --- mm2-2.5.73-1/drivers/char/drm/drm_memory.h 2003-06-22 11:32:35.000000000 -0700 +++ mm2-2.5.73-2/drivers/char/drm/drm_memory.h 2003-06-28 03:11:37.000000000 -0700 @@ -123,7 +123,7 @@ static inline unsigned long drm_follow_page (void *vaddr) { pgd_t *pgd = pgd_offset_k((unsigned long) vaddr); - pmd_t *pmd = pmd_offset(pgd, (unsigned long) vaddr); + pmd_t *pmd = pmd_offset_kernel(pgd, (unsigned long)vaddr); pte_t *ptep = pte_offset_kernel(pmd, (unsigned long) vaddr); return pte_pfn(*ptep) << PAGE_SHIFT; } diff -prauN mm2-2.5.73-1/fs/exec.c mm2-2.5.73-2/fs/exec.c --- mm2-2.5.73-1/fs/exec.c 2003-06-28 03:09:53.000000000 -0700 +++ mm2-2.5.73-2/fs/exec.c 2003-06-28 03:11:37.000000000 -0700 @@ -304,10 +304,10 @@ void put_dirty_page(struct task_struct * if (!pte_chain) goto out_sig; spin_lock(&tsk->mm->page_table_lock); - pmd = pmd_alloc(tsk->mm, pgd, address); + pmd = pmd_alloc_map(tsk->mm, pgd, address); if (!pmd) goto out; - pte = pte_alloc_map(tsk->mm, pmd, address); + pte = pte_alloc_map(tsk->mm, &pmd, address); if (!pte) goto out; if (!pte_none(*pte)) { @@ -319,6 +319,7 @@ void put_dirty_page(struct task_struct * set_pte(pte, pte_mkdirty(pte_mkwrite(mk_pte(page, prot)))); pte_chain = page_add_rmap(page, pte, pte_chain); pte_unmap(pte); + pmd_unmap(pmd); tsk->mm->rss++; spin_unlock(&tsk->mm->page_table_lock); @@ -326,6 +327,8 @@ void put_dirty_page(struct task_struct * pte_chain_free(pte_chain); return; out: + if (pmd) + pmd_unmap(pmd); spin_unlock(&tsk->mm->page_table_lock); out_sig: __free_page(page); diff -prauN mm2-2.5.73-1/include/asm-alpha/pgtable.h mm2-2.5.73-2/include/asm-alpha/pgtable.h --- mm2-2.5.73-1/include/asm-alpha/pgtable.h 2003-06-22 11:32:38.000000000 -0700 +++ mm2-2.5.73-2/include/asm-alpha/pgtable.h 2003-06-28 08:20:41.000000000 -0700 @@ -229,9 +229,11 @@ pmd_page_kernel(pmd_t pmd) #define pmd_page(pmd) (mem_map + ((pmd_val(pmd) & _PFN_MASK) >> 32)) #endif -extern inline unsigned long pgd_page(pgd_t pgd) +extern inline unsigned long __pgd_page(pgd_t pgd) { return PAGE_OFFSET + ((pgd_val(pgd) & _PFN_MASK) >> (32-PAGE_SHIFT)); } +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) + extern inline int pte_none(pte_t pte) { return !pte_val(pte); } extern inline int pte_present(pte_t pte) { return pte_val(pte) & _PAGE_VALID; } extern inline void pte_clear(pte_t *ptep) { pte_val(*ptep) = 0; } @@ -280,7 +282,7 @@ extern inline pte_t pte_mkyoung(pte_t pt /* Find an entry in the second-level page table.. */ extern inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address) { - return (pmd_t *) pgd_page(*dir) + ((address >> PMD_SHIFT) & (PTRS_PER_PAGE - 1)); + return (pmd_t *)__pgd_page(*dir) + ((address >> PMD_SHIFT) & (PTRS_PER_PAGE - 1)); } /* Find an entry in the third-level page table.. */ diff -prauN mm2-2.5.73-1/include/asm-arm/pgtable.h mm2-2.5.73-2/include/asm-arm/pgtable.h --- mm2-2.5.73-1/include/asm-arm/pgtable.h 2003-06-22 11:32:38.000000000 -0700 +++ mm2-2.5.73-2/include/asm-arm/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -125,6 +125,11 @@ extern struct page *empty_zero_page; /* Find an entry in the second-level page table.. */ #define pmd_offset(dir, addr) ((pmd_t *)(dir)) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) /* Find an entry in the third-level page table.. */ #define __pte_index(addr) (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)) diff -prauN mm2-2.5.73-1/include/asm-arm26/pgtable.h mm2-2.5.73-2/include/asm-arm26/pgtable.h --- mm2-2.5.73-1/include/asm-arm26/pgtable.h 2003-06-22 11:32:32.000000000 -0700 +++ mm2-2.5.73-2/include/asm-arm26/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -189,6 +189,12 @@ extern struct page *empty_zero_page; #define pte_unmap(pte) do { } while (0) #define pte_unmap_nested(pte) do { } while (0) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pgd, addr) do { } while (0) +#define pmd_unmap_nested(pgd, addr) do { } while (0) + #define _PAGE_PRESENT 0x01 #define _PAGE_READONLY 0x02 diff -prauN mm2-2.5.73-1/include/asm-h8300/pgtable.h mm2-2.5.73-2/include/asm-h8300/pgtable.h --- mm2-2.5.73-1/include/asm-h8300/pgtable.h 2003-06-22 11:32:42.000000000 -0700 +++ mm2-2.5.73-2/include/asm-h8300/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -15,6 +15,11 @@ typedef pte_t *pte_addr_t; #define pgd_clear(pgdp) #define kern_addr_valid(addr) (1) #define pmd_offset(a, b) ((void *)0) +#define pmd_offset_kernel(a,b) pmd_offset(a,b) +#define pmd_offset_map(a,b) pmd_offset(a,b) +#define pmd_offset_map_nested(a,b) pmd_offset(a,b) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) #define PAGE_NONE __pgprot(0) /* these mean nothing to NO_MM */ #define PAGE_SHARED __pgprot(0) /* these mean nothing to NO_MM */ diff -prauN mm2-2.5.73-1/include/asm-i386/kmap_types.h mm2-2.5.73-2/include/asm-i386/kmap_types.h --- mm2-2.5.73-1/include/asm-i386/kmap_types.h 2003-06-22 11:33:01.000000000 -0700 +++ mm2-2.5.73-2/include/asm-i386/kmap_types.h 2003-06-28 03:11:37.000000000 -0700 @@ -17,14 +17,16 @@ D(3) KM_USER0, D(4) KM_USER1, D(5) KM_BIO_SRC_IRQ, D(6) KM_BIO_DST_IRQ, -D(7) KM_PTE0, -D(8) KM_PTE1, -D(9) KM_PTE2, -D(10) KM_IRQ0, -D(11) KM_IRQ1, -D(12) KM_SOFTIRQ0, -D(13) KM_SOFTIRQ1, -D(14) KM_TYPE_NR +D(7) KM_PMD0, +D(8) KM_PMD1, +D(9) KM_PTE0, +D(10) KM_PTE1, +D(11) KM_PTE2, +D(12) KM_IRQ0, +D(13) KM_IRQ1, +D(14) KM_SOFTIRQ0, +D(15) KM_SOFTIRQ1, +D(16) KM_TYPE_NR }; #undef D diff -prauN mm2-2.5.73-1/include/asm-i386/pgalloc.h mm2-2.5.73-2/include/asm-i386/pgalloc.h --- mm2-2.5.73-1/include/asm-i386/pgalloc.h 2003-06-22 11:32:31.000000000 -0700 +++ mm2-2.5.73-2/include/asm-i386/pgalloc.h 2003-06-28 08:06:24.000000000 -0700 @@ -46,6 +46,7 @@ static inline void pte_free(struct page */ #define pmd_alloc_one(mm, addr) ({ BUG(); ((pmd_t *)2); }) +#define pmd_alloc_one_kernel(mm, addr) ({ BUG(); ((pmd_t *)2); }) #define pmd_free(x) do { } while (0) #define __pmd_free_tlb(tlb,x) do { } while (0) #define pgd_populate(mm, pmd, pte) BUG() diff -prauN mm2-2.5.73-1/include/asm-i386/pgtable-2level.h mm2-2.5.73-2/include/asm-i386/pgtable-2level.h --- mm2-2.5.73-1/include/asm-i386/pgtable-2level.h 2003-06-22 11:32:55.000000000 -0700 +++ mm2-2.5.73-2/include/asm-i386/pgtable-2level.h 2003-06-28 03:11:37.000000000 -0700 @@ -48,13 +48,15 @@ static inline int pgd_present(pgd_t pgd) #define set_pmd(pmdptr, pmdval) (*(pmdptr) = pmdval) #define set_pgd(pgdptr, pgdval) (*(pgdptr) = pgdval) -#define pgd_page(pgd) \ -((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +#define pgd_page(pgd) pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT) + +#define pmd_offset_map(pgd, addr) ({ (pmd_t *)(pgd); }) +#define pmd_offset_map_nested(pgd, addr) pmd_offset_map(pgd, addr) +#define pmd_offset_kernel(pgd, addr) pmd_offset_map(pgd, addr) + +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) -static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address) -{ - return (pmd_t *) dir; -} #define ptep_get_and_clear(xp) __pte(xchg(&(xp)->pte_low, 0)) #define pte_same(a, b) ((a).pte_low == (b).pte_low) #define pte_page(x) pfn_to_page(pte_pfn(x)) diff -prauN mm2-2.5.73-1/include/asm-i386/pgtable-3level.h mm2-2.5.73-2/include/asm-i386/pgtable-3level.h --- mm2-2.5.73-1/include/asm-i386/pgtable-3level.h 2003-06-28 03:09:54.000000000 -0700 +++ mm2-2.5.73-2/include/asm-i386/pgtable-3level.h 2003-06-28 08:21:14.000000000 -0700 @@ -64,12 +64,32 @@ static inline void set_pte(pte_t *ptep, */ static inline void pgd_clear (pgd_t * pgd) { } -#define pgd_page(pgd) \ -((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +static inline unsigned long pgd_pfn(pgd_t pgd) +{ + return pgd_val(pgd) >> PAGE_SHIFT; +} + +#define pgd_page(pgd) pfn_to_page(pgd_pfn(pgd)) + +#define pmd_offset_kernel(pgd, addr) \ + ((pmd_t *)__va(pgd_val(*(pgd)) & PAGE_MASK) + pmd_index(addr)) /* Find an entry in the second-level page table.. */ -#define pmd_offset(dir, address) ((pmd_t *) pgd_page(*(dir)) + \ - pmd_index(address)) +#ifdef CONFIG_HIGHPMD +#define __pmd_offset(pgd, addr, type) \ + ((pmd_t *)kmap_atomic(pgd_page(*(pgd)), type) + pmd_index(addr)) +#define __pmd_unmap(pmd, type) kunmap_atomic(pmd, type) +#else +#define __pmd_offset(pgd, addr, type) \ + ((pmd_t *)__va(pgd_val(*(pgd)) & PAGE_MASK) + pmd_index(addr)) +#define __pmd_unmap(pmd, type) do { } while (0) +#endif + +#define pmd_offset_map(pgd, addr) __pmd_offset(pgd, addr, KM_PMD0) +#define pmd_offset_map_nested(pgd, addr) __pmd_offset(pgd, addr, KM_PMD1) + +#define pmd_unmap(pmd) __pmd_unmap(pmd, KM_PMD0); +#define pmd_unmap_nested(pmd) __pmd_unmap(pmd, KM_PMD1); static inline pte_t ptep_get_and_clear(pte_t *ptep) { diff -prauN mm2-2.5.73-1/include/asm-i386/pgtable.h mm2-2.5.73-2/include/asm-i386/pgtable.h --- mm2-2.5.73-1/include/asm-i386/pgtable.h 2003-06-28 03:09:54.000000000 -0700 +++ mm2-2.5.73-2/include/asm-i386/pgtable.h 2003-06-28 08:18:44.000000000 -0700 @@ -33,11 +33,9 @@ extern unsigned long empty_zero_page[1024]; extern pgd_t swapper_pg_dir[1024]; extern kmem_cache_t *pgd_cache; -extern kmem_cache_t *pmd_cache; extern spinlock_t pgd_lock; extern struct list_head pgd_list; -void pmd_ctor(void *, kmem_cache_t *, unsigned long); void pgd_ctor(void *, kmem_cache_t *, unsigned long); void pgd_dtor(void *, kmem_cache_t *, unsigned long); void pgtable_cache_init(void); diff -prauN mm2-2.5.73-1/include/asm-ia64/pgtable.h mm2-2.5.73-2/include/asm-ia64/pgtable.h --- mm2-2.5.73-1/include/asm-ia64/pgtable.h 2003-06-22 11:32:39.000000000 -0700 +++ mm2-2.5.73-2/include/asm-ia64/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -257,7 +257,8 @@ ia64_phys_addr_valid (unsigned long addr #define pgd_bad(pgd) (!ia64_phys_addr_valid(pgd_val(pgd))) #define pgd_present(pgd) (pgd_val(pgd) != 0UL) #define pgd_clear(pgdp) (pgd_val(*(pgdp)) = 0UL) -#define pgd_page(pgd) ((unsigned long) __va(pgd_val(pgd) & _PFN_MASK)) +#define __pgd_page(pgd) ((unsigned long)__va(pgd_val(pgd) & _PFN_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) /* * The following have defined behavior only work if pte_present() is true. @@ -326,7 +327,13 @@ pgd_offset (struct mm_struct *mm, unsign /* Find an entry in the second-level page table.. */ #define pmd_offset(dir,addr) \ - ((pmd_t *) pgd_page(*(dir)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))) + ((pmd_t *)__pgd_page(*(dir)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))) + +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) /* * Find an entry in the third-level page table. This looks more complicated than it diff -prauN mm2-2.5.73-1/include/asm-m68k/motorola_pgtable.h mm2-2.5.73-2/include/asm-m68k/motorola_pgtable.h --- mm2-2.5.73-1/include/asm-m68k/motorola_pgtable.h 2003-06-22 11:32:57.000000000 -0700 +++ mm2-2.5.73-2/include/asm-m68k/motorola_pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -115,6 +115,7 @@ extern inline void pgd_set(pgd_t * pgdp, #define __pte_page(pte) ((unsigned long)__va(pte_val(pte) & PAGE_MASK)) #define __pmd_page(pmd) ((unsigned long)__va(pmd_val(pmd) & _TABLE_MASK)) #define __pgd_page(pgd) ((unsigned long)__va(pgd_val(pgd) & _TABLE_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) #define pte_none(pte) (!pte_val(pte)) diff -prauN mm2-2.5.73-1/include/asm-m68knommu/pgtable.h mm2-2.5.73-2/include/asm-m68knommu/pgtable.h --- mm2-2.5.73-1/include/asm-m68knommu/pgtable.h 2003-06-22 11:32:56.000000000 -0700 +++ mm2-2.5.73-2/include/asm-m68knommu/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -21,7 +21,12 @@ typedef pte_t *pte_addr_t; #define pgd_bad(pgd) (0) #define pgd_clear(pgdp) #define kern_addr_valid(addr) (1) -#define pmd_offset(a, b) ((void *)0) +#define pmd_offset(a, b) ((void *)0) +#define pmd_offset_kernel(a, b) pmd_offset(a, b) +#define pmd_offset_map(a, b) pmd_offset(a, b) +#define pmd_offset_map_nested(a, b) pmd_offset(a, b) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) #define PAGE_NONE __pgprot(0) #define PAGE_SHARED __pgprot(0) diff -prauN mm2-2.5.73-1/include/asm-mips64/pgtable.h mm2-2.5.73-2/include/asm-mips64/pgtable.h --- mm2-2.5.73-1/include/asm-mips64/pgtable.h 2003-06-28 03:09:55.000000000 -0700 +++ mm2-2.5.73-2/include/asm-mips64/pgtable.h 2003-06-28 03:16:19.000000000 -0700 @@ -155,11 +155,13 @@ extern pmd_t empty_bad_pmd_table[2*PAGE_ #define pmd_page(pmd) (pfn_to_page(pmd_phys(pmd) >> PAGE_SHIFT)) #define pmd_page_kernel(pmd) pmd_val(pmd) -static inline unsigned long pgd_page(pgd_t pgd) +static inline unsigned long __pgd_page(pgd_t pgd) { return pgd_val(pgd); } +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) + static inline int pte_none(pte_t pte) { return !(pte_val(pte) & ~_PAGE_GLOBAL); @@ -397,7 +399,7 @@ static inline pte_t pte_modify(pte_t pte /* Find an entry in the second-level page table.. */ static inline pmd_t *pmd_offset(pgd_t * dir, unsigned long address) { - return (pmd_t *) pgd_page(*dir) + + return (pmd_t *)__pgd_page(*dir) + ((address >> PMD_SHIFT) & (PTRS_PER_PMD - 1)); } diff -prauN mm2-2.5.73-1/include/asm-parisc/pgtable.h mm2-2.5.73-2/include/asm-parisc/pgtable.h --- mm2-2.5.73-1/include/asm-parisc/pgtable.h 2003-06-22 11:33:15.000000000 -0700 +++ mm2-2.5.73-2/include/asm-parisc/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -242,7 +242,8 @@ extern unsigned long *empty_zero_page; #ifdef __LP64__ -#define pgd_page(pgd) ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +#define __pgd_page(pgd) ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) /* For 64 bit we have three level tables */ @@ -339,11 +340,17 @@ extern inline pte_t pte_modify(pte_t pte #ifdef __LP64__ #define pmd_offset(dir,address) \ -((pmd_t *) pgd_page(*(dir)) + (((address)>>PMD_SHIFT) & (PTRS_PER_PMD-1))) +((pmd_t *)__pgd_page(*(dir)) + (((address)>>PMD_SHIFT) & (PTRS_PER_PMD-1))) #else #define pmd_offset(dir,addr) ((pmd_t *) dir) #endif +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) + /* Find an entry in the third-level page table.. */ #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1)) #define pte_offset_kernel(pmd, address) \ diff -prauN mm2-2.5.73-1/include/asm-ppc/pgtable.h mm2-2.5.73-2/include/asm-ppc/pgtable.h --- mm2-2.5.73-1/include/asm-ppc/pgtable.h 2003-06-22 11:32:37.000000000 -0700 +++ mm2-2.5.73-2/include/asm-ppc/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -370,8 +370,9 @@ static inline int pgd_bad(pgd_t pgd) { static inline int pgd_present(pgd_t pgd) { return 1; } #define pgd_clear(xp) do { } while (0) -#define pgd_page(pgd) \ +#define __pgd_page(pgd) \ ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) /* * The following only work if pte_present() is true. diff -prauN mm2-2.5.73-1/include/asm-ppc64/pgtable.h mm2-2.5.73-2/include/asm-ppc64/pgtable.h --- mm2-2.5.73-1/include/asm-ppc64/pgtable.h 2003-06-22 11:33:18.000000000 -0700 +++ mm2-2.5.73-2/include/asm-ppc64/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -190,7 +190,8 @@ extern unsigned long empty_zero_page[PAG #define pgd_bad(pgd) ((pgd_val(pgd)) == 0) #define pgd_present(pgd) (pgd_val(pgd) != 0UL) #define pgd_clear(pgdp) (pgd_val(*(pgdp)) = 0UL) -#define pgd_page(pgd) (__bpn_to_ba(pgd_val(pgd))) +#define __pgd_page(pgd) (__bpn_to_ba(pgd_val(pgd))) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) /* * Find an entry in a page-table-directory. We combine the address region @@ -203,12 +204,18 @@ extern unsigned long empty_zero_page[PAG /* Find an entry in the second-level page table.. */ #define pmd_offset(dir,addr) \ - ((pmd_t *) pgd_page(*(dir)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))) + ((pmd_t *)__pgd_page(*(dir)) + (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))) /* Find an entry in the third-level page table.. */ #define pte_offset_kernel(dir,addr) \ ((pte_t *) pmd_page_kernel(*(dir)) + (((addr) >> PAGE_SHIFT) & (PTRS_PER_PTE - 1))) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) + #define pte_offset_map(dir,addr) pte_offset_kernel((dir), (addr)) #define pte_offset_map_nested(dir,addr) pte_offset_kernel((dir), (addr)) #define pte_unmap(pte) do { } while(0) diff -prauN mm2-2.5.73-1/include/asm-s390/pgtable.h mm2-2.5.73-2/include/asm-s390/pgtable.h --- mm2-2.5.73-1/include/asm-s390/pgtable.h 2003-06-22 11:33:07.000000000 -0700 +++ mm2-2.5.73-2/include/asm-s390/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -613,6 +613,7 @@ static inline pte_t mk_pte_phys(unsigned /* to find an entry in a page-table-directory */ #define pgd_index(address) ((address >> PGDIR_SHIFT) & (PTRS_PER_PGD-1)) #define pgd_offset(mm, address) ((mm)->pgd+pgd_index(address)) +#define pgd_page(pgd) virt_to_page(pgd_page_kernel(pgd)) /* to find an entry in a kernel page-table-directory */ #define pgd_offset_k(address) pgd_offset(&init_mm, address) @@ -634,6 +635,12 @@ extern inline pmd_t * pmd_offset(pgd_t * #endif /* __s390x__ */ +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) + /* Find an entry in the third-level page table.. */ #define pte_index(address) (((address) >> PAGE_SHIFT) & (PTRS_PER_PTE-1)) #define pte_offset_kernel(pmd, address) \ diff -prauN mm2-2.5.73-1/include/asm-sh/pgtable-2level.h mm2-2.5.73-2/include/asm-sh/pgtable-2level.h --- mm2-2.5.73-1/include/asm-sh/pgtable-2level.h 2003-06-22 11:33:32.000000000 -0700 +++ mm2-2.5.73-2/include/asm-sh/pgtable-2level.h 2003-06-28 03:11:37.000000000 -0700 @@ -48,8 +48,9 @@ static inline void pgd_clear (pgd_t * pg #define set_pmd(pmdptr, pmdval) (*(pmdptr) = pmdval) #define set_pgd(pgdptr, pgdval) (*(pgdptr) = pgdval) -#define pgd_page(pgd) \ +#define __pgd_page(pgd) \ ((unsigned long) __va(pgd_val(pgd) & PAGE_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) static inline pmd_t * pmd_offset(pgd_t * dir, unsigned long address) { diff -prauN mm2-2.5.73-1/include/asm-sparc/pgtable.h mm2-2.5.73-2/include/asm-sparc/pgtable.h --- mm2-2.5.73-1/include/asm-sparc/pgtable.h 2003-06-22 11:32:56.000000000 -0700 +++ mm2-2.5.73-2/include/asm-sparc/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -202,10 +202,11 @@ extern unsigned long empty_zero_page; /* */ BTFIXUPDEF_CALL_CONST(struct page *, pmd_page, pmd_t) -BTFIXUPDEF_CALL_CONST(unsigned long, pgd_page, pgd_t) +BTFIXUPDEF_CALL_CONST(unsigned long, __pgd_page, pgd_t) #define pmd_page(pmd) BTFIXUP_CALL(pmd_page)(pmd) -#define pgd_page(pgd) BTFIXUP_CALL(pgd_page)(pgd) +#define __pgd_page(pgd) BTFIXUP_CALL(__pgd_page)(pgd) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) BTFIXUPDEF_SETHI(none_mask) BTFIXUPDEF_CALL_CONST(int, pte_present, pte_t) @@ -352,6 +353,11 @@ extern __inline__ pte_t pte_modify(pte_t /* Find an entry in the second-level page table.. */ BTFIXUPDEF_CALL(pmd_t *, pmd_offset, pgd_t *, unsigned long) #define pmd_offset(dir,addr) BTFIXUP_CALL(pmd_offset)(dir,addr) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) /* Find an entry in the third-level page table.. */ BTFIXUPDEF_CALL(pte_t *, pte_offset_kernel, pmd_t *, unsigned long) diff -prauN mm2-2.5.73-1/include/asm-sparc64/pgtable.h mm2-2.5.73-2/include/asm-sparc64/pgtable.h --- mm2-2.5.73-1/include/asm-sparc64/pgtable.h 2003-06-22 11:32:31.000000000 -0700 +++ mm2-2.5.73-2/include/asm-sparc64/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -228,7 +228,8 @@ static inline pte_t pte_modify(pte_t ori (pgd_val(*(pgdp)) = (__pa((unsigned long) (pmdp)) >> 11UL)) #define __pmd_page(pmd) ((unsigned long) __va((pmd_val(pmd)<<11UL))) #define pmd_page(pmd) virt_to_page((void *)__pmd_page(pmd)) -#define pgd_page(pgd) ((unsigned long) __va((pgd_val(pgd)<<11UL))) +#define __pgd_page(pgd) ((unsigned long) __va((pgd_val(pgd)<<11UL))) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) #define pte_none(pte) (!pte_val(pte)) #define pte_present(pte) (pte_val(pte) & _PAGE_PRESENT) #define pte_clear(pte) (pte_val(*(pte)) = 0UL) @@ -270,8 +271,13 @@ static inline pte_t pte_modify(pte_t ori #define pgd_offset_k(address) pgd_offset(&init_mm, address) /* Find an entry in the second-level page table.. */ -#define pmd_offset(dir, address) ((pmd_t *) pgd_page(*(dir)) + \ +#define pmd_offset(dir, address) ((pmd_t *)__pgd_page(*(dir)) + \ ((address >> PMD_SHIFT) & (REAL_PTRS_PER_PMD-1))) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) /* Find an entry in the third-level page table.. */ #define pte_index(dir, address) ((pte_t *) __pmd_page(*(dir)) + \ diff -prauN mm2-2.5.73-1/include/asm-v850/pgtable.h mm2-2.5.73-2/include/asm-v850/pgtable.h --- mm2-2.5.73-1/include/asm-v850/pgtable.h 2003-06-22 11:32:58.000000000 -0700 +++ mm2-2.5.73-2/include/asm-v850/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -13,6 +13,11 @@ typedef pte_t *pte_addr_t; #define pgd_clear(pgdp) ((void)0) #define pmd_offset(a, b) ((void *)0) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) #define kern_addr_valid(addr) (1) diff -prauN mm2-2.5.73-1/include/asm-x86_64/pgtable.h mm2-2.5.73-2/include/asm-x86_64/pgtable.h --- mm2-2.5.73-1/include/asm-x86_64/pgtable.h 2003-06-28 03:09:57.000000000 -0700 +++ mm2-2.5.73-2/include/asm-x86_64/pgtable.h 2003-06-28 03:11:37.000000000 -0700 @@ -98,8 +98,9 @@ static inline void set_pml4(pml4_t *dst, pml4_val(*dst) = pml4_val(val); } -#define pgd_page(pgd) \ +#define __pgd_page(pgd) \ ((unsigned long) __va(pgd_val(pgd) & PHYSICAL_PAGE_MASK)) +#define pgd_page(pgd) virt_to_page(__pgd_page(pgd)) #define ptep_get_and_clear(xp) __pte(xchg(&(xp)->pte, 0)) #define pte_same(a, b) ((a).pte == (b).pte) @@ -332,8 +333,13 @@ static inline pgd_t *current_pgd_offset_ #define pmd_page(pmd) (pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)) #define pmd_index(address) (((address) >> PMD_SHIFT) & (PTRS_PER_PMD-1)) -#define pmd_offset(dir, address) ((pmd_t *) pgd_page(*(dir)) + \ +#define pmd_offset(dir, address) ((pmd_t *)__pgd_page(*(dir)) + \ pmd_index(address)) +#define pmd_offset_kernel(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map(pgd, addr) pmd_offset(pgd, addr) +#define pmd_offset_map_nested(pgd, addr) pmd_offset(pgd, addr) +#define pmd_unmap(pmd) do { } while (0) +#define pmd_unmap_nested(pmd) do { } while (0) #define pmd_none(x) (!pmd_val(x)) #define pmd_present(x) (pmd_val(x) & _PAGE_PRESENT) #define pmd_clear(xp) do { set_pmd(xp, __pmd(0)); } while (0) diff -prauN mm2-2.5.73-1/include/linux/mm.h mm2-2.5.73-2/include/linux/mm.h --- mm2-2.5.73-1/include/linux/mm.h 2003-06-28 03:09:57.000000000 -0700 +++ mm2-2.5.73-2/include/linux/mm.h 2003-06-28 03:11:37.000000000 -0700 @@ -426,8 +426,9 @@ extern void invalidate_mmap_range(struct loff_t const holelen); extern int vmtruncate(struct inode * inode, loff_t offset); extern pmd_t *FASTCALL(__pmd_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)); +pmd_t *FASTCALL(__pmd_alloc_kernel(struct mm_struct *mm, pgd_t *pmd, unsigned long address)); extern pte_t *FASTCALL(pte_alloc_kernel(struct mm_struct *mm, pmd_t *pmd, unsigned long address)); -extern pte_t *FASTCALL(pte_alloc_map(struct mm_struct *mm, pmd_t *pmd, unsigned long address)); +pte_t *FASTCALL(pte_alloc_map(struct mm_struct *mm, pmd_t **pmd, unsigned long address)); extern int install_page(struct mm_struct *mm, struct vm_area_struct *vma, unsigned long addr, struct page *page, pgprot_t prot); extern int handle_mm_fault(struct mm_struct *mm,struct vm_area_struct *vma, unsigned long address, int write_access); extern int make_pages_present(unsigned long addr, unsigned long end); @@ -488,12 +489,11 @@ static inline int set_page_dirty(struct * inlining and the symmetry break with pte_alloc_map() that does all * of this out-of-line. */ -static inline pmd_t *pmd_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) -{ - if (pgd_none(*pgd)) - return __pmd_alloc(mm, pgd, address); - return pmd_offset(pgd, address); -} +#define pmd_alloc_map(mm, pgd, addr) \ + (pgd_none(*(pgd))? __pmd_alloc(mm,pgd,addr): pmd_offset_map(pgd,addr)) + +#define pmd_alloc_kernel(mm, pgd, addr) \ + (pgd_none(*(pgd))? __pmd_alloc_kernel(mm,pgd,addr): pmd_offset_kernel(pgd,addr)) extern void free_area_init(unsigned long * zones_size); extern void free_area_init_node(int nid, pg_data_t *pgdat, struct page *pmap, diff -prauN mm2-2.5.73-1/mm/fremap.c mm2-2.5.73-2/mm/fremap.c --- mm2-2.5.73-1/mm/fremap.c 2003-06-22 11:32:31.000000000 -0700 +++ mm2-2.5.73-2/mm/fremap.c 2003-06-28 03:11:37.000000000 -0700 @@ -67,11 +67,11 @@ int install_page(struct mm_struct *mm, s pgd = pgd_offset(mm, addr); spin_lock(&mm->page_table_lock); - pmd = pmd_alloc(mm, pgd, addr); + pmd = pmd_alloc_map(mm, pgd, addr); if (!pmd) goto err_unlock; - pte = pte_alloc_map(mm, pmd, addr); + pte = pte_alloc_map(mm, &pmd, addr); if (!pte) goto err_unlock; @@ -82,6 +82,7 @@ int install_page(struct mm_struct *mm, s set_pte(pte, mk_pte(page, prot)); pte_chain = page_add_rmap(page, pte, pte_chain); pte_unmap(pte); + pmd_unmap(pmd); if (flush) flush_tlb_page(vma, addr); update_mmu_cache(vma, addr, *pte); diff -prauN mm2-2.5.73-1/mm/memory.c mm2-2.5.73-2/mm/memory.c --- mm2-2.5.73-1/mm/memory.c 2003-06-28 03:09:58.000000000 -0700 +++ mm2-2.5.73-2/mm/memory.c 2003-06-28 08:35:17.000000000 -0700 @@ -103,6 +103,7 @@ static inline void free_one_pmd(struct m static inline void free_one_pgd(struct mmu_gather *tlb, pgd_t * dir) { pmd_t * pmd, * md, * emd; + struct page *page; if (pgd_none(*dir)) return; @@ -111,7 +112,8 @@ static inline void free_one_pgd(struct m pgd_clear(dir); return; } - pmd = pmd_offset(dir, 0); + page = pgd_page(*dir); + pmd = pmd_offset_map(dir, 0); pgd_clear(dir); /* * Beware if changing the loop below. It once used int j, @@ -128,7 +130,8 @@ static inline void free_one_pgd(struct m */ for (md = pmd, emd = pmd + PTRS_PER_PMD; md < emd; md++) free_one_pmd(tlb,md); - pmd_free_tlb(tlb, pmd); + pmd_unmap(pmd); + pmd_free_tlb(tlb, page); } /* @@ -148,30 +151,40 @@ void clear_page_tables(struct mmu_gather } while (--nr); } -pte_t * pte_alloc_map(struct mm_struct *mm, pmd_t *pmd, unsigned long address) +/* + * error return happens with pmd unmapped + */ +pte_t *pte_alloc_map(struct mm_struct *mm, pmd_t **pmd, unsigned long address) { - if (!pmd_present(*pmd)) { + if (!pmd_present(**pmd)) { + pgd_t *pgd; struct page *new; + pmd_unmap(*pmd); spin_unlock(&mm->page_table_lock); new = pte_alloc_one(mm, address); spin_lock(&mm->page_table_lock); - if (!new) + if (!new) { + *pmd = NULL; return NULL; + } + + pgd = pgd_offset(mm, address); + *pmd = pmd_offset_map(pgd, address); /* * Because we dropped the lock, we should re-check the * entry, as somebody else could have populated it.. */ - if (pmd_present(*pmd)) { + if (pmd_present(**pmd)) { pte_free(new); goto out; } pgtable_add_rmap(new, mm, address); - pmd_populate(mm, pmd, new); + pmd_populate(mm, *pmd, new); } out: - return pte_offset_map(pmd, address); + return pte_offset_map(*pmd, address); } pte_t * pte_alloc_kernel(struct mm_struct *mm, pmd_t *pmd, unsigned long address) @@ -256,11 +269,10 @@ skip_copy_pmd_range: address = (address continue; } - src_pmd = pmd_offset(src_pgd, address); - dst_pmd = pmd_alloc(dst, dst_pgd, address); + dst_pmd = pmd_alloc_map(dst, dst_pgd, address); if (!dst_pmd) goto nomem; - + src_pmd = pmd_offset_map_nested(src_pgd, address); do { pte_t * src_pte, * dst_pte; @@ -273,15 +285,20 @@ skip_copy_pmd_range: address = (address pmd_clear(src_pmd); skip_copy_pte_range: address = (address + PMD_SIZE) & PMD_MASK; - if (address >= end) + if (address >= end) { + pmd_unmap(dst_pmd); + pmd_unmap_nested(src_pmd); goto out; + } goto cont_copy_pmd_range; } - dst_pte = pte_alloc_map(dst, dst_pmd, address); + pmd_unmap_nested(src_pmd); + dst_pte = pte_alloc_map(dst, &dst_pmd, address); if (!dst_pte) goto nomem; spin_lock(&src->page_table_lock); + src_pmd = pmd_offset_map_nested(src_pgd, address); src_pte = pte_offset_map_nested(src_pmd, address); do { pte_t pte = *src_pte; @@ -348,6 +365,8 @@ skip_copy_pte_range: */ pte_unmap_nested(src_pte); pte_unmap(dst_pte); + pmd_unmap_nested(src_pmd); + pmd_unmap(dst_pmd); spin_unlock(&src->page_table_lock); spin_unlock(&dst->page_table_lock); pte_chain = pte_chain_alloc(GFP_KERNEL); @@ -355,12 +374,16 @@ skip_copy_pte_range: if (!pte_chain) goto nomem; spin_lock(&src->page_table_lock); + dst_pmd = pmd_offset_map(dst_pgd, address); + src_pmd = pmd_offset_map_nested(src_pgd, address); dst_pte = pte_offset_map(dst_pmd, address); src_pte = pte_offset_map_nested(src_pmd, address); cont_copy_pte_range_noset: address += PAGE_SIZE; if (address >= end) { + pmd_unmap(dst_pmd); + pmd_unmap_nested(src_pmd); pte_unmap_nested(src_pte); pte_unmap(dst_pte); goto out_unlock; @@ -376,6 +399,8 @@ cont_copy_pmd_range: src_pmd++; dst_pmd++; } while ((unsigned long)src_pmd & PMD_TABLE_MASK); + pmd_unmap_nested(src_pmd-1); + pmd_unmap(dst_pmd-1); } out_unlock: spin_unlock(&src->page_table_lock); @@ -451,7 +476,7 @@ zap_pmd_range(struct mmu_gather *tlb, pg pgd_clear(dir); return; } - pmd = pmd_offset(dir, address); + pmd = pmd_offset_map(dir, address); end = address + size; if (end > ((address + PGDIR_SIZE) & PGDIR_MASK)) end = ((address + PGDIR_SIZE) & PGDIR_MASK); @@ -460,6 +485,7 @@ zap_pmd_range(struct mmu_gather *tlb, pg address = (address + PMD_SIZE) & PMD_MASK; pmd++; } while (address < end); + pmd_unmap(pmd - 1); } void unmap_page_range(struct mmu_gather *tlb, struct vm_area_struct *vma, @@ -641,20 +667,27 @@ follow_page(struct mm_struct *mm, unsign if (pgd_none(*pgd) || pgd_bad(*pgd)) goto out; - pmd = pmd_offset(pgd, address); + pmd = pmd_offset_map(pgd, address); if (pmd_none(*pmd)) - goto out; - if (pmd_huge(*pmd)) - return follow_huge_pmd(mm, address, pmd, write); - if (pmd_bad(*pmd)) - goto out; + goto out_unmap; + if (pmd_bad(*pmd)) { + pmd_ERROR(*pmd); + pmd_clear(pmd); + goto out_unmap; + } + if (pmd_huge(*pmd)) { + struct page *page = follow_huge_pmd(mm, address, pmd, write); + pmd_unmap(pmd); + return page; + } ptep = pte_offset_map(pmd, address); if (!ptep) - goto out; + goto out_unmap; pte = *ptep; pte_unmap(ptep); + pmd_unmap(pmd); if (pte_present(pte)) { if (!write || (pte_write(pte) && pte_dirty(pte))) { pfn = pte_pfn(pte); @@ -665,6 +698,9 @@ follow_page(struct mm_struct *mm, unsign out: return NULL; +out_unmap: + pmd_unmap(pmd); + goto out; } /* @@ -723,7 +759,7 @@ int get_user_pages(struct task_struct *t pgd = pgd_offset_k(pg); if (!pgd) return i ? : -EFAULT; - pmd = pmd_offset(pgd, pg); + pmd = pmd_offset_kernel(pgd, pg); if (!pmd) return i ? : -EFAULT; pte = pte_offset_kernel(pmd, pg); @@ -815,8 +851,8 @@ static void zeromap_pte_range(pte_t * pt } while (address && (address < end)); } -static inline int zeromap_pmd_range(struct mm_struct *mm, pmd_t * pmd, unsigned long address, - unsigned long size, pgprot_t prot) +static inline int zeromap_pmd_range(struct mm_struct *mm, pmd_t **pmd, + unsigned long address, unsigned long size, pgprot_t prot) { unsigned long end; @@ -825,13 +861,13 @@ static inline int zeromap_pmd_range(stru if (end > PGDIR_SIZE) end = PGDIR_SIZE; do { - pte_t * pte = pte_alloc_map(mm, pmd, address); + pte_t *pte = pte_alloc_map(mm, pmd, address); if (!pte) return -ENOMEM; zeromap_pte_range(pte, address, end - address, prot); pte_unmap(pte); address = (address + PMD_SIZE) & PMD_MASK; - pmd++; + (*pmd)++; } while (address && (address < end)); return 0; } @@ -851,13 +887,14 @@ int zeromap_page_range(struct vm_area_st spin_lock(&mm->page_table_lock); do { - pmd_t *pmd = pmd_alloc(mm, dir, address); + pmd_t *pmd = pmd_alloc_map(mm, dir, address); error = -ENOMEM; if (!pmd) break; - error = zeromap_pmd_range(mm, pmd, address, end - address, prot); + error = zeromap_pmd_range(mm, &pmd, address, end - address, prot); if (error) break; + pmd_unmap(pmd - 1); address = (address + PGDIR_SIZE) & PGDIR_MASK; dir++; } while (address && (address < end)); @@ -892,8 +929,9 @@ static inline void remap_pte_range(pte_t } while (address && (address < end)); } -static inline int remap_pmd_range(struct mm_struct *mm, pmd_t * pmd, unsigned long address, unsigned long size, - unsigned long phys_addr, pgprot_t prot) +static inline int remap_pmd_range(struct mm_struct *mm, pmd_t **pmd, + unsigned long address, unsigned long size, + unsigned long phys_addr, pgprot_t prot) { unsigned long base, end; @@ -904,13 +942,13 @@ static inline int remap_pmd_range(struct end = PGDIR_SIZE; phys_addr -= address; do { - pte_t * pte = pte_alloc_map(mm, pmd, base + address); + pte_t *pte = pte_alloc_map(mm, pmd, base + address); if (!pte) return -ENOMEM; remap_pte_range(pte, base + address, end - address, address + phys_addr, prot); pte_unmap(pte); address = (address + PMD_SIZE) & PMD_MASK; - pmd++; + (*pmd)++; } while (address && (address < end)); return 0; } @@ -932,13 +970,14 @@ int remap_page_range(struct vm_area_stru spin_lock(&mm->page_table_lock); do { - pmd_t *pmd = pmd_alloc(mm, dir, from); + pmd_t *pmd = pmd_alloc_map(mm, dir, from); error = -ENOMEM; if (!pmd) break; - error = remap_pmd_range(mm, pmd, from, end - from, phys_addr + from, prot); + error = remap_pmd_range(mm, &pmd, from, end - from, phys_addr + from, prot); if (error) break; + pmd_unmap(pmd - 1); from = (from + PGDIR_SIZE) & PGDIR_MASK; dir++; } while (from && (from < end)); @@ -1008,6 +1047,7 @@ static int do_wp_page(struct mm_struct * * data, but for the moment just pretend this is OOM. */ pte_unmap(page_table); + pmd_unmap(pmd); printk(KERN_ERR "do_wp_page: bogus page at address %08lx\n", address); goto oom; @@ -1022,11 +1062,13 @@ static int do_wp_page(struct mm_struct * establish_pte(vma, address, page_table, pte_mkyoung(pte_mkdirty(pte_mkwrite(pte)))); pte_unmap(page_table); + pmd_unmap(pmd); ret = VM_FAULT_MINOR; goto out; } } pte_unmap(page_table); + pmd_unmap(pmd); /* * Ok, we need to copy. Oh, well.. @@ -1046,6 +1088,7 @@ static int do_wp_page(struct mm_struct * * Re-check the pte - we dropped the lock */ spin_lock(&mm->page_table_lock); + pmd = pmd_offset_map(pgd_offset(mm, address), address); page_table = pte_offset_map(pmd, address); if (pte_same(*page_table, pte)) { if (PageReserved(old_page)) @@ -1059,6 +1102,7 @@ static int do_wp_page(struct mm_struct * new_page = old_page; } pte_unmap(page_table); + pmd_unmap(pmd); page_cache_release(new_page); page_cache_release(old_page); ret = VM_FAULT_MINOR; @@ -1227,6 +1271,7 @@ static int do_swap_page(struct mm_struct struct pte_chain *pte_chain = NULL; pte_unmap(page_table); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); page = lookup_swap_cache(entry); if (!page) { @@ -1238,12 +1283,14 @@ static int do_swap_page(struct mm_struct * we released the page table lock. */ spin_lock(&mm->page_table_lock); + pmd = pmd_offset_map(pgd_offset(mm, address), address); page_table = pte_offset_map(pmd, address); if (pte_same(*page_table, orig_pte)) ret = VM_FAULT_OOM; else ret = VM_FAULT_MINOR; pte_unmap(page_table); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); goto out; } @@ -1266,9 +1313,11 @@ static int do_swap_page(struct mm_struct * released the page table lock. */ spin_lock(&mm->page_table_lock); + pmd = pmd_offset_map(pgd_offset(mm, address), address); page_table = pte_offset_map(pmd, address); if (!pte_same(*page_table, orig_pte)) { pte_unmap(page_table); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); unlock_page(page); page_cache_release(page); @@ -1294,6 +1343,7 @@ static int do_swap_page(struct mm_struct /* No need to invalidate - it was non-present before */ update_mmu_cache(vma, address, pte); + pmd_unmap(pmd); pte_unmap(page_table); spin_unlock(&mm->page_table_lock); out: @@ -1319,11 +1369,13 @@ do_anonymous_page(struct mm_struct *mm, pte_chain = pte_chain_alloc(GFP_ATOMIC); if (!pte_chain) { pte_unmap(page_table); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); pte_chain = pte_chain_alloc(GFP_KERNEL); if (!pte_chain) goto no_mem; spin_lock(&mm->page_table_lock); + pmd = pmd_offset_map(pgd_offset(mm, addr), addr); page_table = pte_offset_map(pmd, addr); } @@ -1334,6 +1386,7 @@ do_anonymous_page(struct mm_struct *mm, if (write_access) { /* Allocate our own private page. */ pte_unmap(page_table); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); page = alloc_page(GFP_HIGHUSER); @@ -1342,9 +1395,11 @@ do_anonymous_page(struct mm_struct *mm, clear_user_highpage(page, addr); spin_lock(&mm->page_table_lock); + pmd = pmd_offset_map(pgd_offset(mm, addr), addr); page_table = pte_offset_map(pmd, addr); if (!pte_none(*page_table)) { + pmd_unmap(pmd); pte_unmap(page_table); page_cache_release(page); spin_unlock(&mm->page_table_lock); @@ -1360,6 +1415,7 @@ do_anonymous_page(struct mm_struct *mm, set_pte(page_table, entry); /* ignores ZERO_PAGE */ pte_chain = page_add_rmap(page, page_table, pte_chain); + pmd_unmap(pmd); pte_unmap(page_table); /* No need to invalidate - it was non-present before */ @@ -1402,6 +1458,7 @@ do_no_page(struct mm_struct *mm, struct return do_anonymous_page(mm, vma, page_table, pmd, write_access, address); pte_unmap(page_table); + pmd_unmap(pmd); mapping = vma->vm_file->f_dentry->d_inode->i_mapping; sequence = atomic_read(&mapping->truncate_count); @@ -1446,6 +1503,7 @@ retry: page_cache_release(new_page); goto retry; } + pmd = pmd_offset_map(pgd_offset(mm, address), address); page_table = pte_offset_map(pmd, address); /* @@ -1468,9 +1526,11 @@ retry: set_pte(page_table, entry); pte_chain = page_add_rmap(new_page, page_table, pte_chain); pte_unmap(page_table); + pmd_unmap(pmd); } else { /* One of our sibling threads was faster, back out. */ pte_unmap(page_table); + pmd_unmap(pmd); page_cache_release(new_page); spin_unlock(&mm->page_table_lock); ret = VM_FAULT_MINOR; @@ -1514,6 +1574,7 @@ static int do_file_page(struct mm_struct pgoff = pte_to_pgoff(*pte); pte_unmap(pte); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); err = vma->vm_ops->populate(vma, address & PAGE_MASK, PAGE_SIZE, vma->vm_page_prot, pgoff, 0); @@ -1574,6 +1635,7 @@ static inline int handle_pte_fault(struc entry = pte_mkyoung(entry); establish_pte(vma, address, pte, entry); pte_unmap(pte); + pmd_unmap(pmd); spin_unlock(&mm->page_table_lock); return VM_FAULT_MINOR; } @@ -1600,10 +1662,10 @@ int handle_mm_fault(struct mm_struct *mm * and the SMP-safe atomic PTE updates. */ spin_lock(&mm->page_table_lock); - pmd = pmd_alloc(mm, pgd, address); + pmd = pmd_alloc_map(mm, pgd, address); if (pmd) { - pte_t * pte = pte_alloc_map(mm, pmd, address); + pte_t *pte = pte_alloc_map(mm, &pmd, address); if (pte) return handle_pte_fault(mm, vma, address, write_access, pte, pmd); } @@ -1640,7 +1702,30 @@ pmd_t *__pmd_alloc(struct mm_struct *mm, } pgd_populate(mm, pgd, new); out: - return pmd_offset(pgd, address); + return pmd_offset_map(pgd, address); +} + +pmd_t *__pmd_alloc_kernel(struct mm_struct *mm, pgd_t *pgd, unsigned long address) +{ + pmd_t *new; + + spin_unlock(&mm->page_table_lock); + new = pmd_alloc_one_kernel(mm, address); + spin_lock(&mm->page_table_lock); + if (!new) + return NULL; + + /* + * Because we dropped the lock, we should re-check the + * entry, as somebody else could have populated it.. + */ + if (pgd_present(*pgd)) { + pmd_free(new); + goto out; + } + pgd_populate(mm, pgd, new); +out: + return pmd_offset_kernel(pgd, address); } int make_pages_present(unsigned long addr, unsigned long end) @@ -1672,7 +1757,7 @@ struct page * vmalloc_to_page(void * vma pte_t *ptep, pte; if (!pgd_none(*pgd)) { - pmd = pmd_offset(pgd, addr); + pmd = pmd_offset_map(pgd, addr); if (!pmd_none(*pmd)) { preempt_disable(); ptep = pte_offset_map(pmd, addr); @@ -1682,6 +1767,7 @@ struct page * vmalloc_to_page(void * vma pte_unmap(ptep); preempt_enable(); } + pmd_unmap(pmd); } return page; } diff -prauN mm2-2.5.73-1/mm/mprotect.c mm2-2.5.73-2/mm/mprotect.c --- mm2-2.5.73-1/mm/mprotect.c 2003-06-28 03:09:58.000000000 -0700 +++ mm2-2.5.73-2/mm/mprotect.c 2003-06-28 03:11:37.000000000 -0700 @@ -73,7 +73,7 @@ change_pmd_range(pgd_t *pgd, unsigned lo pgd_clear(pgd); return; } - pmd = pmd_offset(pgd, address); + pmd = pmd_offset_map(pgd, address); address &= ~PGDIR_MASK; end = address + size; if (end > PGDIR_SIZE) @@ -83,6 +83,7 @@ change_pmd_range(pgd_t *pgd, unsigned lo address = (address + PMD_SIZE) & PMD_MASK; pmd++; } while (address && (address < end)); + pmd_unmap(pmd - 1); } static void diff -prauN mm2-2.5.73-1/mm/mremap.c mm2-2.5.73-2/mm/mremap.c --- mm2-2.5.73-1/mm/mremap.c 2003-06-28 03:09:58.000000000 -0700 +++ mm2-2.5.73-2/mm/mremap.c 2003-06-28 03:11:37.000000000 -0700 @@ -38,7 +38,7 @@ static pte_t *get_one_pte_map_nested(str goto end; } - pmd = pmd_offset(pgd, addr); + pmd = pmd_offset_map_nested(pgd, addr); if (pmd_none(*pmd)) goto end; if (pmd_bad(*pmd)) { @@ -53,6 +53,7 @@ static pte_t *get_one_pte_map_nested(str pte = NULL; } end: + pmd_unmap_nested(pmd); return pte; } @@ -61,12 +62,15 @@ static inline int page_table_present(str { pgd_t *pgd; pmd_t *pmd; + int ret; pgd = pgd_offset(mm, addr); if (pgd_none(*pgd)) return 0; - pmd = pmd_offset(pgd, addr); - return pmd_present(*pmd); + pmd = pmd_offset_map(pgd, addr); + ret = pmd_present(*pmd); + pmd_unmap(pmd); + return ret != 0; } #else #define page_table_present(mm, addr) (1) @@ -77,9 +81,10 @@ static inline pte_t *alloc_one_pte_map(s pmd_t *pmd; pte_t *pte = NULL; - pmd = pmd_alloc(mm, pgd_offset(mm, addr), addr); + pmd = pmd_alloc_map(mm, pgd_offset(mm, addr), addr); if (pmd) - pte = pte_alloc_map(mm, pmd, addr); + pte = pte_alloc_map(mm, &pmd, addr); + pmd_unmap(pmd); return pte; } diff -prauN mm2-2.5.73-1/mm/msync.c mm2-2.5.73-2/mm/msync.c --- mm2-2.5.73-1/mm/msync.c 2003-06-22 11:32:42.000000000 -0700 +++ mm2-2.5.73-2/mm/msync.c 2003-06-28 03:11:37.000000000 -0700 @@ -82,7 +82,7 @@ static inline int filemap_sync_pmd_range pgd_clear(pgd); return 0; } - pmd = pmd_offset(pgd, address); + pmd = pmd_offset_map(pgd, address); if ((address & PGDIR_MASK) != (end & PGDIR_MASK)) end = (address & PGDIR_MASK) + PGDIR_SIZE; error = 0; @@ -91,6 +91,7 @@ static inline int filemap_sync_pmd_range address = (address + PMD_SIZE) & PMD_MASK; pmd++; } while (address && (address < end)); + pmd_unmap(pmd - 1); return error; } diff -prauN mm2-2.5.73-1/mm/slab.c mm2-2.5.73-2/mm/slab.c --- mm2-2.5.73-1/mm/slab.c 2003-06-28 03:09:58.000000000 -0700 +++ mm2-2.5.73-2/mm/slab.c 2003-06-28 03:27:00.000000000 -0700 @@ -2717,7 +2717,7 @@ void ptrinfo(unsigned long addr) printk("No pgd.\n"); break; } - pmd = pmd_offset(pgd, addr); + pmd = pmd_offset_kernel(pgd, addr); if (pmd_none(*pmd)) { printk("No pmd.\n"); break; diff -prauN mm2-2.5.73-1/mm/swapfile.c mm2-2.5.73-2/mm/swapfile.c --- mm2-2.5.73-1/mm/swapfile.c 2003-06-28 03:09:58.000000000 -0700 +++ mm2-2.5.73-2/mm/swapfile.c 2003-06-28 03:11:37.000000000 -0700 @@ -448,7 +448,7 @@ static int unuse_pgd(struct vm_area_stru pgd_clear(dir); return 0; } - pmd = pmd_offset(dir, address); + pmd = pmd_offset_map(dir, address); offset = address & PGDIR_MASK; address &= ~PGDIR_MASK; end = address + size; @@ -463,6 +463,7 @@ static int unuse_pgd(struct vm_area_stru address = (address + PMD_SIZE) & PMD_MASK; pmd++; } while (address && (address < end)); + pmd_unmap(pmd - 1); return 0; } diff -prauN mm2-2.5.73-1/mm/vmalloc.c mm2-2.5.73-2/mm/vmalloc.c --- mm2-2.5.73-1/mm/vmalloc.c 2003-06-22 11:32:56.000000000 -0700 +++ mm2-2.5.73-2/mm/vmalloc.c 2003-06-28 03:11:37.000000000 -0700 @@ -70,7 +70,7 @@ static void unmap_area_pmd(pgd_t *dir, u return; } - pmd = pmd_offset(dir, address); + pmd = pmd_offset_kernel(dir, address); address &= ~PGDIR_MASK; end = address + size; if (end > PGDIR_SIZE) @@ -159,7 +159,7 @@ int map_vm_area(struct vm_struct *area, dir = pgd_offset_k(address); spin_lock(&init_mm.page_table_lock); do { - pmd_t *pmd = pmd_alloc(&init_mm, dir, address); + pmd_t *pmd = pmd_alloc_kernel(&init_mm, dir, address); if (!pmd) { err = -ENOMEM; break; -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-28 15:54 ` 2.5.73-mm2 William Lee Irwin III @ 2003-06-28 16:08 ` Christoph Hellwig -1 siblings, 0 replies; 49+ messages in thread From: Christoph Hellwig @ 2003-06-28 16:08 UTC (permalink / raw) To: William Lee Irwin III, Andrew Morton, linux-kernel, linux-mm On Sat, Jun 28, 2003 at 08:54:36AM -0700, William Lee Irwin III wrote: > +config HIGHPMD > + bool "Allocate 2nd-level pagetables from highmem" > + depends on HIGHMEM64G > + help > + The VM uses one pmd entry for each pagetable page of physical > + memory allocated. For systems with extreme amounts of highmem, > + this cannot be tolerated. Setting this option will put > + userspace 2nd-level pagetables in highmem. Does this make sense for !HIGHPTE? In fact does it make sense to carry along HIGHPTE as an option still? .. > +#ifndef CONFIG_HIGHPMD /* Oh boy. Error reporting is going to blow major goats. */ Any chance you can rearragne the code to avoid the ifndef in favour of an ifdef? > set_pte(dst_pte, entry); > + pmd_unmap(dst_pte); > + pmd_unmap_nested(src_pte); <Lots more pmd_unmap* calls snipped> Looks like you changed some API so that pmds are now returned mapped? It might make sense to change their names into foo_map then so the breakage is at the API level if someone misses updates for the changes. > +#ifdef CONFIG_HIGHPMD > +#define GFP_PMD (__GFP_REPEAT|__GFP_HIGHMEM|GFP_KERNEL) > +#else > +#define GFP_PMD (__GFP_REPEAT|GFP_KERNEL) > +#endif So what? Do you want to use a space or tab after the #define? :) Also Given that GFP_PMD is used just once it's argueable whether it makes sense to get rid of the defintion and use the expanded values directly. Otherwise the patch looks fine to me and should allow to get some more free lowmem on those insanely big 32bit machines.. :) ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-06-28 16:08 ` Christoph Hellwig 0 siblings, 0 replies; 49+ messages in thread From: Christoph Hellwig @ 2003-06-28 16:08 UTC (permalink / raw) To: William Lee Irwin III, Andrew Morton, linux-kernel, linux-mm On Sat, Jun 28, 2003 at 08:54:36AM -0700, William Lee Irwin III wrote: > +config HIGHPMD > + bool "Allocate 2nd-level pagetables from highmem" > + depends on HIGHMEM64G > + help > + The VM uses one pmd entry for each pagetable page of physical > + memory allocated. For systems with extreme amounts of highmem, > + this cannot be tolerated. Setting this option will put > + userspace 2nd-level pagetables in highmem. Does this make sense for !HIGHPTE? In fact does it make sense to carry along HIGHPTE as an option still? .. > +#ifndef CONFIG_HIGHPMD /* Oh boy. Error reporting is going to blow major goats. */ Any chance you can rearragne the code to avoid the ifndef in favour of an ifdef? > set_pte(dst_pte, entry); > + pmd_unmap(dst_pte); > + pmd_unmap_nested(src_pte); <Lots more pmd_unmap* calls snipped> Looks like you changed some API so that pmds are now returned mapped? It might make sense to change their names into foo_map then so the breakage is at the API level if someone misses updates for the changes. > +#ifdef CONFIG_HIGHPMD > +#define GFP_PMD (__GFP_REPEAT|__GFP_HIGHMEM|GFP_KERNEL) > +#else > +#define GFP_PMD (__GFP_REPEAT|GFP_KERNEL) > +#endif So what? Do you want to use a space or tab after the #define? :) Also Given that GFP_PMD is used just once it's argueable whether it makes sense to get rid of the defintion and use the expanded values directly. Otherwise the patch looks fine to me and should allow to get some more free lowmem on those insanely big 32bit machines.. :) -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-28 16:08 ` 2.5.73-mm2 Christoph Hellwig (?) @ 2003-06-28 20:49 ` William Lee Irwin III -1 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-06-28 20:49 UTC (permalink / raw) To: Christoph Hellwig, Andrew Morton, linux-kernel, linux-mm [-- Attachment #1: Type: text/plain, Size: 3005 bytes --] On Sat, Jun 28, 2003 at 08:54:36AM -0700, William Lee Irwin III wrote: >> +config HIGHPMD >> + bool "Allocate 2nd-level pagetables from highmem" >> + depends on HIGHMEM64G >> + help >> + The VM uses one pmd entry for each pagetable page of physical >> + memory allocated. For systems with extreme amounts of highmem, >> + this cannot be tolerated. Setting this option will put >> + userspace 2nd-level pagetables in highmem. On Sat, Jun 28, 2003 at 05:08:37PM +0100, Christoph Hellwig wrote: > Does this make sense for !HIGHPTE? In fact does it make sense to > carry along HIGHPTE as an option still? .. It's both possible and functional, if that's the question. I basically decided I'd enforce the minimum policy, though it is possible to entangle the two. I'll make it depends on HIGHMEM64G && HIGHPTE for now. On Sat, Jun 28, 2003 at 08:54:36AM -0700, William Lee Irwin III wrote: >> +#ifndef CONFIG_HIGHPMD /* Oh boy. Error reporting is going to blow major goats. */ On Sat, Jun 28, 2003 at 05:08:37PM +0100, Christoph Hellwig wrote: > Any chance you can rearragne the code to avoid the ifndef in favour > of an ifdef? Okay. On Sat, Jun 28, 2003 at 08:54:36AM -0700, William Lee Irwin III wrote: >> set_pte(dst_pte, entry); >> + pmd_unmap(dst_pte); >> + pmd_unmap_nested(src_pte); On Sat, Jun 28, 2003 at 08:54:36AM -0700, William Lee Irwin III wrote: > <Lots more pmd_unmap* calls snipped> > Looks like you changed some API so that pmds are now returned mapped? > It might make sense to change their names into foo_map then so the > breakage is at the API level if someone misses updates for the changes. It's entirely analogous to pte_offset_map()/pte_alloc_map(). I stubbed out the things for all architectures already, so it should only be a question of additional functionality for 32-bit architectures supporting PAE-like mechanisms. It looks like some mips64, m68k, and other bits fell out of my tree. They're restored in the real content of this followup with matching definition counts for all arches. On Sat, Jun 28, 2003 at 05:08:37PM +0100, Christoph Hellwig wrote: >> +#ifdef CONFIG_HIGHPMD >> +#define GFP_PMD (__GFP_REPEAT|__GFP_HIGHMEM|GFP_KERNEL) >> +#else >> +#define GFP_PMD (__GFP_REPEAT|GFP_KERNEL) >> +#endif On Sat, Jun 28, 2003 at 08:54:36AM -0700, William Lee Irwin III wrote: > So what? Do you want to use a space or tab after the #define? :) > Also Given that GFP_PMD is used just once it's argueable whether it makes > sense to get rid of the defintion and use the expanded values directly. > Otherwise the patch looks fine to me and should allow to get some more > free lowmem on those insanely big 32bit machines.. :) I guess if I'm going to bother at all, shoving both GFP_PMD and GFP_PTE into gfp.h and updating all arches to cut out the duplicated #ifdefs would be the way to do it. I have a feeling that won't fly, so it's ripped out of the following: (compressed and MIME-attached, unfortunately some MTA's are barfing on it) -- wli [-- Attachment #2: hpmd-2.5.73-mm2-2.gz --] [-- Type: application/octet-stream, Size: 14021 bytes --] ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-28 16:08 ` 2.5.73-mm2 Christoph Hellwig @ 2003-06-29 0:34 ` Martin J. Bligh -1 siblings, 0 replies; 49+ messages in thread From: Martin J. Bligh @ 2003-06-29 0:34 UTC (permalink / raw) To: Christoph Hellwig, William Lee Irwin III, Andrew Morton, linux-kernel, linux-mm --Christoph Hellwig <hch@infradead.org> wrote (on Saturday, June 28, 2003 17:08:37 +0100): > On Sat, Jun 28, 2003 at 08:54:36AM -0700, William Lee Irwin III wrote: >> +config HIGHPMD >> + bool "Allocate 2nd-level pagetables from highmem" >> + depends on HIGHMEM64G >> + help >> + The VM uses one pmd entry for each pagetable page of physical >> + memory allocated. For systems with extreme amounts of highmem, >> + this cannot be tolerated. Setting this option will put >> + userspace 2nd-level pagetables in highmem. > > Does this make sense for !HIGHPTE? In fact does it make sense to > carry along HIGHPTE as an option still? .. Last time I measured it, it had about a 10% overhead in kernel time. Seems like a good thing to keep as an option to me. Bill said he had some other code to alleviate the overhead, but I don't think it's merged ... I'd rather see UKVA (permanently map the pagetables on a per-process basis) merged before it becomes "not an option" - that gets rid of all the kmapping. M. ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-06-29 0:34 ` Martin J. Bligh 0 siblings, 0 replies; 49+ messages in thread From: Martin J. Bligh @ 2003-06-29 0:34 UTC (permalink / raw) To: Christoph Hellwig, William Lee Irwin III, Andrew Morton, linux-kernel, linux-mm --Christoph Hellwig <hch@infradead.org> wrote (on Saturday, June 28, 2003 17:08:37 +0100): > On Sat, Jun 28, 2003 at 08:54:36AM -0700, William Lee Irwin III wrote: >> +config HIGHPMD >> + bool "Allocate 2nd-level pagetables from highmem" >> + depends on HIGHMEM64G >> + help >> + The VM uses one pmd entry for each pagetable page of physical >> + memory allocated. For systems with extreme amounts of highmem, >> + this cannot be tolerated. Setting this option will put >> + userspace 2nd-level pagetables in highmem. > > Does this make sense for !HIGHPTE? In fact does it make sense to > carry along HIGHPTE as an option still? .. Last time I measured it, it had about a 10% overhead in kernel time. Seems like a good thing to keep as an option to me. Bill said he had some other code to alleviate the overhead, but I don't think it's merged ... I'd rather see UKVA (permanently map the pagetables on a per-process basis) merged before it becomes "not an option" - that gets rid of all the kmapping. M. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-29 0:34 ` 2.5.73-mm2 Martin J. Bligh @ 2003-06-29 2:18 ` William Lee Irwin III -1 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-06-29 2:18 UTC (permalink / raw) To: Martin J. Bligh; +Cc: Christoph Hellwig, Andrew Morton, linux-kernel, linux-mm On Sat, Jun 28, 2003 at 05:34:05PM -0700, Martin J. Bligh wrote: > Last time I measured it, it had about a 10% overhead in kernel time. > Seems like a good thing to keep as an option to me. Bill said he > had some other code to alleviate the overhead, but I don't think > it's merged ... I'd rather see UKVA (permanently map the pagetables > on a per-process basis) merged before it becomes "not an option" - > that gets rid of all the kmapping. There are several orthogonal things going on here. One is dropping the hooks in the right places to get various concrete tasks done. Another is general resource scalability vs. raw overhead tradeoffs. The last one is gathering a wide enough repertoire of core hooks that arches can use "advanced" techniques like recursive pagetables when they require various kinds of intervention by the kernel to use. This is just another set of hooks we'll need for our end goal, with a fully functional implementation. It has direct applications and is completely usable now for resource scalability albeit with some overhead. Things are all headed in the appropriate directions; the hooks do not conflict with and do not require any core modifications whatsoever in order to use in combination with recursive pagetables; they can simply recover information from already-available places and transparently replace the highpmd and highpte arch code. I can work directly with Dave to arrange a proper demonstration of this (i.e. fully functional implementation) if need be. I've largely avoided interceding in recursive pagetable mechanics in order not to duplicate work. -- wli ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-06-29 2:18 ` William Lee Irwin III 0 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-06-29 2:18 UTC (permalink / raw) To: Martin J. Bligh; +Cc: Christoph Hellwig, Andrew Morton, linux-kernel, linux-mm On Sat, Jun 28, 2003 at 05:34:05PM -0700, Martin J. Bligh wrote: > Last time I measured it, it had about a 10% overhead in kernel time. > Seems like a good thing to keep as an option to me. Bill said he > had some other code to alleviate the overhead, but I don't think > it's merged ... I'd rather see UKVA (permanently map the pagetables > on a per-process basis) merged before it becomes "not an option" - > that gets rid of all the kmapping. There are several orthogonal things going on here. One is dropping the hooks in the right places to get various concrete tasks done. Another is general resource scalability vs. raw overhead tradeoffs. The last one is gathering a wide enough repertoire of core hooks that arches can use "advanced" techniques like recursive pagetables when they require various kinds of intervention by the kernel to use. This is just another set of hooks we'll need for our end goal, with a fully functional implementation. It has direct applications and is completely usable now for resource scalability albeit with some overhead. Things are all headed in the appropriate directions; the hooks do not conflict with and do not require any core modifications whatsoever in order to use in combination with recursive pagetables; they can simply recover information from already-available places and transparently replace the highpmd and highpte arch code. I can work directly with Dave to arrange a proper demonstration of this (i.e. fully functional implementation) if need be. I've largely avoided interceding in recursive pagetable mechanics in order not to duplicate work. -- wli -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-29 2:18 ` 2.5.73-mm2 William Lee Irwin III @ 2003-06-29 3:07 ` Martin J. Bligh -1 siblings, 0 replies; 49+ messages in thread From: Martin J. Bligh @ 2003-06-29 3:07 UTC (permalink / raw) To: William Lee Irwin III Cc: Christoph Hellwig, Andrew Morton, linux-kernel, linux-mm --William Lee Irwin III <wli@holomorphy.com> wrote (on Saturday, June 28, 2003 19:18:09 -0700): > On Sat, Jun 28, 2003 at 05:34:05PM -0700, Martin J. Bligh wrote: >> Last time I measured it, it had about a 10% overhead in kernel time. >> Seems like a good thing to keep as an option to me. Bill said he >> had some other code to alleviate the overhead, but I don't think >> it's merged ... I'd rather see UKVA (permanently map the pagetables >> on a per-process basis) merged before it becomes "not an option" - >> that gets rid of all the kmapping. > > There are several orthogonal things going on here. One is dropping the > hooks in the right places to get various concrete tasks done. Another > is general resource scalability vs. raw overhead tradeoffs. The last > one is gathering a wide enough repertoire of core hooks that arches can > use "advanced" techniques like recursive pagetables when they require > various kinds of intervention by the kernel to use. > > This is just another set of hooks we'll need for our end goal, with a > fully functional implementation. It has direct applications and is > completely usable now for resource scalability albeit with some > overhead. Things are all headed in the appropriate directions; the > hooks do not conflict with and do not require any core modifications > whatsoever in order to use in combination with recursive pagetables; > they can simply recover information from already-available places and > transparently replace the highpmd and highpte arch code. > > I can work directly with Dave to arrange a proper demonstration of this > (i.e. fully functional implementation) if need be. I've largely avoided > interceding in recursive pagetable mechanics in order not to duplicate > work. Right, I'm not against what you're doing - I'm totally for it. My only concern was that whilst it has some overhead, it should stay as a config option (which you did). That lets people make the call of overhead vs resource scaling. Your patch is fine - just the talk of removing the config option scared me ;-) M. ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-06-29 3:07 ` Martin J. Bligh 0 siblings, 0 replies; 49+ messages in thread From: Martin J. Bligh @ 2003-06-29 3:07 UTC (permalink / raw) To: William Lee Irwin III Cc: Christoph Hellwig, Andrew Morton, linux-kernel, linux-mm --William Lee Irwin III <wli@holomorphy.com> wrote (on Saturday, June 28, 2003 19:18:09 -0700): > On Sat, Jun 28, 2003 at 05:34:05PM -0700, Martin J. Bligh wrote: >> Last time I measured it, it had about a 10% overhead in kernel time. >> Seems like a good thing to keep as an option to me. Bill said he >> had some other code to alleviate the overhead, but I don't think >> it's merged ... I'd rather see UKVA (permanently map the pagetables >> on a per-process basis) merged before it becomes "not an option" - >> that gets rid of all the kmapping. > > There are several orthogonal things going on here. One is dropping the > hooks in the right places to get various concrete tasks done. Another > is general resource scalability vs. raw overhead tradeoffs. The last > one is gathering a wide enough repertoire of core hooks that arches can > use "advanced" techniques like recursive pagetables when they require > various kinds of intervention by the kernel to use. > > This is just another set of hooks we'll need for our end goal, with a > fully functional implementation. It has direct applications and is > completely usable now for resource scalability albeit with some > overhead. Things are all headed in the appropriate directions; the > hooks do not conflict with and do not require any core modifications > whatsoever in order to use in combination with recursive pagetables; > they can simply recover information from already-available places and > transparently replace the highpmd and highpte arch code. > > I can work directly with Dave to arrange a proper demonstration of this > (i.e. fully functional implementation) if need be. I've largely avoided > interceding in recursive pagetable mechanics in order not to duplicate > work. Right, I'm not against what you're doing - I'm totally for it. My only concern was that whilst it has some overhead, it should stay as a config option (which you did). That lets people make the call of overhead vs resource scaling. Your patch is fine - just the talk of removing the config option scared me ;-) M. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-28 15:54 ` 2.5.73-mm2 William Lee Irwin III @ 2003-06-28 23:00 ` Andrew Morton -1 siblings, 0 replies; 49+ messages in thread From: Andrew Morton @ 2003-06-28 23:00 UTC (permalink / raw) To: William Lee Irwin III; +Cc: linux-kernel, linux-mm William Lee Irwin III <wli@holomorphy.com> wrote: > > Here's highpmd. I taught patch-scripts a new trick: check_patch() { if grep "^+.*[ ]$" $P/patches/$1.patch then echo warning: $1 adds trailing whitespace fi } + if (pmd_table != pmd_offset_kernel(pgd, 0)) + pmd = pmd_offset_kernel(pgd, address); +#define __pgd_page(pgd) (__bpn_to_ba(pgd_val(pgd))) warning: highpmd adds trailing whitespace You're far from the worst. There's some editor out there which adds trailing tabs all over the place. I edited the diff. What architectures has this been tested on? ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-06-28 23:00 ` Andrew Morton 0 siblings, 0 replies; 49+ messages in thread From: Andrew Morton @ 2003-06-28 23:00 UTC (permalink / raw) To: William Lee Irwin III; +Cc: linux-kernel, linux-mm William Lee Irwin III <wli@holomorphy.com> wrote: > > Here's highpmd. I taught patch-scripts a new trick: check_patch() { if grep "^+.*[ ]$" $P/patches/$1.patch then echo warning: $1 adds trailing whitespace fi } + if (pmd_table != pmd_offset_kernel(pgd, 0)) + pmd = pmd_offset_kernel(pgd, address); +#define __pgd_page(pgd) (__bpn_to_ba(pgd_val(pgd))) warning: highpmd adds trailing whitespace You're far from the worst. There's some editor out there which adds trailing tabs all over the place. I edited the diff. What architectures has this been tested on? -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-28 23:00 ` 2.5.73-mm2 Andrew Morton @ 2003-06-28 23:11 ` William Lee Irwin III -1 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-06-28 23:11 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-mm William Lee Irwin III <wli@holomorphy.com> wrote: >> Here's highpmd. On Sat, Jun 28, 2003 at 04:00:13PM -0700, Andrew Morton wrote: > I taught patch-scripts a new trick: > check_patch() > { > if grep "^+.*[ ]$" $P/patches/$1.patch > then > echo warning: $1 adds trailing whitespace > fi > } William Lee Irwin III <wli@holomorphy.com> wrote: > + if (pmd_table != pmd_offset_kernel(pgd, 0)) > + pmd = pmd_offset_kernel(pgd, address); > +#define __pgd_page(pgd) (__bpn_to_ba(pgd_val(pgd))) On Sat, Jun 28, 2003 at 04:00:13PM -0700, Andrew Morton wrote: > warning: highpmd adds trailing whitespace > You're far from the worst. There's some editor out there which > adds trailing tabs all over the place. I edited the diff. This is not my editor; it is either a manual screwup or cut and paste (inside vi(1) and/or ed(1) buffers) of code which did so beforehand. Thanks for cleaning that up for me; it's one of my own largest pet peeves and I'm not terribly pleased to hear I've committed it (though I'll not go so far as to shoot the messenger). On Sat, Jun 28, 2003 at 04:00:13PM -0700, Andrew Morton wrote: > What architectures has this been tested on? i386 only, CONFIG_HIGHMEM64G with various combinations of highpte & highpmd, and nohighmem. No CONFIG_HIGHMEM4G or non-i386 machines that can run 2.5.x are within my grasp (obviously CONFIG_HIGHMEM4G machines could, I just don't have them, and the discontig code barfs on mem=). -- wli ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-06-28 23:11 ` William Lee Irwin III 0 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-06-28 23:11 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-mm William Lee Irwin III <wli@holomorphy.com> wrote: >> Here's highpmd. On Sat, Jun 28, 2003 at 04:00:13PM -0700, Andrew Morton wrote: > I taught patch-scripts a new trick: > check_patch() > { > if grep "^+.*[ ]$" $P/patches/$1.patch > then > echo warning: $1 adds trailing whitespace > fi > } William Lee Irwin III <wli@holomorphy.com> wrote: > + if (pmd_table != pmd_offset_kernel(pgd, 0)) > + pmd = pmd_offset_kernel(pgd, address); > +#define __pgd_page(pgd) (__bpn_to_ba(pgd_val(pgd))) On Sat, Jun 28, 2003 at 04:00:13PM -0700, Andrew Morton wrote: > warning: highpmd adds trailing whitespace > You're far from the worst. There's some editor out there which > adds trailing tabs all over the place. I edited the diff. This is not my editor; it is either a manual screwup or cut and paste (inside vi(1) and/or ed(1) buffers) of code which did so beforehand. Thanks for cleaning that up for me; it's one of my own largest pet peeves and I'm not terribly pleased to hear I've committed it (though I'll not go so far as to shoot the messenger). On Sat, Jun 28, 2003 at 04:00:13PM -0700, Andrew Morton wrote: > What architectures has this been tested on? i386 only, CONFIG_HIGHMEM64G with various combinations of highpte & highpmd, and nohighmem. No CONFIG_HIGHMEM4G or non-i386 machines that can run 2.5.x are within my grasp (obviously CONFIG_HIGHMEM4G machines could, I just don't have them, and the discontig code barfs on mem=). -- wli -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-28 23:11 ` 2.5.73-mm2 William Lee Irwin III @ 2003-06-29 12:45 ` Zwane Mwaikambo -1 siblings, 0 replies; 49+ messages in thread From: Zwane Mwaikambo @ 2003-06-29 12:45 UTC (permalink / raw) To: William Lee Irwin III; +Cc: Andrew Morton, linux-kernel, linux-mm On Sat, 28 Jun 2003, William Lee Irwin III wrote: > On Sat, Jun 28, 2003 at 04:00:13PM -0700, Andrew Morton wrote: > > What architectures has this been tested on? > > i386 only, CONFIG_HIGHMEM64G with various combinations of highpte & > highpmd, and nohighmem. No CONFIG_HIGHMEM4G or non-i386 machines that > can run 2.5.x are within my grasp (obviously CONFIG_HIGHMEM4G machines > could, I just don't have them, and the discontig code barfs on mem=). It comes up fine on a CONFIG_HIGHMEM4G (16G) box. -- function.linuxpower.ca ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-06-29 12:45 ` Zwane Mwaikambo 0 siblings, 0 replies; 49+ messages in thread From: Zwane Mwaikambo @ 2003-06-29 12:45 UTC (permalink / raw) To: William Lee Irwin III; +Cc: Andrew Morton, linux-kernel, linux-mm On Sat, 28 Jun 2003, William Lee Irwin III wrote: > On Sat, Jun 28, 2003 at 04:00:13PM -0700, Andrew Morton wrote: > > What architectures has this been tested on? > > i386 only, CONFIG_HIGHMEM64G with various combinations of highpte & > highpmd, and nohighmem. No CONFIG_HIGHMEM4G or non-i386 machines that > can run 2.5.x are within my grasp (obviously CONFIG_HIGHMEM4G machines > could, I just don't have them, and the discontig code barfs on mem=). It comes up fine on a CONFIG_HIGHMEM4G (16G) box. -- function.linuxpower.ca -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-28 15:54 ` 2.5.73-mm2 William Lee Irwin III ` (2 preceding siblings ...) (?) @ 2003-07-02 3:11 ` William Lee Irwin III -1 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-07-02 3:11 UTC (permalink / raw) To: Andrew Morton, linux-kernel, linux-mm [-- Attachment #1: Type: text/plain, Size: 848 bytes --] On Fri, Jun 27, 2003 at 08:21:30PM -0700, Andrew Morton wrote: >> ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/ >> Just bits and pieces. On Sat, Jun 28, 2003 at 08:54:36AM -0700, William Lee Irwin III wrote: > Here's highpmd. This allocates L2 pagetables from highmem, decreasing > the per-process lowmem overhead on CONFIG_HIGHMEM64G from 20KB to 8KB. > Some attempts were made to update non-i386 architectures to the new > API's, though they're entirely untested. It's been tested for a while > in -wli on i386 machines, both lowmem and highmem boxen. Here's highpmd again, but with the bash-shared-mappings oops fixed. Some missing s/pmd_alloc()/pmd_alloc_map()/ conversions in non-i386 code are also included in the update. Included as a MIME attachment to prevent MTA's from barfing on its size. -- wli [-- Attachment #2: highpmd-2.5.73-mm2-2.bz2 --] [-- Type: application/octet-stream, Size: 14399 bytes --] ^ permalink raw reply [flat|nested] 49+ messages in thread
* [patch] 2.5.73-mm2: let CONFIG_TC35815 depend on CONFIG_TOSHIBA_JMR3927 2003-06-28 3:21 ` 2.5.73-mm2 Andrew Morton ` (2 preceding siblings ...) (?) @ 2003-06-29 19:04 ` Adrian Bunk 2003-06-29 21:04 ` Ralf Baechle -1 siblings, 1 reply; 49+ messages in thread From: Adrian Bunk @ 2003-06-29 19:04 UTC (permalink / raw) To: Andrew Morton, ralf; +Cc: linux-kernel, trivial The following problem seems to come from Linus' tree: I got an error at the final linking with CONFIG_TC35815 enabled since the variables tc_readl and tc_writel are not available. The only place where they are defined is arch/mips/pci/ops-jmr3927.c, so I assume the following was intended: --- linux-2.5.73-mm2/drivers/net/Kconfig.old 2003-06-28 11:14:16.000000000 +0200 +++ linux-2.5.73-mm2/drivers/net/Kconfig 2003-06-29 20:55:16.000000000 +0200 @@ -1397,7 +1397,7 @@ config TC35815 tristate "TOSHIBA TC35815 Ethernet support" - depends on NET_PCI && PCI + depends on NET_PCI && PCI && TOSHIBA_JMR3927 config DGRS tristate "Digi Intl. RightSwitch SE-X support" cu Adrian -- "Is there not promise of rain?" Ling Tan asked suddenly out of the darkness. There had been need of rain for many days. "Only a promise," Lao Er said. Pearl S. Buck - Dragon Seed ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: [patch] 2.5.73-mm2: let CONFIG_TC35815 depend on CONFIG_TOSHIBA_JMR3927 2003-06-29 19:04 ` [patch] 2.5.73-mm2: let CONFIG_TC35815 depend on CONFIG_TOSHIBA_JMR3927 Adrian Bunk @ 2003-06-29 21:04 ` Ralf Baechle 0 siblings, 0 replies; 49+ messages in thread From: Ralf Baechle @ 2003-06-29 21:04 UTC (permalink / raw) To: Adrian Bunk; +Cc: Andrew Morton, linux-kernel, trivial On Sun, Jun 29, 2003 at 09:04:32PM +0200, Adrian Bunk wrote: > The following problem seems to come from Linus' tree: > > I got an error at the final linking with CONFIG_TC35815 enabled since > the variables tc_readl and tc_writel are not available. > > The only place where they are defined is arch/mips/pci/ops-jmr3927.c, so > I assume the following was intended: Not really intended but it makes sense as this particular boards seems to be the only user of that chip - which already has vanished from Toshiba's pages anyway ... Ralf ^ permalink raw reply [flat|nested] 49+ messages in thread
* [patch] 2.5.73-mm2: let CONFIG_TC35815 depend on CONFIG_TOSHIBA_JMR3927 2003-06-28 3:21 ` 2.5.73-mm2 Andrew Morton ` (3 preceding siblings ...) (?) @ 2003-06-29 19:04 ` Adrian Bunk -1 siblings, 0 replies; 49+ messages in thread From: Adrian Bunk @ 2003-06-29 19:04 UTC (permalink / raw) To: Andrew Morton, ralf; +Cc: linux-kernel, trivial The following problem seems to come from Linus' tree: I got an error at the final linking with CONFIG_TC35815 enabled since the variables tc_readl and tc_writel are not available. The only place where they are defined is arch/mips/pci/ops-jmr3927.c, so I assume the following was intended: --- linux-2.5.73-mm2/drivers/net/Kconfig.old 2003-06-28 11:14:16.000000000 +0200 +++ linux-2.5.73-mm2/drivers/net/Kconfig 2003-06-29 20:55:16.000000000 +0200 @@ -1397,7 +1397,7 @@ config TC35815 tristate "TOSHIBA TC35815 Ethernet support" - depends on NET_PCI && PCI + depends on NET_PCI && PCI && TOSHIBA_JMR3927 config DGRS tristate "Digi Intl. RightSwitch SE-X support" cu Adrian -- "Is there not promise of rain?" Ling Tan asked suddenly out of the darkness. There had been need of rain for many days. "Only a promise," Lao Er said. Pearl S. Buck - Dragon Seed ^ permalink raw reply [flat|nested] 49+ messages in thread
* [linux-lvm] Re: 2.5.73-mm2 2003-06-28 3:21 ` 2.5.73-mm2 Andrew Morton ` (4 preceding siblings ...) (?) @ 2003-06-30 9:51 ` Wiktor Wodecki 2003-06-29 0:30 ` Andrew Morton -1 siblings, 1 reply; 49+ messages in thread From: Wiktor Wodecki @ 2003-06-30 9:51 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-lvm [-- Attachment #1: Type: text/plain, Size: 1222 bytes --] Hello, On Fri, Jun 27, 2003 at 08:21:30PM -0700, Andrew Morton wrote: > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/ > > Just bits and pieces. I have problems with lvm2 and the 2.5.73-mm2 patchset. I have two volume groups, one with 4 lv's (lvm2 metadata) the other one with only one lv (lvm2, too). When I boot the 2.5.73 stock kernel everything seems to be fine, however when I boot into -mm2 only the first lv of the first vg is found. A subsequent lvscan doesn't find anything. I've seen this bug a couple of kernel releases ago in your tree (I think it was 2.5.69-mmsomething) but I didn't had time to look into it then, I had other 2.5 problems back then. There are more things in my lvm configuration (some pv's beeing md devices) but I don't think that is important, since the raid stuff comes up properly. Booting into stock 2.5.73 brings up the lv's as expected. I'm running debian testing on x86 with the lvm2 package installed. LVM version: 1.95.15 (2003-01-10) Library version: 0.96.07-ioctl-cvs (2002-11-21) Driver version: 1.0.6 If you need more infos or patches to be tested I'd be happy to help. -- Regards, Wiktor Wodecki [-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --] ^ permalink raw reply [flat|nested] 49+ messages in thread
* [linux-lvm] Re: 2.5.73-mm2 2003-06-30 9:51 ` [linux-lvm] Re: 2.5.73-mm2 Wiktor Wodecki @ 2003-06-29 0:30 ` Andrew Morton 2003-06-30 9:51 ` Wiktor Wodecki 0 siblings, 1 reply; 49+ messages in thread From: Andrew Morton @ 2003-06-29 0:30 UTC (permalink / raw) To: Wiktor Wodecki; +Cc: linux-lvm Wiktor Wodecki <wodecki@gmx.de> wrote: > > On Fri, Jun 27, 2003 at 08:21:30PM -0700, Andrew Morton wrote: > > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/ > > > > Just bits and pieces. > > I have problems with lvm2 and the 2.5.73-mm2 patchset. Could be because -mm uses a 64-bit dev_t. Could you please revert ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/broken-out/64-bit-dev_t-kdev_t.patch and retest? I think there's a version of devicemapper userspace which works correctly with 64-bit dev_t. Perhaps the Sistina team can shed some light. ^ permalink raw reply [flat|nested] 49+ messages in thread
* [linux-lvm] Re: 2.5.73-mm2 2003-06-29 0:30 ` Andrew Morton @ 2003-06-30 9:51 ` Wiktor Wodecki 2003-06-30 10:23 ` Joe Thornber 2003-06-30 12:29 ` Christophe Saout 0 siblings, 2 replies; 49+ messages in thread From: Wiktor Wodecki @ 2003-06-30 9:51 UTC (permalink / raw) To: linux-lvm; +Cc: Andrew Morton [-- Attachment #1: Type: text/plain, Size: 671 bytes --] On Sat, Jun 28, 2003 at 10:29:38PM -0700, Andrew Morton wrote: > > I have problems with lvm2 and the 2.5.73-mm2 patchset. > Could be because -mm uses a 64-bit dev_t. > Could you please revert > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/broken-out/64-bit-dev_t-kdev_t.patch > and retest? yes, reverting this patch fixes the problem. Is it sane to run the -mm2 without this patch? > I think there's a version of devicemapper userspace which works correctly > with 64-bit dev_t. Perhaps the Sistina team can shed some light. would be nice if somebody else could comment on this.. -- Regards, Wiktor Wodecki [-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --] ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: [linux-lvm] Re: 2.5.73-mm2 2003-06-30 9:51 ` Wiktor Wodecki @ 2003-06-30 10:23 ` Joe Thornber 2003-06-30 10:41 ` Wiktor Wodecki 2003-06-30 12:29 ` Christophe Saout 1 sibling, 1 reply; 49+ messages in thread From: Joe Thornber @ 2003-06-30 10:23 UTC (permalink / raw) To: linux-lvm, Andrew Morton On Sun, Jun 29, 2003 at 12:32:26PM +0200, Wiktor Wodecki wrote: > On Sat, Jun 28, 2003 at 10:29:38PM -0700, Andrew Morton wrote: > > > I have problems with lvm2 and the 2.5.73-mm2 patchset. > > Could be because -mm uses a 64-bit dev_t. > > Could you please revert > > ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.5/2.5.73/2.5.73-mm2/broken-out/64-bit-dev_t-kdev_t.patch > > and retest? > > yes, reverting this patch fixes the problem. Is it sane to run the -mm2 > without this patch? > > > I think there's a version of devicemapper userspace which works correctly > > with 64-bit dev_t. Perhaps the Sistina team can shed some light. > > would be nice if somebody else could comment on this.. Yes, I think it's the 64bit dev_t. You'll have to wait for the new ioctl interface + userland tools before dm can make this transition. We've been having some problems with the tools, which caused this delay in fixing things. I *hope* the tools will be released this week. - Joe ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: [linux-lvm] Re: 2.5.73-mm2 2003-06-30 10:23 ` Joe Thornber @ 2003-06-30 10:41 ` Wiktor Wodecki 2003-06-30 11:41 ` Heinz J . Mauelshagen 0 siblings, 1 reply; 49+ messages in thread From: Wiktor Wodecki @ 2003-06-30 10:41 UTC (permalink / raw) To: linux-lvm On Mon, Jun 30, 2003 at 04:22:52PM +0100, Joe Thornber wrote: > Yes, I think it's the 64bit dev_t. You'll have to wait for the new > ioctl interface + userland tools before dm can make this transition. > We've been having some problems with the tools, which caused this > delay in fixing things. I *hope* the tools will be released this > week. ah okay, tell me if you want me to test them, I'd be glad to do help. -- Regards, Wiktor Wodecki ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: [linux-lvm] Re: 2.5.73-mm2 2003-06-30 10:41 ` Wiktor Wodecki @ 2003-06-30 11:41 ` Heinz J . Mauelshagen 0 siblings, 0 replies; 49+ messages in thread From: Heinz J . Mauelshagen @ 2003-06-30 11:41 UTC (permalink / raw) To: linux-lvm On Mon, Jun 30, 2003 at 05:43:54PM +0200, Wiktor Wodecki wrote: > On Mon, Jun 30, 2003 at 04:22:52PM +0100, Joe Thornber wrote: > > Yes, I think it's the 64bit dev_t. You'll have to wait for the new > > ioctl interface + userland tools before dm can make this transition. > > We've been having some problems with the tools, which caused this > > delay in fixing things. I *hope* the tools will be released this > > week. > > ah okay, tell me if you want me to test them, I'd be glad to do help. Thx :) > > -- > Regards, > > Wiktor Wodecki > > _______________________________________________ > linux-lvm mailing list > linux-lvm@sistina.com > http://lists.sistina.com/mailman/listinfo/linux-lvm > read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/ -- Regards, Heinz -- The LVM Guy -- *** Software bugs are stupid. Nevertheless it needs not so stupid people to solve them *** =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Heinz Mauelshagen Sistina Software Inc. Senior Consultant/Developer Am Sonnenhang 11 56242 Marienrachdorf Germany Mauelshagen@Sistina.com +49 2626 141200 FAX 924446 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: [linux-lvm] Re: 2.5.73-mm2 2003-06-30 9:51 ` Wiktor Wodecki 2003-06-30 10:23 ` Joe Thornber @ 2003-06-30 12:29 ` Christophe Saout 1 sibling, 0 replies; 49+ messages in thread From: Christophe Saout @ 2003-06-30 12:29 UTC (permalink / raw) To: linux-lvm; +Cc: Andrew Morton Am So, 2003-06-29 um 12.32 schrieb Wiktor Wodecki: > > I think there's a version of devicemapper userspace which works correctly > > with 64-bit dev_t. Perhaps the Sistina team can shed some light. > > would be nice if somebody else could comment on this.. Simply recompiling libdevmapper against the 2.5.7#-mm# kernel headers helps for now. But the plan for the future is that the dm ioctl interface will use 64 bit by default, so that it will work with any kernel version. -- Christophe Saout <christophe@saout.de> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-28 3:21 ` 2.5.73-mm2 Andrew Morton @ 2003-07-01 0:39 ` William Lee Irwin III -1 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-07-01 0:39 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-mm On Fri, Jun 27, 2003 at 08:21:30PM -0700, Andrew Morton wrote: > Just bits and pieces. It was suggested during my last round of OOM killer fixes that one of my patches, which just checked nr_free_buffer_pages() > 0, should also consider userspace (i.e. reclaimable at will) memory free. This patch implements that suggestion. Lightly tested, and expected to fall within the "relatively trivial" category. We're still not out of hot water here yet, since the minimum thresholds will still send all processes into perpetual torpor under persistent low memory conditions while fooling the OOM heuristics. But this is at least better than complete ignorance of ZONE_NORMAL exhaustion. -- wli diff -prauN wli-2.5.73-31/include/linux/mm.h wli-2.5.73-32/include/linux/mm.h --- wli-2.5.73-31/include/linux/mm.h 2003-06-29 01:39:42.000000000 -0700 +++ wli-2.5.73-32/include/linux/mm.h 2003-06-30 16:42:28.000000000 -0700 @@ -605,7 +605,8 @@ static inline struct vm_area_struct * fi extern struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr); -extern unsigned int nr_used_zone_pages(void); +unsigned int nr_used_low_pages(void); +unsigned int nr_used_zone_pages(void); extern struct page * vmalloc_to_page(void *addr); extern struct page * follow_page(struct mm_struct *mm, unsigned long address, diff -prauN wli-2.5.73-31/mm/oom_kill.c wli-2.5.73-32/mm/oom_kill.c --- wli-2.5.73-31/mm/oom_kill.c 2003-06-22 11:32:55.000000000 -0700 +++ wli-2.5.73-32/mm/oom_kill.c 2003-06-30 16:46:49.000000000 -0700 @@ -217,9 +217,9 @@ void out_of_memory(void) unsigned long now, since; /* - * Enough swap space left? Not OOM. + * Enough swap space and ZONE_NORMAL left? Not OOM. */ - if (nr_swap_pages > 0) + if (nr_swap_pages > 0 && nr_free_buffer_pages() + nr_used_low_pages() > 0) return; spin_lock(&oom_lock); diff -prauN wli-2.5.73-31/mm/page_alloc.c wli-2.5.73-32/mm/page_alloc.c --- wli-2.5.73-31/mm/page_alloc.c 2003-06-23 10:53:46.000000000 -0700 +++ wli-2.5.73-32/mm/page_alloc.c 2003-06-30 17:06:20.000000000 -0700 @@ -738,17 +738,6 @@ unsigned int nr_free_pages(void) } EXPORT_SYMBOL(nr_free_pages); -unsigned int nr_used_zone_pages(void) -{ - unsigned int pages = 0; - struct zone *zone; - - for_each_zone(zone) - pages += zone->nr_active + zone->nr_inactive; - - return pages; -} - #ifdef CONFIG_NUMA unsigned int nr_free_pages_pgdat(pg_data_t *pgdat) { @@ -782,6 +771,28 @@ static unsigned int nr_free_zone_pages(i return sum; } +static unsigned int __nr_used_zone_pages(int offset) +{ + struct zone *zone; + unsigned int sum = 0; + + for_each_zone(zone) + if (zone - zone->zone_pgdat->node_zones <= offset) + sum += zone->nr_active + zone->nr_inactive; + + return sum; +} + +unsigned int nr_used_zone_pages(void) +{ + return __nr_used_zone_pages(GFP_HIGHUSER & GFP_ZONEMASK); +} + +unsigned int nr_used_low_pages(void) +{ + return __nr_used_zone_pages(GFP_USER & GFP_ZONEMASK); +} + /* * Amount of free RAM allocatable within ZONE_DMA and ZONE_NORMAL */ ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-07-01 0:39 ` William Lee Irwin III 0 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-07-01 0:39 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-mm On Fri, Jun 27, 2003 at 08:21:30PM -0700, Andrew Morton wrote: > Just bits and pieces. It was suggested during my last round of OOM killer fixes that one of my patches, which just checked nr_free_buffer_pages() > 0, should also consider userspace (i.e. reclaimable at will) memory free. This patch implements that suggestion. Lightly tested, and expected to fall within the "relatively trivial" category. We're still not out of hot water here yet, since the minimum thresholds will still send all processes into perpetual torpor under persistent low memory conditions while fooling the OOM heuristics. But this is at least better than complete ignorance of ZONE_NORMAL exhaustion. -- wli diff -prauN wli-2.5.73-31/include/linux/mm.h wli-2.5.73-32/include/linux/mm.h --- wli-2.5.73-31/include/linux/mm.h 2003-06-29 01:39:42.000000000 -0700 +++ wli-2.5.73-32/include/linux/mm.h 2003-06-30 16:42:28.000000000 -0700 @@ -605,7 +605,8 @@ static inline struct vm_area_struct * fi extern struct vm_area_struct *find_extend_vma(struct mm_struct *mm, unsigned long addr); -extern unsigned int nr_used_zone_pages(void); +unsigned int nr_used_low_pages(void); +unsigned int nr_used_zone_pages(void); extern struct page * vmalloc_to_page(void *addr); extern struct page * follow_page(struct mm_struct *mm, unsigned long address, diff -prauN wli-2.5.73-31/mm/oom_kill.c wli-2.5.73-32/mm/oom_kill.c --- wli-2.5.73-31/mm/oom_kill.c 2003-06-22 11:32:55.000000000 -0700 +++ wli-2.5.73-32/mm/oom_kill.c 2003-06-30 16:46:49.000000000 -0700 @@ -217,9 +217,9 @@ void out_of_memory(void) unsigned long now, since; /* - * Enough swap space left? Not OOM. + * Enough swap space and ZONE_NORMAL left? Not OOM. */ - if (nr_swap_pages > 0) + if (nr_swap_pages > 0 && nr_free_buffer_pages() + nr_used_low_pages() > 0) return; spin_lock(&oom_lock); diff -prauN wli-2.5.73-31/mm/page_alloc.c wli-2.5.73-32/mm/page_alloc.c --- wli-2.5.73-31/mm/page_alloc.c 2003-06-23 10:53:46.000000000 -0700 +++ wli-2.5.73-32/mm/page_alloc.c 2003-06-30 17:06:20.000000000 -0700 @@ -738,17 +738,6 @@ unsigned int nr_free_pages(void) } EXPORT_SYMBOL(nr_free_pages); -unsigned int nr_used_zone_pages(void) -{ - unsigned int pages = 0; - struct zone *zone; - - for_each_zone(zone) - pages += zone->nr_active + zone->nr_inactive; - - return pages; -} - #ifdef CONFIG_NUMA unsigned int nr_free_pages_pgdat(pg_data_t *pgdat) { @@ -782,6 +771,28 @@ static unsigned int nr_free_zone_pages(i return sum; } +static unsigned int __nr_used_zone_pages(int offset) +{ + struct zone *zone; + unsigned int sum = 0; + + for_each_zone(zone) + if (zone - zone->zone_pgdat->node_zones <= offset) + sum += zone->nr_active + zone->nr_inactive; + + return sum; +} + +unsigned int nr_used_zone_pages(void) +{ + return __nr_used_zone_pages(GFP_HIGHUSER & GFP_ZONEMASK); +} + +unsigned int nr_used_low_pages(void) +{ + return __nr_used_zone_pages(GFP_USER & GFP_ZONEMASK); +} + /* * Amount of free RAM allocatable within ZONE_DMA and ZONE_NORMAL */ -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-07-01 0:39 ` 2.5.73-mm2 William Lee Irwin III @ 2003-07-01 2:14 ` Andrew Morton -1 siblings, 0 replies; 49+ messages in thread From: Andrew Morton @ 2003-07-01 2:14 UTC (permalink / raw) To: William Lee Irwin III; +Cc: linux-kernel, linux-mm William Lee Irwin III <wli@holomorphy.com> wrote: > > @@ -217,9 +217,9 @@ void out_of_memory(void) > unsigned long now, since; > > /* > - * Enough swap space left? Not OOM. > + * Enough swap space and ZONE_NORMAL left? Not OOM. > */ > - if (nr_swap_pages > 0) > + if (nr_swap_pages > 0 && nr_free_buffer_pages() + nr_used_low_pages() > 0) > return; a) if someone is trying to allocate some ZONE_DMA pages and there are still swappable or free ZONE_NORMAL pages, nobody gets killed. b) If there are free ZONE_NORMAL pages then why on earth did we call out_of_memory()? Does nr_free_buffer_pages() ever return non-zero in here? It will do so for a ZONE_DMA allocation, but you're not doing them... Generally, I'm thinking that this test should just be removed. It is the responsibility of try_to_free_pages() to work out whether the allocation can succeed. If try_to_free_pages() calls out_of_memory() when there are still swappable, reclaimable or free pages in the relevant zones then try_to_free_pages() goofed, and needs mending. out_of_memory() shouldn't be cleaning up after try_to_free_pages()'s mistakes. I have a bad feeling that it _will_ goof. A long time ago I looked at the amount of scanning we're doing in there and decided that it was way overkill and reduced it by a lot. I may have gone overboard. So how's about I and thy take that test out, see how things get along? ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-07-01 2:14 ` Andrew Morton 0 siblings, 0 replies; 49+ messages in thread From: Andrew Morton @ 2003-07-01 2:14 UTC (permalink / raw) To: William Lee Irwin III; +Cc: linux-kernel, linux-mm William Lee Irwin III <wli@holomorphy.com> wrote: > > @@ -217,9 +217,9 @@ void out_of_memory(void) > unsigned long now, since; > > /* > - * Enough swap space left? Not OOM. > + * Enough swap space and ZONE_NORMAL left? Not OOM. > */ > - if (nr_swap_pages > 0) > + if (nr_swap_pages > 0 && nr_free_buffer_pages() + nr_used_low_pages() > 0) > return; a) if someone is trying to allocate some ZONE_DMA pages and there are still swappable or free ZONE_NORMAL pages, nobody gets killed. b) If there are free ZONE_NORMAL pages then why on earth did we call out_of_memory()? Does nr_free_buffer_pages() ever return non-zero in here? It will do so for a ZONE_DMA allocation, but you're not doing them... Generally, I'm thinking that this test should just be removed. It is the responsibility of try_to_free_pages() to work out whether the allocation can succeed. If try_to_free_pages() calls out_of_memory() when there are still swappable, reclaimable or free pages in the relevant zones then try_to_free_pages() goofed, and needs mending. out_of_memory() shouldn't be cleaning up after try_to_free_pages()'s mistakes. I have a bad feeling that it _will_ goof. A long time ago I looked at the amount of scanning we're doing in there and decided that it was way overkill and reduced it by a lot. I may have gone overboard. So how's about I and thy take that test out, see how things get along? -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-07-01 2:14 ` 2.5.73-mm2 Andrew Morton @ 2003-07-01 2:46 ` William Lee Irwin III -1 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-07-01 2:46 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-mm William Lee Irwin III <wli@holomorphy.com> wrote: >> @@ -217,9 +217,9 @@ void out_of_memory(void) >> unsigned long now, since; >> >> /* >> - * Enough swap space left? Not OOM. >> + * Enough swap space and ZONE_NORMAL left? Not OOM. >> */ >> - if (nr_swap_pages > 0) >> + if (nr_swap_pages > 0 && nr_free_buffer_pages() + nr_used_low_pages() > 0) >> return; On Mon, Jun 30, 2003 at 07:14:56PM -0700, Andrew Morton wrote: > a) if someone is trying to allocate some ZONE_DMA pages and there are > still swappable or free ZONE_NORMAL pages, nobody gets killed. This is yet another problem for the method above. =( On Mon, Jun 30, 2003 at 07:14:56PM -0700, Andrew Morton wrote: > b) If there are free ZONE_NORMAL pages then why on earth did we call > out_of_memory()? Does nr_free_buffer_pages() ever return non-zero in > here? It will do so for a ZONE_DMA allocation, but you're not doing > them... Allocations will enter this path if free memory is below the minimum page thresholds, since the allocation will be sort of artificially failed. Basically, with this in place it's more likely to livelock than to go on killing sprees. There's a small amount of empirical evidence suggesting this avoids livelocking in some common scenarios, though that really isn't good enough for this kind of affair. On Mon, Jun 30, 2003 at 07:14:56PM -0700, Andrew Morton wrote: > Generally, I'm thinking that this test should just be removed. It is > the responsibility of try_to_free_pages() to work out whether the > allocation can succeed. > If try_to_free_pages() calls out_of_memory() when there are still > swappable, reclaimable or free pages in the relevant zones then > try_to_free_pages() goofed, and needs mending. out_of_memory() > shouldn't be cleaning up after try_to_free_pages()'s mistakes. > I have a bad feeling that it _will_ goof. A long time ago I looked > at the amount of scanning we're doing in there and decided that it > was way overkill and reduced it by a lot. I may have gone overboard. > So how's about I and thy take that test out, see how things get along? I'm not particularly attached to the method, only the result, so I'm game. -- wli ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-07-01 2:46 ` William Lee Irwin III 0 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-07-01 2:46 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-mm William Lee Irwin III <wli@holomorphy.com> wrote: >> @@ -217,9 +217,9 @@ void out_of_memory(void) >> unsigned long now, since; >> >> /* >> - * Enough swap space left? Not OOM. >> + * Enough swap space and ZONE_NORMAL left? Not OOM. >> */ >> - if (nr_swap_pages > 0) >> + if (nr_swap_pages > 0 && nr_free_buffer_pages() + nr_used_low_pages() > 0) >> return; On Mon, Jun 30, 2003 at 07:14:56PM -0700, Andrew Morton wrote: > a) if someone is trying to allocate some ZONE_DMA pages and there are > still swappable or free ZONE_NORMAL pages, nobody gets killed. This is yet another problem for the method above. =( On Mon, Jun 30, 2003 at 07:14:56PM -0700, Andrew Morton wrote: > b) If there are free ZONE_NORMAL pages then why on earth did we call > out_of_memory()? Does nr_free_buffer_pages() ever return non-zero in > here? It will do so for a ZONE_DMA allocation, but you're not doing > them... Allocations will enter this path if free memory is below the minimum page thresholds, since the allocation will be sort of artificially failed. Basically, with this in place it's more likely to livelock than to go on killing sprees. There's a small amount of empirical evidence suggesting this avoids livelocking in some common scenarios, though that really isn't good enough for this kind of affair. On Mon, Jun 30, 2003 at 07:14:56PM -0700, Andrew Morton wrote: > Generally, I'm thinking that this test should just be removed. It is > the responsibility of try_to_free_pages() to work out whether the > allocation can succeed. > If try_to_free_pages() calls out_of_memory() when there are still > swappable, reclaimable or free pages in the relevant zones then > try_to_free_pages() goofed, and needs mending. out_of_memory() > shouldn't be cleaning up after try_to_free_pages()'s mistakes. > I have a bad feeling that it _will_ goof. A long time ago I looked > at the amount of scanning we're doing in there and decided that it > was way overkill and reduced it by a lot. I may have gone overboard. > So how's about I and thy take that test out, see how things get along? I'm not particularly attached to the method, only the result, so I'm game. -- wli -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-07-01 0:39 ` 2.5.73-mm2 William Lee Irwin III @ 2003-07-01 10:46 ` Hugh Dickins -1 siblings, 0 replies; 49+ messages in thread From: Hugh Dickins @ 2003-07-01 10:46 UTC (permalink / raw) To: William Lee Irwin III; +Cc: Andrew Morton, linux-kernel, linux-mm On Mon, 30 Jun 2003, William Lee Irwin III wrote: > > It was suggested during my last round of OOM killer fixes that one of > my patches, which just checked nr_free_buffer_pages() > 0, should also > consider userspace (i.e. reclaimable at will) memory free. If you pursued it, wouldn't your patch also need to change nr_free_buffer_pages() to do what you think it does, count the free lowmem pages? It, and nr_free_pagecache_pages(), and nr_free_zone_pages(), are horribly badly named. They count present_pages-pages_high, they don't count free pages: okay for initialization estimates, useless for anything dynamic. Hugh p.s. any chance of some more imaginative Subject lines :-? ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-07-01 10:46 ` Hugh Dickins 0 siblings, 0 replies; 49+ messages in thread From: Hugh Dickins @ 2003-07-01 10:46 UTC (permalink / raw) To: William Lee Irwin III; +Cc: Andrew Morton, linux-kernel, linux-mm On Mon, 30 Jun 2003, William Lee Irwin III wrote: > > It was suggested during my last round of OOM killer fixes that one of > my patches, which just checked nr_free_buffer_pages() > 0, should also > consider userspace (i.e. reclaimable at will) memory free. If you pursued it, wouldn't your patch also need to change nr_free_buffer_pages() to do what you think it does, count the free lowmem pages? It, and nr_free_pagecache_pages(), and nr_free_zone_pages(), are horribly badly named. They count present_pages-pages_high, they don't count free pages: okay for initialization estimates, useless for anything dynamic. Hugh p.s. any chance of some more imaginative Subject lines :-? -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-07-01 10:46 ` 2.5.73-mm2 Hugh Dickins @ 2003-07-01 10:51 ` William Lee Irwin III -1 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-07-01 10:51 UTC (permalink / raw) To: Hugh Dickins; +Cc: Andrew Morton, linux-kernel, linux-mm On Mon, 30 Jun 2003, William Lee Irwin III wrote: >> It was suggested during my last round of OOM killer fixes that one of >> my patches, which just checked nr_free_buffer_pages() > 0, should also >> consider userspace (i.e. reclaimable at will) memory free. On Tue, Jul 01, 2003 at 11:46:34AM +0100, Hugh Dickins wrote: > If you pursued it, wouldn't your patch also need to change > nr_free_buffer_pages() to do what you think it does, count > the free lowmem pages? It, and nr_free_pagecache_pages(), > and nr_free_zone_pages(), are horribly badly named. They > count present_pages-pages_high, they don't count free pages: > okay for initialization estimates, useless for anything dynamic. > Hugh > p.s. any chance of some more imaginative Subject lines :-? Well, I was mostly looking for getting handed back 0 when lowmem is empty; I actually did realize they didn't give entirely accurate counts of free lowmem pages. -- wli ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-07-01 10:51 ` William Lee Irwin III 0 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-07-01 10:51 UTC (permalink / raw) To: Hugh Dickins; +Cc: Andrew Morton, linux-kernel, linux-mm On Mon, 30 Jun 2003, William Lee Irwin III wrote: >> It was suggested during my last round of OOM killer fixes that one of >> my patches, which just checked nr_free_buffer_pages() > 0, should also >> consider userspace (i.e. reclaimable at will) memory free. On Tue, Jul 01, 2003 at 11:46:34AM +0100, Hugh Dickins wrote: > If you pursued it, wouldn't your patch also need to change > nr_free_buffer_pages() to do what you think it does, count > the free lowmem pages? It, and nr_free_pagecache_pages(), > and nr_free_zone_pages(), are horribly badly named. They > count present_pages-pages_high, they don't count free pages: > okay for initialization estimates, useless for anything dynamic. > Hugh > p.s. any chance of some more imaginative Subject lines :-? Well, I was mostly looking for getting handed back 0 when lowmem is empty; I actually did realize they didn't give entirely accurate counts of free lowmem pages. -- wli -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-07-01 10:51 ` 2.5.73-mm2 William Lee Irwin III @ 2003-07-01 11:08 ` Hugh Dickins -1 siblings, 0 replies; 49+ messages in thread From: Hugh Dickins @ 2003-07-01 11:08 UTC (permalink / raw) To: William Lee Irwin III; +Cc: Andrew Morton, linux-kernel, linux-mm On Tue, 1 Jul 2003, William Lee Irwin III wrote: > On Tue, Jul 01, 2003 at 11:46:34AM +0100, Hugh Dickins wrote: > > If you pursued it, wouldn't your patch also need to change > > nr_free_buffer_pages() to do what you think it does, count > > the free lowmem pages? It, and nr_free_pagecache_pages(), > > and nr_free_zone_pages(), are horribly badly named. They > > count present_pages-pages_high, they don't count free pages: > > okay for initialization estimates, useless for anything dynamic. > > Well, I was mostly looking for getting handed back 0 when lowmem is > empty; I actually did realize they didn't give entirely accurate counts > of free lowmem pages. I'm not pleading for complete accuracy, but nr_free_buffer_pages() will never hand back 0 (if your system managed to boot). It's a static count of present_pages (adjusted), not of free pages. Or am I misreading nr_free_zone_pages()? Hugh ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-07-01 11:08 ` Hugh Dickins 0 siblings, 0 replies; 49+ messages in thread From: Hugh Dickins @ 2003-07-01 11:08 UTC (permalink / raw) To: William Lee Irwin III; +Cc: Andrew Morton, linux-kernel, linux-mm On Tue, 1 Jul 2003, William Lee Irwin III wrote: > On Tue, Jul 01, 2003 at 11:46:34AM +0100, Hugh Dickins wrote: > > If you pursued it, wouldn't your patch also need to change > > nr_free_buffer_pages() to do what you think it does, count > > the free lowmem pages? It, and nr_free_pagecache_pages(), > > and nr_free_zone_pages(), are horribly badly named. They > > count present_pages-pages_high, they don't count free pages: > > okay for initialization estimates, useless for anything dynamic. > > Well, I was mostly looking for getting handed back 0 when lowmem is > empty; I actually did realize they didn't give entirely accurate counts > of free lowmem pages. I'm not pleading for complete accuracy, but nr_free_buffer_pages() will never hand back 0 (if your system managed to boot). It's a static count of present_pages (adjusted), not of free pages. Or am I misreading nr_free_zone_pages()? Hugh -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-07-01 11:08 ` 2.5.73-mm2 Hugh Dickins @ 2003-07-01 11:08 ` William Lee Irwin III -1 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-07-01 11:08 UTC (permalink / raw) To: Hugh Dickins; +Cc: Andrew Morton, linux-kernel, linux-mm On Tue, 1 Jul 2003, William Lee Irwin III wrote: >> Well, I was mostly looking for getting handed back 0 when lowmem is >> empty; I actually did realize they didn't give entirely accurate counts >> of free lowmem pages. On Tue, Jul 01, 2003 at 12:08:03PM +0100, Hugh Dickins wrote: > I'm not pleading for complete accuracy, but nr_free_buffer_pages() > will never hand back 0 (if your system managed to boot). > It's a static count of present_pages (adjusted), not of > free pages. Or am I misreading nr_free_zone_pages()? You're right. Wow, that's even more worse than I suspected. -- wli ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-07-01 11:08 ` William Lee Irwin III 0 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-07-01 11:08 UTC (permalink / raw) To: Hugh Dickins; +Cc: Andrew Morton, linux-kernel, linux-mm On Tue, 1 Jul 2003, William Lee Irwin III wrote: >> Well, I was mostly looking for getting handed back 0 when lowmem is >> empty; I actually did realize they didn't give entirely accurate counts >> of free lowmem pages. On Tue, Jul 01, 2003 at 12:08:03PM +0100, Hugh Dickins wrote: > I'm not pleading for complete accuracy, but nr_free_buffer_pages() > will never hand back 0 (if your system managed to boot). > It's a static count of present_pages (adjusted), not of > free pages. Or am I misreading nr_free_zone_pages()? You're right. Wow, that's even more worse than I suspected. -- wli -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-07-01 11:08 ` 2.5.73-mm2 William Lee Irwin III @ 2003-07-01 12:39 ` Nikita Danilov -1 siblings, 0 replies; 49+ messages in thread From: Nikita Danilov @ 2003-07-01 12:39 UTC (permalink / raw) To: William Lee Irwin III; +Cc: Hugh Dickins, Andrew Morton, linux-kernel, linux-mm William Lee Irwin III writes: > On Tue, 1 Jul 2003, William Lee Irwin III wrote: > >> Well, I was mostly looking for getting handed back 0 when lowmem is > >> empty; I actually did realize they didn't give entirely accurate counts > >> of free lowmem pages. > > On Tue, Jul 01, 2003 at 12:08:03PM +0100, Hugh Dickins wrote: > > I'm not pleading for complete accuracy, but nr_free_buffer_pages() > > will never hand back 0 (if your system managed to boot). > > It's a static count of present_pages (adjusted), not of > > free pages. Or am I misreading nr_free_zone_pages()? > > You're right. Wow, that's even more worse than I suspected. > Another thing is that if one boots with mem=X, nr_free_pagecache_pages() returns X. However part of X (occupied by kernel image, etc) is not part of any zone. As a result, zone actually contains fewer pages than reported by nr_free_pagecache_pages(). With X small enough (comparable with kernel image size, for example) this can confuse balance_dirty_pages() enough so that throttling would never start, and VM will oom_kill(). > > -- wli Nikita. ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 @ 2003-07-01 12:39 ` Nikita Danilov 0 siblings, 0 replies; 49+ messages in thread From: Nikita Danilov @ 2003-07-01 12:39 UTC (permalink / raw) To: William Lee Irwin III; +Cc: Hugh Dickins, Andrew Morton, linux-kernel, linux-mm William Lee Irwin III writes: > On Tue, 1 Jul 2003, William Lee Irwin III wrote: > >> Well, I was mostly looking for getting handed back 0 when lowmem is > >> empty; I actually did realize they didn't give entirely accurate counts > >> of free lowmem pages. > > On Tue, Jul 01, 2003 at 12:08:03PM +0100, Hugh Dickins wrote: > > I'm not pleading for complete accuracy, but nr_free_buffer_pages() > > will never hand back 0 (if your system managed to boot). > > It's a static count of present_pages (adjusted), not of > > free pages. Or am I misreading nr_free_zone_pages()? > > You're right. Wow, that's even more worse than I suspected. > Another thing is that if one boots with mem=X, nr_free_pagecache_pages() returns X. However part of X (occupied by kernel image, etc) is not part of any zone. As a result, zone actually contains fewer pages than reported by nr_free_pagecache_pages(). With X small enough (comparable with kernel image size, for example) this can confuse balance_dirty_pages() enough so that throttling would never start, and VM will oom_kill(). > > -- wli Nikita. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a> ^ permalink raw reply [flat|nested] 49+ messages in thread
* Re: 2.5.73-mm2 2003-06-28 3:21 ` 2.5.73-mm2 Andrew Morton ` (6 preceding siblings ...) (?) @ 2003-07-01 5:56 ` William Lee Irwin III -1 siblings, 0 replies; 49+ messages in thread From: William Lee Irwin III @ 2003-07-01 5:56 UTC (permalink / raw) To: Andrew Morton; +Cc: linux-kernel, linux-mm [-- Attachment #1: brief message --] [-- Type: text/plain, Size: 563 bytes --] On Fri, Jun 27, 2003 at 08:21:30PM -0700, Andrew Morton wrote: > Just bits and pieces. And here is cpumask_t. This enables architectures with NR_CPUS > BITS_PER_LONG to utilize all those cpus. Tested on ppc64. This unfortunately has not undergone compiletesting for all architectures, so some amount of source-level breakage is implied. However, the fixups required should be very simple once some kind of compiletesting is done. This patch is sent as a MIME attachment and compressed in order to avoid various MTA's barfing on messages of this size. -- wli [-- Attachment #2: cpumask_t-2.5.73-mm2-1.bz2 --] [-- Type: application/octet-stream, Size: 36623 bytes --] ^ permalink raw reply [flat|nested] 49+ messages in thread
end of thread, other threads:[~2003-07-02 2:57 UTC | newest] Thread overview: 49+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2003-06-28 3:21 2.5.73-mm2 Andrew Morton 2003-06-28 3:21 ` 2.5.73-mm2 Andrew Morton 2003-06-28 8:56 ` 2.5.73-mm2 William Lee Irwin III 2003-06-28 8:56 ` 2.5.73-mm2 William Lee Irwin III 2003-06-28 15:54 ` 2.5.73-mm2 William Lee Irwin III 2003-06-28 15:54 ` 2.5.73-mm2 William Lee Irwin III 2003-06-28 16:08 ` 2.5.73-mm2 Christoph Hellwig 2003-06-28 16:08 ` 2.5.73-mm2 Christoph Hellwig 2003-06-28 20:49 ` 2.5.73-mm2 William Lee Irwin III 2003-06-29 0:34 ` 2.5.73-mm2 Martin J. Bligh 2003-06-29 0:34 ` 2.5.73-mm2 Martin J. Bligh 2003-06-29 2:18 ` 2.5.73-mm2 William Lee Irwin III 2003-06-29 2:18 ` 2.5.73-mm2 William Lee Irwin III 2003-06-29 3:07 ` 2.5.73-mm2 Martin J. Bligh 2003-06-29 3:07 ` 2.5.73-mm2 Martin J. Bligh 2003-06-28 23:00 ` 2.5.73-mm2 Andrew Morton 2003-06-28 23:00 ` 2.5.73-mm2 Andrew Morton 2003-06-28 23:11 ` 2.5.73-mm2 William Lee Irwin III 2003-06-28 23:11 ` 2.5.73-mm2 William Lee Irwin III 2003-06-29 12:45 ` 2.5.73-mm2 Zwane Mwaikambo 2003-06-29 12:45 ` 2.5.73-mm2 Zwane Mwaikambo 2003-07-02 3:11 ` 2.5.73-mm2 William Lee Irwin III 2003-06-29 19:04 ` [patch] 2.5.73-mm2: let CONFIG_TC35815 depend on CONFIG_TOSHIBA_JMR3927 Adrian Bunk 2003-06-29 21:04 ` Ralf Baechle 2003-06-29 19:04 ` Adrian Bunk 2003-06-30 9:51 ` [linux-lvm] Re: 2.5.73-mm2 Wiktor Wodecki 2003-06-29 0:30 ` Andrew Morton 2003-06-30 9:51 ` Wiktor Wodecki 2003-06-30 10:23 ` Joe Thornber 2003-06-30 10:41 ` Wiktor Wodecki 2003-06-30 11:41 ` Heinz J . Mauelshagen 2003-06-30 12:29 ` Christophe Saout 2003-07-01 0:39 ` 2.5.73-mm2 William Lee Irwin III 2003-07-01 0:39 ` 2.5.73-mm2 William Lee Irwin III 2003-07-01 2:14 ` 2.5.73-mm2 Andrew Morton 2003-07-01 2:14 ` 2.5.73-mm2 Andrew Morton 2003-07-01 2:46 ` 2.5.73-mm2 William Lee Irwin III 2003-07-01 2:46 ` 2.5.73-mm2 William Lee Irwin III 2003-07-01 10:46 ` 2.5.73-mm2 Hugh Dickins 2003-07-01 10:46 ` 2.5.73-mm2 Hugh Dickins 2003-07-01 10:51 ` 2.5.73-mm2 William Lee Irwin III 2003-07-01 10:51 ` 2.5.73-mm2 William Lee Irwin III 2003-07-01 11:08 ` 2.5.73-mm2 Hugh Dickins 2003-07-01 11:08 ` 2.5.73-mm2 Hugh Dickins 2003-07-01 11:08 ` 2.5.73-mm2 William Lee Irwin III 2003-07-01 11:08 ` 2.5.73-mm2 William Lee Irwin III 2003-07-01 12:39 ` 2.5.73-mm2 Nikita Danilov 2003-07-01 12:39 ` 2.5.73-mm2 Nikita Danilov 2003-07-01 5:56 ` 2.5.73-mm2 William Lee Irwin III
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.