Merge branch 'stable/swiotlb-0.9' of git://git.kernel.org/pub/scm/linux/kernel/git...
authorLinus Torvalds <torvalds@linux-foundation.org>
Thu, 21 Oct 2010 21:04:25 +0000 (14:04 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 21 Oct 2010 21:04:25 +0000 (14:04 -0700)
* 'stable/swiotlb-0.9' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/swiotlb-2.6:
  swiotlb: Use page alignment for early buffer allocation
  swiotlb: make io_tlb_overflow static

708 files changed:
Documentation/DocBook/kernel-locking.tmpl
Documentation/RCU/checklist.txt
Documentation/RCU/stallwarn.txt
Documentation/RCU/trace.txt
Documentation/cputopology.txt
Documentation/feature-removal-schedule.txt
Documentation/kernel-parameters.txt
Documentation/kprobes.txt
Documentation/networking/e1000.txt
Documentation/networking/e1000e.txt [new file with mode: 0644]
Documentation/networking/ixgbevf.txt [changed mode: 0755->0644]
Documentation/vm/page-types.c
MAINTAINERS
Makefile
arch/Kconfig
arch/alpha/Kconfig
arch/alpha/include/asm/perf_event.h
arch/alpha/kernel/perf_event.c
arch/alpha/kernel/time.c
arch/arm/Kconfig
arch/arm/include/asm/perf_event.h
arch/arm/kernel/kprobes-decode.c
arch/arm/kernel/perf_event.c
arch/arm/mach-at91/include/mach/system.h
arch/arm/mach-bcmring/dma.c
arch/arm/mach-ep93xx/dma-m2p.c
arch/arm/mach-imx/Kconfig
arch/arm/mach-imx/mach-cpuimx27.c
arch/arm/mach-s5p6440/cpu.c
arch/arm/mach-s5p6442/cpu.c
arch/arm/mach-s5pc100/cpu.c
arch/arm/mach-s5pv210/clock.c
arch/arm/mach-s5pv210/cpu.c
arch/arm/mach-vexpress/ct-ca9x4.c
arch/arm/mach-vexpress/v2m.c
arch/arm/mm/ioremap.c
arch/arm/mm/mmu.c
arch/arm/mm/proc-v7.S
arch/arm/oprofile/Makefile
arch/arm/oprofile/common.c
arch/arm/plat-omap/iommu.c
arch/arm/plat-samsung/adc.c
arch/arm/plat-samsung/clock.c
arch/frv/Kconfig
arch/frv/lib/Makefile
arch/frv/lib/perf_event.c [deleted file]
arch/ia64/include/asm/hardirq.h
arch/ia64/include/asm/system.h
arch/m32r/include/asm/elf.h
arch/m32r/kernel/.gitignore [new file with mode: 0644]
arch/m32r/kernel/signal.c
arch/mips/Kbuild
arch/mips/Kconfig
arch/mips/boot/compressed/Makefile
arch/mips/dec/Platform
arch/mips/include/asm/fcntl.h
arch/mips/include/asm/siginfo.h
arch/mips/jz4740/Platform
arch/mips/kernel/branch.c
arch/mips/kernel/mips-mt-fpaff.c
arch/mips/kernel/ptrace.c
arch/mips/kernel/scall32-o32.S
arch/mips/kernel/scall64-64.S
arch/mips/kernel/scall64-n32.S
arch/mips/kernel/scall64-o32.S
arch/mips/kernel/signal.c
arch/mips/kernel/signal_n32.c
arch/mips/kernel/unaligned.c
arch/parisc/Kconfig
arch/parisc/include/asm/perf_event.h
arch/powerpc/Kconfig
arch/powerpc/include/asm/paca.h
arch/powerpc/include/asm/system.h
arch/powerpc/kernel/perf_callchain.c
arch/powerpc/kernel/perf_event.c
arch/powerpc/kernel/perf_event_fsl_emb.c
arch/powerpc/kernel/time.c
arch/s390/Kconfig
arch/s390/include/asm/hardirq.h
arch/s390/include/asm/perf_event.h
arch/s390/include/asm/system.h
arch/s390/include/asm/topology.h
arch/s390/kernel/topology.c
arch/sh/Kconfig
arch/sh/include/asm/perf_event.h
arch/sh/kernel/perf_callchain.c
arch/sh/kernel/perf_event.c
arch/sh/oprofile/Makefile
arch/sh/oprofile/common.c
arch/sh/oprofile/op_impl.h [deleted file]
arch/sparc/Kconfig
arch/sparc/include/asm/jump_label.h [new file with mode: 0644]
arch/sparc/include/asm/perf_event.h
arch/sparc/kernel/Makefile
arch/sparc/kernel/jump_label.c [new file with mode: 0644]
arch/sparc/kernel/module.c
arch/sparc/kernel/pcr.c
arch/sparc/kernel/perf_event.c
arch/um/drivers/hostaudio_kern.c
arch/um/drivers/ubd_kern.c
arch/x86/Kconfig
arch/x86/Kconfig.debug
arch/x86/Makefile
arch/x86/ia32/ia32_aout.c
arch/x86/include/asm/alternative.h
arch/x86/include/asm/amd_iommu.h
arch/x86/include/asm/amd_iommu_proto.h
arch/x86/include/asm/amd_iommu_types.h
arch/x86/include/asm/amd_nb.h [moved from arch/x86/include/asm/k8.h with 61% similarity]
arch/x86/include/asm/apb_timer.h
arch/x86/include/asm/cpu.h
arch/x86/include/asm/cpufeature.h
arch/x86/include/asm/dwarf2.h
arch/x86/include/asm/entry_arch.h
arch/x86/include/asm/fixmap.h
arch/x86/include/asm/gart.h
arch/x86/include/asm/hardirq.h
arch/x86/include/asm/hw_irq.h
arch/x86/include/asm/i387.h
arch/x86/include/asm/io.h
arch/x86/include/asm/irq_remapping.h
arch/x86/include/asm/irq_vectors.h
arch/x86/include/asm/jump_label.h [new file with mode: 0644]
arch/x86/include/asm/kvm_host.h
arch/x86/include/asm/mrst.h
arch/x86/include/asm/mwait.h [new file with mode: 0644]
arch/x86/include/asm/olpc_ofw.h
arch/x86/include/asm/page_types.h
arch/x86/include/asm/paravirt.h
arch/x86/include/asm/paravirt_types.h
arch/x86/include/asm/perf_event_p4.h
arch/x86/include/asm/pgtable.h
arch/x86/include/asm/pgtable_64.h
arch/x86/include/asm/processor.h
arch/x86/include/asm/setup.h
arch/x86/include/asm/vmi.h [deleted file]
arch/x86/include/asm/vmi_time.h [deleted file]
arch/x86/kernel/Makefile
arch/x86/kernel/acpi/cstate.c
arch/x86/kernel/alternative.c
arch/x86/kernel/amd_iommu.c
arch/x86/kernel/amd_iommu_init.c
arch/x86/kernel/amd_nb.c [moved from arch/x86/kernel/k8.c with 66% similarity]
arch/x86/kernel/apb_timer.c
arch/x86/kernel/aperture_64.c
arch/x86/kernel/apic/apic.c
arch/x86/kernel/apic/io_apic.c
arch/x86/kernel/apic/probe_64.c
arch/x86/kernel/cpu/amd.c
arch/x86/kernel/cpu/common.c
arch/x86/kernel/cpu/cpu.h
arch/x86/kernel/cpu/intel.c
arch/x86/kernel/cpu/intel_cacheinfo.c
arch/x86/kernel/cpu/mcheck/mce_amd.c
arch/x86/kernel/cpu/mcheck/therm_throt.c
arch/x86/kernel/cpu/mtrr/cleanup.c
arch/x86/kernel/cpu/mtrr/generic.c
arch/x86/kernel/cpu/perf_event.c
arch/x86/kernel/cpu/perf_event_amd.c
arch/x86/kernel/cpu/perf_event_intel.c
arch/x86/kernel/cpu/perf_event_intel_ds.c
arch/x86/kernel/cpu/perf_event_p4.c
arch/x86/kernel/cpu/perfctr-watchdog.c
arch/x86/kernel/cpu/scattered.c
arch/x86/kernel/crash_dump_64.c
arch/x86/kernel/early-quirks.c
arch/x86/kernel/early_printk.c
arch/x86/kernel/early_printk_mrst.c [new file with mode: 0644]
arch/x86/kernel/entry_32.S
arch/x86/kernel/entry_64.S
arch/x86/kernel/ftrace.c
arch/x86/kernel/i387.c
arch/x86/kernel/irq.c
arch/x86/kernel/irq_work.c [new file with mode: 0644]
arch/x86/kernel/irqinit.c
arch/x86/kernel/jump_label.c [new file with mode: 0644]
arch/x86/kernel/kprobes.c
arch/x86/kernel/machine_kexec_64.c
arch/x86/kernel/module.c
arch/x86/kernel/olpc-xo1.c [new file with mode: 0644]
arch/x86/kernel/olpc.c
arch/x86/kernel/olpc_ofw.c
arch/x86/kernel/paravirt.c
arch/x86/kernel/pci-gart_64.c
arch/x86/kernel/pmtimer_64.c [deleted file]
arch/x86/kernel/process_64.c
arch/x86/kernel/reboot.c
arch/x86/kernel/setup.c
arch/x86/kernel/setup_percpu.c
arch/x86/kernel/sfi.c
arch/x86/kernel/smpboot.c
arch/x86/kernel/sys_i386_32.c
arch/x86/kernel/traps.c
arch/x86/kernel/tsc.c
arch/x86/kernel/vmi_32.c [deleted file]
arch/x86/kernel/vmiclock_32.c [deleted file]
arch/x86/kvm/lapic.c
arch/x86/kvm/svm.c
arch/x86/kvm/vmx.c
arch/x86/kvm/x86.c
arch/x86/lib/memcpy_32.c
arch/x86/lib/memcpy_64.S
arch/x86/lib/memmove_64.c
arch/x86/mm/fault.c
arch/x86/mm/init_32.c
arch/x86/mm/init_64.c
arch/x86/mm/k8topology_64.c
arch/x86/mm/kmemcheck/kmemcheck.c
arch/x86/mm/kmemcheck/opcode.c
arch/x86/mm/numa_64.c
arch/x86/mm/pgtable.c
arch/x86/mm/srat_64.c
arch/x86/mm/tlb.c
arch/x86/oprofile/backtrace.c
arch/x86/oprofile/nmi_int.c
arch/x86/pci/olpc.c
arch/x86/xen/mmu.c
block/bsg.c
block/elevator.c
drivers/acpi/acpi_pad.c
drivers/acpi/blacklist.c
drivers/acpi/processor_core.c
drivers/atm/iphase.c
drivers/atm/iphase.h
drivers/atm/solos-pci.c
drivers/base/topology.c
drivers/block/Kconfig
drivers/block/Makefile
drivers/block/ps3disk.c
drivers/block/rbd.c [new file with mode: 0644]
drivers/block/rbd_types.h [new file with mode: 0644]
drivers/block/virtio_blk.c
drivers/char/agp/Kconfig
drivers/char/agp/amd64-agp.c
drivers/char/agp/generic.c
drivers/char/tpm/tpm.c
drivers/char/virtio_console.c
drivers/dma/ioat/dma_v2.c
drivers/edac/Kconfig
drivers/edac/amd64_edac.c
drivers/firewire/ohci.c
drivers/firewire/ohci.h
drivers/firmware/Kconfig
drivers/gpu/drm/i915/i915_dma.c
drivers/gpu/drm/i915/intel_fb.c
drivers/gpu/drm/nouveau/nouveau_fbcon.c
drivers/gpu/drm/nouveau/nouveau_notifier.c
drivers/gpu/drm/radeon/evergreen.c
drivers/gpu/drm/radeon/r100.c
drivers/gpu/drm/radeon/r600.c
drivers/gpu/drm/radeon/r600_blit_kms.c
drivers/gpu/drm/radeon/radeon.h
drivers/gpu/drm/radeon/radeon_atombios.c
drivers/gpu/drm/radeon/radeon_combios.c
drivers/gpu/drm/radeon/radeon_cursor.c
drivers/gpu/drm/radeon/radeon_fb.c
drivers/gpu/drm/radeon/radeon_object.c
drivers/gpu/drm/radeon/radeon_object.h
drivers/gpu/drm/radeon/rs600.c
drivers/gpu/drm/radeon/rs690.c
drivers/gpu/drm/radeon/rv770.c
drivers/gpu/drm/ttm/ttm_bo.c
drivers/hid/hid-cando.c
drivers/hid/hid-core.c
drivers/hid/hid-ids.h
drivers/hid/hidraw.c
drivers/hid/usbhid/hid-quirks.c
drivers/i2c/busses/i2c-cpm.c
drivers/i2c/busses/i2c-davinci.c
drivers/i2c/busses/i2c-ibm_iic.c
drivers/i2c/busses/i2c-imx.c
drivers/i2c/busses/i2c-mpc.c
drivers/i2c/busses/i2c-pca-isa.c
drivers/i2c/busses/i2c-pca-platform.c
drivers/i2c/i2c-core.c
drivers/idle/intel_idle.c
drivers/input/evdev.c
drivers/input/joydev.c
drivers/input/misc/hp_sdc_rtc.c
drivers/input/misc/uinput.c
drivers/input/serio/hil_mlc.c
drivers/input/serio/hp_sdc.c
drivers/input/tablet/wacom_sys.c
drivers/input/tablet/wacom_wac.c
drivers/isdn/sc/interrupt.c
drivers/macintosh/adb.c
drivers/md/bitmap.c
drivers/md/raid1.c
drivers/media/IR/ir-keytable.c
drivers/media/IR/ir-lirc-codec.c
drivers/media/IR/ir-raw-event.c
drivers/media/IR/ir-sysfs.c
drivers/media/IR/keymaps/rc-rc6-mce.c
drivers/media/IR/mceusb.c
drivers/media/dvb/dvb-usb/dib0700_core.c
drivers/media/dvb/dvb-usb/dib0700_devices.c
drivers/media/dvb/dvb-usb/opera1.c
drivers/media/dvb/frontends/dib7000p.c
drivers/media/dvb/frontends/dib7000p.h
drivers/media/dvb/siano/smscoreapi.c
drivers/media/radio/si470x/radio-si470x-i2c.c
drivers/media/video/cx231xx/Makefile
drivers/media/video/cx231xx/cx231xx-cards.c
drivers/media/video/cx25840/cx25840-core.c
drivers/media/video/cx88/Kconfig
drivers/media/video/gspca/gspca.c
drivers/media/video/gspca/sn9c20x.c
drivers/media/video/ivtv/ivtvfb.c
drivers/media/video/mem2mem_testdev.c
drivers/media/video/mt9m111.c
drivers/media/video/mt9v022.c
drivers/media/video/mx2_camera.c
drivers/media/video/pvrusb2/pvrusb2-ctrl.c
drivers/media/video/s5p-fimc/fimc-core.c
drivers/media/video/saa7134/saa7134-cards.c
drivers/media/video/saa7164/saa7164-buffer.c
drivers/media/video/uvc/uvc_driver.c
drivers/media/video/uvc/uvcvideo.h
drivers/media/video/v4l2-compat-ioctl32.c
drivers/media/video/videobuf-dma-contig.c
drivers/media/video/videobuf-dma-sg.c
drivers/misc/bh1780gli.c
drivers/mmc/core/core.c
drivers/mtd/nand/mxc_nand.c
drivers/net/3c527.c
drivers/net/Kconfig
drivers/net/b44.c
drivers/net/bonding/bond_main.c
drivers/net/ehea/ehea_main.c
drivers/net/ehea/ehea_qmr.h
drivers/net/fec.c
drivers/net/hamradio/6pack.c
drivers/net/hamradio/mkiss.c
drivers/net/irda/sir_dev.c
drivers/net/ppp_async.c
drivers/net/r8169.c
drivers/net/skge.c
drivers/net/tg3.c
drivers/net/tg3.h
drivers/net/wan/cosa.c
drivers/net/wimax/i2400m/rx.c
drivers/net/wireless/ath/ath9k/ani.c
drivers/oprofile/oprof.c
drivers/oprofile/oprof.h
drivers/oprofile/oprofile_files.c
drivers/oprofile/oprofile_perf.c [new file with mode: 0644]
drivers/oprofile/oprofilefs.c
drivers/parport/share.c
drivers/platform/x86/intel_ips.c
drivers/regulator/ad5398.c
drivers/regulator/isl6271a-regulator.c
drivers/rtc/rtc-ds3232.c
drivers/scsi/scsi.c
drivers/serial/ioc3_serial.c
drivers/staging/tm6000/Kconfig
drivers/staging/tm6000/tm6000-input.c
drivers/vhost/net.c
drivers/vhost/vhost.c
drivers/vhost/vhost.h
fs/affs/super.c
fs/binfmt_aout.c
fs/ceph/Kconfig
fs/ceph/Makefile
fs/ceph/README [deleted file]
fs/ceph/addr.c
fs/ceph/caps.c
fs/ceph/ceph_frag.c
fs/ceph/debugfs.c
fs/ceph/dir.c
fs/ceph/export.c
fs/ceph/file.c
fs/ceph/inode.c
fs/ceph/ioctl.c
fs/ceph/ioctl.h
fs/ceph/locks.c
fs/ceph/mds_client.c
fs/ceph/mds_client.h
fs/ceph/mdsmap.c
fs/ceph/pagelist.c [deleted file]
fs/ceph/snap.c
fs/ceph/strings.c [moved from fs/ceph/ceph_strings.c with 59% similarity]
fs/ceph/super.c
fs/ceph/super.h
fs/ceph/xattr.c
fs/exec.c
fs/exofs/inode.c
fs/gfs2/Kconfig
fs/gfs2/aops.c
fs/gfs2/bmap.c
fs/gfs2/bmap.h
fs/gfs2/dentry.c
fs/gfs2/dir.c
fs/gfs2/dir.h
fs/gfs2/export.c
fs/gfs2/file.c
fs/gfs2/glock.c
fs/gfs2/glock.h
fs/gfs2/glops.c
fs/gfs2/incore.h
fs/gfs2/inode.c
fs/gfs2/inode.h
fs/gfs2/lock_dlm.c
fs/gfs2/main.c
fs/gfs2/ops_fstype.c
fs/gfs2/ops_inode.c
fs/gfs2/quota.c
fs/gfs2/recovery.c
fs/gfs2/rgrp.c
fs/gfs2/rgrp.h
fs/gfs2/super.c
fs/gfs2/sys.c
fs/gfs2/trace_gfs2.h
fs/gfs2/trans.h
fs/gfs2/xattr.c
fs/hfs/bfind.c
fs/hfs/btree.c
fs/hfs/btree.h
fs/hfsplus/bfind.c
fs/hfsplus/bitmap.c
fs/hfsplus/brec.c
fs/hfsplus/btree.c
fs/hfsplus/catalog.c
fs/hfsplus/dir.c
fs/hfsplus/extents.c
fs/hfsplus/hfsplus_fs.h
fs/hfsplus/hfsplus_raw.h
fs/hfsplus/inode.c
fs/hfsplus/ioctl.c
fs/hfsplus/options.c
fs/hfsplus/part_tbl.c
fs/hfsplus/super.c
fs/hfsplus/unicode.c
fs/hfsplus/wrapper.c
fs/nfsd/nfsfh.h
fs/notify/Kconfig
fs/xfs/linux-2.6/xfs_sync.c
include/asm-generic/hardirq.h
include/asm-generic/pgtable.h
include/asm-generic/vmlinux.lds.h
include/drm/ttm/ttm_bo_api.h
include/linux/Kbuild
include/linux/acpi_pmtmr.h
include/linux/ceph/auth.h [moved from fs/ceph/auth.h with 97% similarity]
include/linux/ceph/buffer.h [moved from fs/ceph/buffer.h with 100% similarity]
include/linux/ceph/ceph_debug.h [moved from fs/ceph/ceph_debug.h with 86% similarity]
include/linux/ceph/ceph_frag.h [moved from fs/ceph/ceph_frag.h with 100% similarity]
include/linux/ceph/ceph_fs.h [moved from fs/ceph/ceph_fs.h with 99% similarity]
include/linux/ceph/ceph_hash.h [moved from fs/ceph/ceph_hash.h with 100% similarity]
include/linux/ceph/debugfs.h [new file with mode: 0644]
include/linux/ceph/decode.h [moved from fs/ceph/decode.h with 96% similarity]
include/linux/ceph/libceph.h [new file with mode: 0644]
include/linux/ceph/mdsmap.h [moved from fs/ceph/mdsmap.h with 100% similarity]
include/linux/ceph/messenger.h [moved from fs/ceph/messenger.h with 95% similarity]
include/linux/ceph/mon_client.h [moved from fs/ceph/mon_client.h with 99% similarity]
include/linux/ceph/msgpool.h [moved from fs/ceph/msgpool.h with 100% similarity]
include/linux/ceph/msgr.h [moved from fs/ceph/msgr.h with 100% similarity]
include/linux/ceph/osd_client.h [moved from fs/ceph/osd_client.h with 76% similarity]
include/linux/ceph/osdmap.h [moved from fs/ceph/osdmap.h with 97% similarity]
include/linux/ceph/pagelist.h [moved from fs/ceph/pagelist.h with 62% similarity]
include/linux/ceph/rados.h [moved from fs/ceph/rados.h with 100% similarity]
include/linux/ceph/types.h [moved from fs/ceph/types.h with 100% similarity]
include/linux/cgroup.h
include/linux/compiler.h
include/linux/coredump.h
include/linux/cred.h
include/linux/crush/crush.h [moved from fs/ceph/crush/crush.h with 100% similarity]
include/linux/crush/hash.h [moved from fs/ceph/crush/hash.h with 100% similarity]
include/linux/crush/mapper.h [moved from fs/ceph/crush/mapper.h with 100% similarity]
include/linux/debug_locks.h
include/linux/dynamic_debug.h
include/linux/elevator.h
include/linux/fdtable.h
include/linux/fs.h
include/linux/ftrace_event.h
include/linux/genhd.h
include/linux/hardirq.h
include/linux/idr.h
include/linux/init_task.h
include/linux/input.h
include/linux/interrupt.h
include/linux/iocontext.h
include/linux/irq_work.h [new file with mode: 0644]
include/linux/jump_label.h [new file with mode: 0644]
include/linux/jump_label_ref.h [new file with mode: 0644]
include/linux/kernel.h
include/linux/key.h
include/linux/kvm_host.h
include/linux/lockdep.h
include/linux/mm_types.h
include/linux/module.h
include/linux/netfilter/nfnetlink_conntrack.h
include/linux/netfilter/xt_SECMARK.h
include/linux/nfs_fs.h
include/linux/notifier.h
include/linux/oprofile.h
include/linux/pci_ids.h
include/linux/percpu-defs.h
include/linux/percpu.h
include/linux/perf_event.h
include/linux/radix-tree.h
include/linux/rculist.h
include/linux/rculist_nulls.h
include/linux/rcupdate.h
include/linux/rcutiny.h
include/linux/rcutree.h
include/linux/sched.h
include/linux/security.h
include/linux/selinux.h
include/linux/srcu.h
include/linux/stop_machine.h
include/linux/sunrpc/auth_gss.h
include/linux/topology.h
include/linux/tracepoint.h
include/linux/types.h
include/media/videobuf-dma-sg.h
include/net/bluetooth/bluetooth.h
include/net/cls_cgroup.h
include/net/netfilter/nf_conntrack.h
include/trace/events/irq.h
include/trace/events/napi.h
include/trace/events/net.h [new file with mode: 0644]
include/trace/events/power.h
include/trace/events/sched.h
include/trace/events/skb.h
init/Kconfig
kernel/Makefile
kernel/cgroup.c
kernel/cpuset.c
kernel/exit.c
kernel/hrtimer.c
kernel/hung_task.c
kernel/hw_breakpoint.c
kernel/irq_work.c [new file with mode: 0644]
kernel/jump_label.c [new file with mode: 0644]
kernel/kprobes.c
kernel/lockdep.c
kernel/module.c
kernel/perf_event.c
kernel/pid.c
kernel/printk.c
kernel/rcupdate.c
kernel/rcutiny.c
kernel/rcutiny_plugin.h
kernel/rcutorture.c
kernel/rcutree.c
kernel/rcutree.h
kernel/rcutree_plugin.h
kernel/rcutree_trace.c
kernel/sched.c
kernel/sched_fair.c
kernel/sched_features.h
kernel/sched_rt.c
kernel/sched_stoptask.c [new file with mode: 0644]
kernel/signal.c
kernel/softirq.c
kernel/srcu.c
kernel/stop_machine.c
kernel/sysctl.c
kernel/sysctl_check.c
kernel/test_kprobes.c
kernel/timer.c
kernel/trace/Kconfig
kernel/trace/ftrace.c
kernel/trace/ring_buffer.c
kernel/trace/trace.c
kernel/trace/trace.h
kernel/trace/trace_event_perf.c
kernel/trace/trace_events.c
kernel/trace/trace_functions_graph.c
kernel/trace/trace_irqsoff.c
kernel/trace/trace_sched_wakeup.c
kernel/trace/trace_workqueue.c
kernel/tracepoint.c
kernel/watchdog.c
lib/Kconfig.debug
lib/dynamic_debug.c
lib/radix-tree.c
mm/memcontrol.c
mm/memory-failure.c
mm/memory.c
mm/page_alloc.c
mm/vmalloc.c
net/Kconfig
net/Makefile
net/atm/mpc.c
net/bluetooth/l2cap.c
net/bluetooth/rfcomm/sock.c
net/caif/caif_socket.c
net/ceph/Kconfig [new file with mode: 0644]
net/ceph/Makefile [new file with mode: 0644]
net/ceph/armor.c [moved from fs/ceph/armor.c with 100% similarity]
net/ceph/auth.c [moved from fs/ceph/auth.c with 97% similarity]
net/ceph/auth_none.c [moved from fs/ceph/auth_none.c with 96% similarity]
net/ceph/auth_none.h [moved from fs/ceph/auth_none.h with 94% similarity]
net/ceph/auth_x.c [moved from fs/ceph/auth_x.c with 99% similarity]
net/ceph/auth_x.h [moved from fs/ceph/auth_x.h with 96% similarity]
net/ceph/auth_x_protocol.h [moved from fs/ceph/auth_x_protocol.h with 100% similarity]
net/ceph/buffer.c [moved from fs/ceph/buffer.c with 86% similarity]
net/ceph/ceph_common.c [new file with mode: 0644]
net/ceph/ceph_fs.c [moved from fs/ceph/ceph_fs.c with 92% similarity]
net/ceph/ceph_hash.c [moved from fs/ceph/ceph_hash.c with 98% similarity]
net/ceph/ceph_strings.c [new file with mode: 0644]
net/ceph/crush/crush.c [moved from fs/ceph/crush/crush.c with 99% similarity]
net/ceph/crush/hash.c [moved from fs/ceph/crush/hash.c with 99% similarity]
net/ceph/crush/mapper.c [moved from fs/ceph/crush/mapper.c with 99% similarity]
net/ceph/crypto.c [moved from fs/ceph/crypto.c with 99% similarity]
net/ceph/crypto.h [moved from fs/ceph/crypto.h with 95% similarity]
net/ceph/debugfs.c [new file with mode: 0644]
net/ceph/messenger.c [moved from fs/ceph/messenger.c with 89% similarity]
net/ceph/mon_client.c [moved from fs/ceph/mon_client.c with 94% similarity]
net/ceph/msgpool.c [moved from fs/ceph/msgpool.c with 95% similarity]
net/ceph/osd_client.c [moved from fs/ceph/osd_client.c with 84% similarity]
net/ceph/osdmap.c [moved from fs/ceph/osdmap.c with 97% similarity]
net/ceph/pagelist.c [new file with mode: 0644]
net/ceph/pagevec.c [new file with mode: 0644]
net/core/datagram.c
net/core/dev.c
net/core/ethtool.c
net/core/net-traces.c
net/core/skbuff.c
net/core/sock.c
net/core/stream.c
net/ipv4/Kconfig
net/ipv4/igmp.c
net/ipv4/netfilter/nf_conntrack_l3proto_ipv4_compat.c
net/ipv4/netfilter/nf_nat_core.c
net/ipv6/route.c
net/mac80211/agg-tx.c
net/mac80211/status.c
net/netfilter/core.c
net/netfilter/nf_conntrack_ecache.c
net/netfilter/nf_conntrack_extend.c
net/netfilter/nf_conntrack_netlink.c
net/netfilter/nf_conntrack_proto.c
net/netfilter/nf_conntrack_standalone.c
net/netfilter/nf_log.c
net/netfilter/nf_queue.c
net/netfilter/xt_CT.c
net/netfilter/xt_SECMARK.c
net/rds/page.c
net/sched/cls_cgroup.c
net/sched/cls_u32.c
net/sctp/auth.c
net/sctp/socket.c
scripts/Makefile
scripts/Makefile.build
scripts/Makefile.lib
scripts/basic/Makefile
scripts/basic/hash.c [deleted file]
scripts/gcc-goto.sh [new file with mode: 0644]
scripts/kconfig/conf.c
scripts/kconfig/expr.h
scripts/kconfig/menu.c
scripts/kconfig/symbol.c
scripts/recordmcount.c [new file with mode: 0644]
scripts/recordmcount.h [new file with mode: 0644]
security/apparmor/.gitignore
security/apparmor/apparmorfs.c
security/capability.c
security/commoncap.c
security/security.c
security/selinux/Makefile
security/selinux/exports.c
security/selinux/hooks.c
security/selinux/include/classmap.h
security/selinux/include/security.h
security/selinux/selinuxfs.c
security/selinux/ss/Makefile [deleted file]
security/selinux/ss/avtab.c
security/selinux/ss/avtab.h
security/selinux/ss/conditional.c
security/selinux/ss/conditional.h
security/selinux/ss/ebitmap.c
security/selinux/ss/ebitmap.h
security/selinux/ss/policydb.c
security/selinux/ss/policydb.h
security/selinux/ss/services.c
security/selinux/ss/status.c [new file with mode: 0644]
security/smack/smack_lsm.c
security/tomoyo/common.c
sound/core/rawmidi.c
sound/oss/soundcard.c
sound/pci/hda/patch_sigmatel.c
tools/perf/Documentation/perf-annotate.txt
tools/perf/Documentation/perf-report.txt
tools/perf/Makefile
tools/perf/builtin-annotate.c
tools/perf/builtin-report.c
tools/perf/feature-tests.mak
tools/perf/perf.h
tools/perf/scripts/python/bin/netdev-times-record [new file with mode: 0644]
tools/perf/scripts/python/bin/netdev-times-report [new file with mode: 0644]
tools/perf/scripts/python/netdev-times.py [new file with mode: 0644]
tools/perf/util/cache.h
tools/perf/util/callchain.c
tools/perf/util/callchain.h
tools/perf/util/hist.c
tools/perf/util/path.c
tools/perf/util/sort.h
tools/perf/util/symbol.c
tools/perf/util/symbol.h
tools/perf/util/ui/browser.c
tools/perf/util/ui/browser.h
tools/perf/util/ui/browsers/annotate.c
tools/perf/util/ui/browsers/hists.c
tools/perf/util/ui/browsers/map.c
tools/perf/util/ui/util.c
tools/perf/util/util.h

index a0d479d..f66f4df 100644 (file)
@@ -1645,7 +1645,9 @@ the amount of locking which needs to be done.
       all the readers who were traversing the list when we deleted the
       element are finished.  We use <function>call_rcu()</function> to
       register a callback which will actually destroy the object once
-      the readers are finished.
+      all pre-existing readers are finished.  Alternatively,
+      <function>synchronize_rcu()</function> may be used to block until
+      all pre-existing are finished.
     </para>
     <para>
       But how does Read Copy Update know when the readers are
@@ -1714,7 +1716,7 @@ the amount of locking which needs to be done.
 -        object_put(obj);
 +        list_del_rcu(&amp;obj-&gt;list);
          cache_num--;
-+        call_rcu(&amp;obj-&gt;rcu, cache_delete_rcu, obj);
++        call_rcu(&amp;obj-&gt;rcu, cache_delete_rcu);
  }
 
  /* Must be holding cache_lock */
@@ -1725,14 +1727,6 @@ the amount of locking which needs to be done.
          if (++cache_num > MAX_CACHE_SIZE) {
                  struct object *i, *outcast = NULL;
                  list_for_each_entry(i, &amp;cache, list) {
-@@ -85,6 +94,7 @@
-         obj-&gt;popularity = 0;
-         atomic_set(&amp;obj-&gt;refcnt, 1); /* The cache holds a reference */
-         spin_lock_init(&amp;obj-&gt;lock);
-+        INIT_RCU_HEAD(&amp;obj-&gt;rcu);
-
-         spin_lock_irqsave(&amp;cache_lock, flags);
-         __cache_add(obj);
 @@ -104,12 +114,11 @@
  struct object *cache_find(int id)
  {
index 790d1a8..0c134f8 100644 (file)
@@ -218,13 +218,22 @@ over a rather long period of time, but improvements are always welcome!
        include:
 
        a.      Keeping a count of the number of data-structure elements
-               used by the RCU-protected data structure, including those
-               waiting for a grace period to elapse.  Enforce a limit
-               on this number, stalling updates as needed to allow
-               previously deferred frees to complete.
-
-               Alternatively, limit only the number awaiting deferred
-               free rather than the total number of elements.
+               used by the RCU-protected data structure, including
+               those waiting for a grace period to elapse.  Enforce a
+               limit on this number, stalling updates as needed to allow
+               previously deferred frees to complete.  Alternatively,
+               limit only the number awaiting deferred free rather than
+               the total number of elements.
+
+               One way to stall the updates is to acquire the update-side
+               mutex.  (Don't try this with a spinlock -- other CPUs
+               spinning on the lock could prevent the grace period
+               from ever ending.)  Another way to stall the updates
+               is for the updates to use a wrapper function around
+               the memory allocator, so that this wrapper function
+               simulates OOM when there is too much memory awaiting an
+               RCU grace period.  There are of course many other
+               variations on this theme.
 
        b.      Limiting update rate.  For example, if updates occur only
                once per hour, then no explicit rate limiting is required,
@@ -365,3 +374,26 @@ over a rather long period of time, but improvements are always welcome!
        and the compiler to freely reorder code into and out of RCU
        read-side critical sections.  It is the responsibility of the
        RCU update-side primitives to deal with this.
+
+17.    Use CONFIG_PROVE_RCU, CONFIG_DEBUG_OBJECTS_RCU_HEAD, and
+       the __rcu sparse checks to validate your RCU code.  These
+       can help find problems as follows:
+
+       CONFIG_PROVE_RCU: check that accesses to RCU-protected data
+               structures are carried out under the proper RCU
+               read-side critical section, while holding the right
+               combination of locks, or whatever other conditions
+               are appropriate.
+
+       CONFIG_DEBUG_OBJECTS_RCU_HEAD: check that you don't pass the
+               same object to call_rcu() (or friends) before an RCU
+               grace period has elapsed since the last time that you
+               passed that same object to call_rcu() (or friends).
+
+       __rcu sparse checks: tag the pointer to the RCU-protected data
+               structure with __rcu, and sparse will warn you if you
+               access that pointer without the services of one of the
+               variants of rcu_dereference().
+
+       These debugging aids can help you find problems that are
+       otherwise extremely difficult to spot.
index 44c6dcc..862c08e 100644 (file)
@@ -80,6 +80,24 @@ o    A CPU looping with bottom halves disabled.  This condition can
 o      For !CONFIG_PREEMPT kernels, a CPU looping anywhere in the kernel
        without invoking schedule().
 
+o      A CPU-bound real-time task in a CONFIG_PREEMPT kernel, which might
+       happen to preempt a low-priority task in the middle of an RCU
+       read-side critical section.   This is especially damaging if
+       that low-priority task is not permitted to run on any other CPU,
+       in which case the next RCU grace period can never complete, which
+       will eventually cause the system to run out of memory and hang.
+       While the system is in the process of running itself out of
+       memory, you might see stall-warning messages.
+
+o      A CPU-bound real-time task in a CONFIG_PREEMPT_RT kernel that
+       is running at a higher priority than the RCU softirq threads.
+       This will prevent RCU callbacks from ever being invoked,
+       and in a CONFIG_TREE_PREEMPT_RCU kernel will further prevent
+       RCU grace periods from ever completing.  Either way, the
+       system will eventually run out of memory and hang.  In the
+       CONFIG_TREE_PREEMPT_RCU case, you might see stall-warning
+       messages.
+
 o      A bug in the RCU implementation.
 
 o      A hardware failure.  This is quite unlikely, but has occurred
index efd8cc9..a851118 100644 (file)
@@ -125,6 +125,17 @@ o  "b" is the batch limit for this CPU.  If more than this number
        of RCU callbacks is ready to invoke, then the remainder will
        be deferred.
 
+o      "ci" is the number of RCU callbacks that have been invoked for
+       this CPU.  Note that ci+ql is the number of callbacks that have
+       been registered in absence of CPU-hotplug activity.
+
+o      "co" is the number of RCU callbacks that have been orphaned due to
+       this CPU going offline.
+
+o      "ca" is the number of RCU callbacks that have been adopted due to
+       other CPUs going offline.  Note that ci+co-ca+ql is the number of
+       RCU callbacks registered on this CPU.
+
 There is also an rcu/rcudata.csv file with the same information in
 comma-separated-variable spreadsheet format.
 
@@ -180,7 +191,7 @@ o   "s" is the "signaled" state that drives force_quiescent_state()'s
 
 o      "jfq" is the number of jiffies remaining for this grace period
        before force_quiescent_state() is invoked to help push things
-       along.  Note that CPUs in dyntick-idle mode thoughout the grace
+       along.  Note that CPUs in dyntick-idle mode throughout the grace
        period will not report on their own, but rather must be check by
        some other CPU via force_quiescent_state().
 
index f1c5c4b..902d315 100644 (file)
@@ -14,25 +14,39 @@ to /proc/cpuinfo.
        identifier (rather than the kernel's).  The actual value is
        architecture and platform dependent.
 
-3) /sys/devices/system/cpu/cpuX/topology/thread_siblings:
+3) /sys/devices/system/cpu/cpuX/topology/book_id:
+
+       the book ID of cpuX. Typically it is the hardware platform's
+       identifier (rather than the kernel's).  The actual value is
+       architecture and platform dependent.
+
+4) /sys/devices/system/cpu/cpuX/topology/thread_siblings:
 
        internel kernel map of cpuX's hardware threads within the same
        core as cpuX
 
-4) /sys/devices/system/cpu/cpuX/topology/core_siblings:
+5) /sys/devices/system/cpu/cpuX/topology/core_siblings:
 
        internal kernel map of cpuX's hardware threads within the same
        physical_package_id.
 
+6) /sys/devices/system/cpu/cpuX/topology/book_siblings:
+
+       internal kernel map of cpuX's hardware threads within the same
+       book_id.
+
 To implement it in an architecture-neutral way, a new source file,
-drivers/base/topology.c, is to export the 4 attributes.
+drivers/base/topology.c, is to export the 4 or 6 attributes. The two book
+related sysfs files will only be created if CONFIG_SCHED_BOOK is selected.
 
 For an architecture to support this feature, it must define some of
 these macros in include/asm-XXX/topology.h:
 #define topology_physical_package_id(cpu)
 #define topology_core_id(cpu)
+#define topology_book_id(cpu)
 #define topology_thread_cpumask(cpu)
 #define topology_core_cpumask(cpu)
+#define topology_book_cpumask(cpu)
 
 The type of **_id is int.
 The type of siblings is (const) struct cpumask *.
@@ -45,6 +59,9 @@ not defined by include/asm-XXX/topology.h:
 3) thread_siblings: just the given CPU
 4) core_siblings: just the given CPU
 
+For architectures that don't support books (CONFIG_SCHED_BOOK) there are no
+default definitions for topology_book_id() and topology_book_cpumask().
+
 Additionally, CPU topology information is provided under
 /sys/devices/system/cpu and includes these files.  The internal
 source for the output is in brackets ("[]").
index 842aa9d..5e2bc4a 100644 (file)
@@ -386,34 +386,6 @@ Who:       Tejun Heo <tj@kernel.org>
 
 ----------------------------
 
-What:  Support for VMware's guest paravirtuliazation technique [VMI] will be
-       dropped.
-When:  2.6.37 or earlier.
-Why:   With the recent innovations in CPU hardware acceleration technologies
-       from Intel and AMD, VMware ran a few experiments to compare these
-       techniques to guest paravirtualization technique on VMware's platform.
-       These hardware assisted virtualization techniques have outperformed the
-       performance benefits provided by VMI in most of the workloads. VMware
-       expects that these hardware features will be ubiquitous in a couple of
-       years, as a result, VMware has started a phased retirement of this
-       feature from the hypervisor. We will be removing this feature from the
-       Kernel too. Right now we are targeting 2.6.37 but can retire earlier if
-       technical reasons (read opportunity to remove major chunk of pvops)
-       arise.
-
-       Please note that VMI has always been an optimization and non-VMI kernels
-       still work fine on VMware's platform.
-       Latest versions of VMware's product which support VMI are,
-       Workstation 7.0 and VSphere 4.0 on ESX side, future maintainence
-       releases for these products will continue supporting VMI.
-
-       For more details about VMI retirement take a look at this,
-       http://blogs.vmware.com/guestosguide/2009/09/vmi-retirement.html
-
-Who:   Alok N Kataria <akataria@vmware.com>
-
-----------------------------
-
 What:  Support for lcd_switch and display_get in asus-laptop driver
 When:  March 2010
 Why:   These two features use non-standard interfaces. There are the
index 8dd7248..3a0009e 100644 (file)
@@ -455,7 +455,7 @@ and is between 256 and 4096 characters. It is defined in the file
                        [ARM] imx_timer1,OSTS,netx_timer,mpu_timer2,
                                pxa_timer,timer3,32k_counter,timer0_1
                        [AVR32] avr32
-                       [X86-32] pit,hpet,tsc,vmi-timer;
+                       [X86-32] pit,hpet,tsc;
                                scx200_hrt on Geode; cyclone on IBM x440
                        [MIPS] MIPS
                        [PARISC] cr16
@@ -2153,6 +2153,11 @@ and is between 256 and 4096 characters. It is defined in the file
                        Reserves a hole at the top of the kernel virtual
                        address space.
 
+       reservelow=     [X86]
+                       Format: nn[K]
+                       Set the amount of memory to reserve for BIOS at
+                       the bottom of the address space.
+
        reset_devices   [KNL] Force drivers to reset the underlying device
                        during initialization.
 
@@ -2435,6 +2440,10 @@ and is between 256 and 4096 characters. It is defined in the file
                        disables clocksource verification at runtime.
                        Used to enable high-resolution timer mode on older
                        hardware, and in virtualized environment.
+                       [x86] noirqtime: Do not use TSC to do irq accounting.
+                       Used to run time disable IRQ_TIME_ACCOUNTING on any
+                       platforms where RDTSC is slow and this accounting
+                       can add overhead.
 
        turbografx.map[2|3]=    [HW,JOY]
                        TurboGraFX parallel port interface
index 1762b81..741fe66 100644 (file)
@@ -542,9 +542,11 @@ Kprobes does not use mutexes or allocate memory except during
 registration and unregistration.
 
 Probe handlers are run with preemption disabled.  Depending on the
-architecture, handlers may also run with interrupts disabled.  In any
-case, your handler should not yield the CPU (e.g., by attempting to
-acquire a semaphore).
+architecture and optimization state, handlers may also run with
+interrupts disabled (e.g., kretprobe handlers and optimized kprobe
+handlers run without interrupt disabled on x86/x86-64).  In any case,
+your handler should not yield the CPU (e.g., by attempting to acquire
+a semaphore).
 
 Since a return probe is implemented by replacing the return
 address with the trampoline's address, stack backtraces and calls
index 2df7186..d9271e7 100644 (file)
@@ -1,82 +1,35 @@
 Linux* Base Driver for the Intel(R) PRO/1000 Family of Adapters
 ===============================================================
 
-September 26, 2006
-
+Intel Gigabit Linux driver.
+Copyright(c) 1999 - 2010 Intel Corporation.
 
 Contents
 ========
 
-- In This Release
 - Identifying Your Adapter
-- Building and Installation
 - Command Line Parameters
 - Speed and Duplex Configuration
 - Additional Configurations
-- Known Issues
 - Support
 
-
-In This Release
-===============
-
-This file describes the Linux* Base Driver for the Intel(R) PRO/1000 Family
-of Adapters.  This driver includes support for Itanium(R)2-based systems.
-
-For questions related to hardware requirements, refer to the documentation
-supplied with your Intel PRO/1000 adapter. All hardware requirements listed
-apply to use with Linux.
-
-The following features are now available in supported kernels:
- - Native VLANs
- - Channel Bonding (teaming)
- - SNMP
-
-Channel Bonding documentation can be found in the Linux kernel source:
-/Documentation/networking/bonding.txt
-
-The driver information previously displayed in the /proc filesystem is not
-supported in this release.  Alternatively, you can use ethtool (version 1.6
-or later), lspci, and ifconfig to obtain the same information.
-
-Instructions on updating ethtool can be found in the section "Additional
-Configurations" later in this document.
-
-NOTE: The Intel(R) 82562v 10/100 Network Connection only provides 10/100
-support.
-
-
 Identifying Your Adapter
 ========================
 
 For more information on how to identify your adapter, go to the Adapter &
 Driver ID Guide at:
 
-    http://support.intel.com/support/network/adapter/pro100/21397.htm
+    http://support.intel.com/support/go/network/adapter/idguide.htm
 
 For the latest Intel network drivers for Linux, refer to the following
 website.  In the search field, enter your adapter name or type, or use the
 networking link on the left to search for your adapter:
 
-    http://downloadfinder.intel.com/scripts-df/support_intel.asp
-
+    http://support.intel.com/support/go/network/adapter/home.htm
 
 Command Line Parameters
 =======================
 
-If the driver is built as a module, the  following optional parameters
-are used by entering them on the command line with the modprobe command
-using this syntax:
-
-     modprobe e1000 [<option>=<VAL1>,<VAL2>,...]
-
-For example, with two PRO/1000 PCI adapters, entering:
-
-     modprobe e1000 TxDescriptors=80,128
-
-loads the e1000 driver with 80 TX descriptors for the first adapter and
-128 TX descriptors for the second adapter.
-
 The default value for each parameter is generally the recommended setting,
 unless otherwise noted.
 
@@ -89,10 +42,6 @@ NOTES:  For more information about the AutoNeg, Duplex, and Speed
         parameters, see the application note at:
         http://www.intel.com/design/network/applnots/ap450.htm
 
-        A descriptor describes a data buffer and attributes related to
-        the data buffer.  This information is accessed by the hardware.
-
-
 AutoNeg
 -------
 (Supported only on adapters with copper connections)
@@ -106,7 +55,6 @@ Duplex parameters must not be specified.
 NOTE:  Refer to the Speed and Duplex section of this readme for more
        information on the AutoNeg parameter.
 
-
 Duplex
 ------
 (Supported only on adapters with copper connections)
@@ -119,7 +67,6 @@ set to auto-negotiate, the board auto-detects the correct duplex.  If the
 link partner is forced (either full or half), Duplex defaults to half-
 duplex.
 
-
 FlowControl
 -----------
 Valid Range:   0-3 (0=none, 1=Rx only, 2=Tx only, 3=Rx&Tx)
@@ -128,16 +75,16 @@ Default Value: Reads flow control settings from the EEPROM
 This parameter controls the automatic generation(Tx) and response(Rx)
 to Ethernet PAUSE frames.
 
-
 InterruptThrottleRate
 ---------------------
 (not supported on Intel(R) 82542, 82543 or 82544-based adapters)
-Valid Range:   0,1,3,100-100000 (0=off, 1=dynamic, 3=dynamic conservative)
+Valid Range:   0,1,3,4,100-100000 (0=off, 1=dynamic, 3=dynamic conservative,
+                                   4=simplified balancing)
 Default Value: 3
 
 The driver can limit the amount of interrupts per second that the adapter
-will generate for incoming packets. It does this by writing a value to the 
-adapter that is based on the maximum amount of interrupts that the adapter 
+will generate for incoming packets. It does this by writing a value to the
+adapter that is based on the maximum amount of interrupts that the adapter
 will generate per second.
 
 Setting InterruptThrottleRate to a value greater or equal to 100
@@ -146,37 +93,43 @@ per second, even if more packets have come in. This reduces interrupt
 load on the system and can lower CPU utilization under heavy load,
 but will increase latency as packets are not processed as quickly.
 
-The default behaviour of the driver previously assumed a static 
-InterruptThrottleRate value of 8000, providing a good fallback value for 
-all traffic types,but lacking in small packet performance and latency. 
-The hardware can handle many more small packets per second however, and 
+The default behaviour of the driver previously assumed a static
+InterruptThrottleRate value of 8000, providing a good fallback value for
+all traffic types,but lacking in small packet performance and latency.
+The hardware can handle many more small packets per second however, and
 for this reason an adaptive interrupt moderation algorithm was implemented.
 
 Since 7.3.x, the driver has two adaptive modes (setting 1 or 3) in which
-it dynamically adjusts the InterruptThrottleRate value based on the traffic 
+it dynamically adjusts the InterruptThrottleRate value based on the traffic
 that it receives. After determining the type of incoming traffic in the last
-timeframe, it will adjust the InterruptThrottleRate to an appropriate value 
+timeframe, it will adjust the InterruptThrottleRate to an appropriate value
 for that traffic.
 
 The algorithm classifies the incoming traffic every interval into
-classes.  Once the class is determined, the InterruptThrottleRate value is 
-adjusted to suit that traffic type the best. There are three classes defined: 
+classes.  Once the class is determined, the InterruptThrottleRate value is
+adjusted to suit that traffic type the best. There are three classes defined:
 "Bulk traffic", for large amounts of packets of normal size; "Low latency",
 for small amounts of traffic and/or a significant percentage of small
-packets; and "Lowest latency", for almost completely small packets or 
+packets; and "Lowest latency", for almost completely small packets or
 minimal traffic.
 
-In dynamic conservative mode, the InterruptThrottleRate value is set to 4000 
-for traffic that falls in class "Bulk traffic". If traffic falls in the "Low 
-latency" or "Lowest latency" class, the InterruptThrottleRate is increased 
+In dynamic conservative mode, the InterruptThrottleRate value is set to 4000
+for traffic that falls in class "Bulk traffic". If traffic falls in the "Low
+latency" or "Lowest latency" class, the InterruptThrottleRate is increased
 stepwise to 20000. This default mode is suitable for most applications.
 
 For situations where low latency is vital such as cluster or
 grid computing, the algorithm can reduce latency even more when
 InterruptThrottleRate is set to mode 1. In this mode, which operates
-the same as mode 3, the InterruptThrottleRate will be increased stepwise to 
+the same as mode 3, the InterruptThrottleRate will be increased stepwise to
 70000 for traffic in class "Lowest latency".
 
+In simplified mode the interrupt rate is based on the ratio of Tx and
+Rx traffic.  If the bytes per second rate is approximately equal, the
+interrupt rate will drop as low as 2000 interrupts per second.  If the
+traffic is mostly transmit or mostly receive, the interrupt rate could
+be as high as 8000.
+
 Setting InterruptThrottleRate to 0 turns off any interrupt moderation
 and may improve small packet latency, but is generally not suitable
 for bulk throughput traffic.
@@ -212,8 +165,6 @@ NOTE:  When e1000 is loaded with default settings and multiple adapters
        be platform-specific.  If CPU utilization is not a concern, use
        RX_POLLING (NAPI) and default driver settings.
 
-
-
 RxDescriptors
 -------------
 Valid Range:   80-256 for 82542 and 82543-based adapters
@@ -225,15 +176,14 @@ by the driver.  Increasing this value allows the driver to buffer more
 incoming packets, at the expense of increased system memory utilization.
 
 Each descriptor is 16 bytes.  A receive buffer is also allocated for each
-descriptor and can be either 2048, 4096, 8192, or 16384 bytes, depending 
+descriptor and can be either 2048, 4096, 8192, or 16384 bytes, depending
 on the MTU setting. The maximum MTU size is 16110.
 
-NOTE:  MTU designates the frame size.  It only needs to be set for Jumbo 
-       Frames.  Depending on the available system resources, the request 
-       for a higher number of receive descriptors may be denied.  In this 
+NOTE:  MTU designates the frame size.  It only needs to be set for Jumbo
+       Frames.  Depending on the available system resources, the request
+       for a higher number of receive descriptors may be denied.  In this
        case, use a lower number.
 
-
 RxIntDelay
 ----------
 Valid Range:   0-65535 (0=off)
@@ -254,7 +204,6 @@ CAUTION:  When setting RxIntDelay to a value other than 0, adapters may
           restoring the network connection.  To eliminate the potential
           for the hang ensure that RxIntDelay is set to 0.
 
-
 RxAbsIntDelay
 -------------
 (This parameter is supported only on 82540, 82545 and later adapters.)
@@ -268,7 +217,6 @@ packet is received within the set amount of time.  Proper tuning,
 along with RxIntDelay, may improve traffic throughput in specific network
 conditions.
 
-
 Speed
 -----
 (This parameter is supported only on adapters with copper connections.)
@@ -280,7 +228,6 @@ Speed forces the line speed to the specified value in megabits per second
 partner is set to auto-negotiate, the board will auto-detect the correct
 speed.  Duplex should also be set when Speed is set to either 10 or 100.
 
-
 TxDescriptors
 -------------
 Valid Range:   80-256 for 82542 and 82543-based adapters
@@ -295,6 +242,36 @@ NOTE:  Depending on the available system resources, the request for a
        higher number of transmit descriptors may be denied.  In this case,
        use a lower number.
 
+TxDescriptorStep
+----------------
+Valid Range:    1 (use every Tx Descriptor)
+               4 (use every 4th Tx Descriptor)
+
+Default Value:  1 (use every Tx Descriptor)
+
+On certain non-Intel architectures, it has been observed that intense TX
+traffic bursts of short packets may result in an improper descriptor
+writeback. If this occurs, the driver will report a "TX Timeout" and reset
+the adapter, after which the transmit flow will restart, though data may
+have stalled for as much as 10 seconds before it resumes.
+
+The improper writeback does not occur on the first descriptor in a system
+memory cache-line, which is typically 32 bytes, or 4 descriptors long.
+
+Setting TxDescriptorStep to a value of 4 will ensure that all TX descriptors
+are aligned to the start of a system memory cache line, and so this problem
+will not occur.
+
+NOTES: Setting TxDescriptorStep to 4 effectively reduces the number of
+       TxDescriptors available for transmits to 1/4 of the normal allocation.
+       This has a possible negative performance impact, which may be
+       compensated for by allocating more descriptors using the TxDescriptors
+       module parameter.
+
+       There are other conditions which may result in "TX Timeout", which will
+       not be resolved by the use of the TxDescriptorStep parameter. As the
+       issue addressed by this parameter has never been observed on Intel
+       Architecture platforms, it should not be used on Intel platforms.
 
 TxIntDelay
 ----------
@@ -307,7 +284,6 @@ efficiency if properly tuned for specific network traffic.  If the
 system is reporting dropped transmits, this value may be set too high
 causing the driver to run out of available transmit descriptors.
 
-
 TxAbsIntDelay
 -------------
 (This parameter is supported only on 82540, 82545 and later adapters.)
@@ -330,6 +306,35 @@ Default Value: 1
 A value of '1' indicates that the driver should enable IP checksum
 offload for received packets (both UDP and TCP) to the adapter hardware.
 
+Copybreak
+---------
+Valid Range:   0-xxxxxxx (0=off)
+Default Value: 256
+Usage: insmod e1000.ko copybreak=128
+
+Driver copies all packets below or equaling this size to a fresh Rx
+buffer before handing it up the stack.
+
+This parameter is different than other parameters, in that it is a
+single (not 1,1,1 etc.) parameter applied to all driver instances and
+it is also available during runtime at
+/sys/module/e1000/parameters/copybreak
+
+SmartPowerDownEnable
+--------------------
+Valid Range: 0-1
+Default Value:  0 (disabled)
+
+Allows PHY to turn off in lower power states. The user can turn off
+this parameter in supported chipsets.
+
+KumeranLockLoss
+---------------
+Valid Range: 0-1
+Default Value: 1 (enabled)
+
+This workaround skips resetting the PHY at shutdown for the initial
+silicon releases of ICH8 systems.
 
 Speed and Duplex Configuration
 ==============================
@@ -385,40 +390,9 @@ If the link partner is forced to a specific speed and duplex, then this
 parameter should not be used.  Instead, use the Speed and Duplex parameters
 previously mentioned to force the adapter to the same speed and duplex.
 
-
 Additional Configurations
 =========================
 
-  Configuring the Driver on Different Distributions
-  -------------------------------------------------
-  Configuring a network driver to load properly when the system is started
-  is distribution dependent.  Typically, the configuration process involves
-  adding an alias line to /etc/modules.conf or /etc/modprobe.conf as well
-  as editing other system startup scripts and/or configuration files.  Many
-  popular Linux distributions ship with tools to make these changes for you.
-  To learn the proper way to configure a network device for your system,
-  refer to your distribution documentation.  If during this process you are
-  asked for the driver or module name, the name for the Linux Base Driver
-  for the Intel(R) PRO/1000 Family of Adapters is e1000.
-
-  As an example, if you install the e1000 driver for two PRO/1000 adapters
-  (eth0 and eth1) and set the speed and duplex to 10full and 100half, add
-  the following to modules.conf or or modprobe.conf:
-
-       alias eth0 e1000
-       alias eth1 e1000
-       options e1000 Speed=10,100 Duplex=2,1
-
-  Viewing Link Messages
-  ---------------------
-  Link messages will not be displayed to the console if the distribution is
-  restricting system messages.  In order to see network driver link messages
-  on your console, set dmesg to eight by entering the following:
-
-       dmesg -n 8
-
-  NOTE: This setting is not saved across reboots.
-
   Jumbo Frames
   ------------
   Jumbo Frames support is enabled by changing the MTU to a value larger than
@@ -437,9 +411,11 @@ Additional Configurations
    setting in a different location.
 
   Notes:
-
-  - To enable Jumbo Frames, increase the MTU size on the interface beyond
-    1500.
+  Degradation in throughput performance may be observed in some Jumbo frames
+  environments. If this is observed, increasing the application's socket buffer
+  size and/or increasing the /proc/sys/net/ipv4/tcp_*mem entry values may help.
+  See the specific application manual and /usr/src/linux*/Documentation/
+  networking/ip-sysctl.txt for more details.
 
   - The maximum MTU setting for Jumbo Frames is 16110.  This value coincides
     with the maximum Jumbo Frames size of 16128.
@@ -447,40 +423,11 @@ Additional Configurations
   - Using Jumbo Frames at 10 or 100 Mbps may result in poor performance or
     loss of link.
 
-  - Some Intel gigabit adapters that support Jumbo Frames have a frame size
-    limit of 9238 bytes, with a corresponding MTU size limit of 9216 bytes.
-    The adapters with this limitation are based on the Intel(R) 82571EB,
-    82572EI, 82573L and 80003ES2LAN controller.  These correspond to the
-    following product names:
-     Intel(R) PRO/1000 PT Server Adapter
-     Intel(R) PRO/1000 PT Desktop Adapter
-     Intel(R) PRO/1000 PT Network Connection
-     Intel(R) PRO/1000 PT Dual Port Server Adapter
-     Intel(R) PRO/1000 PT Dual Port Network Connection
-     Intel(R) PRO/1000 PF Server Adapter
-     Intel(R) PRO/1000 PF Network Connection
-     Intel(R) PRO/1000 PF Dual Port Server Adapter
-     Intel(R) PRO/1000 PB Server Connection
-     Intel(R) PRO/1000 PL Network Connection
-     Intel(R) PRO/1000 EB Network Connection with I/O Acceleration
-     Intel(R) PRO/1000 EB Backplane Connection with I/O Acceleration
-     Intel(R) PRO/1000 PT Quad Port Server Adapter
-
   - Adapters based on the Intel(R) 82542 and 82573V/E controller do not
     support Jumbo Frames. These correspond to the following product names:
      Intel(R) PRO/1000 Gigabit Server Adapter
      Intel(R) PRO/1000 PM Network Connection
 
-  - The following adapters do not support Jumbo Frames:
-     Intel(R) 82562V 10/100 Network Connection
-     Intel(R) 82566DM Gigabit Network Connection
-     Intel(R) 82566DC Gigabit Network Connection
-     Intel(R) 82566MM Gigabit Network Connection
-     Intel(R) 82566MC Gigabit Network Connection
-     Intel(R) 82562GT 10/100 Network Connection
-     Intel(R) 82562G 10/100 Network Connection
-
-
   Ethtool
   -------
   The driver utilizes the ethtool interface for driver configuration and
@@ -490,142 +437,14 @@ Additional Configurations
   The latest release of ethtool can be found from
   http://sourceforge.net/projects/gkernel.
 
-  NOTE: Ethtool 1.6 only supports a limited set of ethtool options.  Support
-  for a more complete ethtool feature set can be enabled by upgrading
-  ethtool to ethtool-1.8.1.
-
   Enabling Wake on LAN* (WoL)
   ---------------------------
-  WoL is configured through the Ethtool* utility.  Ethtool is included with
-  all versions of Red Hat after Red Hat 7.2.  For other Linux distributions,
-  download and install Ethtool from the following website:
-  http://sourceforge.net/projects/gkernel.
-
-  For instructions on enabling WoL with Ethtool, refer to the website listed
-  above.
+  WoL is configured through the Ethtool* utility.
 
   WoL will be enabled on the system during the next shut down or reboot.
   For this driver version, in order to enable WoL, the e1000 driver must be
   loaded when shutting down or rebooting the system.
 
-  Wake On LAN is only supported on port A for the following devices:
-  Intel(R) PRO/1000 PT Dual Port Network Connection
-  Intel(R) PRO/1000 PT Dual Port Server Connection
-  Intel(R) PRO/1000 PT Dual Port Server Adapter
-  Intel(R) PRO/1000 PF Dual Port Server Adapter
-  Intel(R) PRO/1000 PT Quad Port Server Adapter
-
-  NAPI
-  ----
-  NAPI (Rx polling mode) is enabled in the e1000 driver.
-
-  See www.cyberus.ca/~hadi/usenix-paper.tgz for more information on NAPI.
-
-
-Known Issues
-============
-
-Dropped Receive Packets on Half-duplex 10/100 Networks
-------------------------------------------------------
-If you have an Intel PCI Express adapter running at 10mbps or 100mbps, half-
-duplex, you may observe occasional dropped receive packets.  There are no
-workarounds for this problem in this network configuration.  The network must
-be updated to operate in full-duplex, and/or 1000mbps only.
-
-Jumbo Frames System Requirement
--------------------------------
-Memory allocation failures have been observed on Linux systems with 64 MB
-of RAM or less that are running Jumbo Frames.  If you are using Jumbo
-Frames, your system may require more than the advertised minimum
-requirement of 64 MB of system memory.
-
-Performance Degradation with Jumbo Frames
------------------------------------------
-Degradation in throughput performance may be observed in some Jumbo frames
-environments.  If this is observed, increasing the application's socket
-buffer size and/or increasing the /proc/sys/net/ipv4/tcp_*mem entry values
-may help.  See the specific application manual and
-/usr/src/linux*/Documentation/
-networking/ip-sysctl.txt for more details.
-
-Jumbo Frames on Foundry BigIron 8000 switch
--------------------------------------------
-There is a known issue using Jumbo frames when connected to a Foundry
-BigIron 8000 switch.  This is a 3rd party limitation.  If you experience
-loss of packets, lower the MTU size.
-
-Allocating Rx Buffers when Using Jumbo Frames 
----------------------------------------------
-Allocating Rx buffers when using Jumbo Frames on 2.6.x kernels may fail if 
-the available memory is heavily fragmented. This issue may be seen with PCI-X 
-adapters or with packet split disabled. This can be reduced or eliminated 
-by changing the amount of available memory for receive buffer allocation, by
-increasing /proc/sys/vm/min_free_kbytes. 
-
-Multiple Interfaces on Same Ethernet Broadcast Network
-------------------------------------------------------
-Due to the default ARP behavior on Linux, it is not possible to have
-one system on two IP networks in the same Ethernet broadcast domain
-(non-partitioned switch) behave as expected.  All Ethernet interfaces
-will respond to IP traffic for any IP address assigned to the system.
-This results in unbalanced receive traffic.
-
-If you have multiple interfaces in a server, either turn on ARP
-filtering by entering:
-
-    echo 1 > /proc/sys/net/ipv4/conf/all/arp_filter
-(this only works if your kernel's version is higher than 2.4.5),
-
-NOTE: This setting is not saved across reboots.  The configuration
-change can be made permanent by adding the line:
-    net.ipv4.conf.all.arp_filter = 1
-to the file /etc/sysctl.conf
-
-      or,
-
-install the interfaces in separate broadcast domains (either in
-different switches or in a switch partitioned to VLANs).
-
-82541/82547 can't link or are slow to link with some link partners
------------------------------------------------------------------
-There is a known compatibility issue with 82541/82547 and some
-low-end switches where the link will not be established, or will
-be slow to establish.  In particular, these switches are known to
-be incompatible with 82541/82547:
-
-    Planex FXG-08TE
-    I-O Data ETG-SH8
-
-To workaround this issue, the driver can be compiled with an override
-of the PHY's master/slave setting.  Forcing master or forcing slave
-mode will improve time-to-link.
-
-    # make CFLAGS_EXTRA=-DE1000_MASTER_SLAVE=<n>
-
-Where <n> is:
-
-    0 = Hardware default
-    1 = Master mode
-    2 = Slave mode
-    3 = Auto master/slave
-
-Disable rx flow control with ethtool
-------------------------------------
-In order to disable receive flow control using ethtool, you must turn
-off auto-negotiation on the same command line.
-
-For example:
-
-   ethtool -A eth? autoneg off rx off
-
-Unplugging network cable while ethtool -p is running
-----------------------------------------------------
-In kernel versions 2.5.50 and later (including 2.6 kernel), unplugging
-the network cable while ethtool -p is running will cause the system to
-become unresponsive to keyboard commands, except for control-alt-delete.
-Restarting the system appears to be the only remedy.
-
-
 Support
 =======
 
diff --git a/Documentation/networking/e1000e.txt b/Documentation/networking/e1000e.txt
new file mode 100644 (file)
index 0000000..6aa048b
--- /dev/null
@@ -0,0 +1,302 @@
+Linux* Driver for Intel(R) Network Connection
+===============================================================
+
+Intel Gigabit Linux driver.
+Copyright(c) 1999 - 2010 Intel Corporation.
+
+Contents
+========
+
+- Identifying Your Adapter
+- Command Line Parameters
+- Additional Configurations
+- Support
+
+Identifying Your Adapter
+========================
+
+The e1000e driver supports all PCI Express Intel(R) Gigabit Network
+Connections, except those that are 82575, 82576 and 82580-based*.
+
+* NOTE: The Intel(R) PRO/1000 P Dual Port Server Adapter is supported by
+  the e1000 driver, not the e1000e driver due to the 82546 part being used
+  behind a PCI Express bridge.
+
+For more information on how to identify your adapter, go to the Adapter &
+Driver ID Guide at:
+
+    http://support.intel.com/support/go/network/adapter/idguide.htm
+
+For the latest Intel network drivers for Linux, refer to the following
+website.  In the search field, enter your adapter name or type, or use the
+networking link on the left to search for your adapter:
+
+    http://support.intel.com/support/go/network/adapter/home.htm
+
+Command Line Parameters
+=======================
+
+The default value for each parameter is generally the recommended setting,
+unless otherwise noted.
+
+NOTES:  For more information about the InterruptThrottleRate,
+        RxIntDelay, TxIntDelay, RxAbsIntDelay, and TxAbsIntDelay
+        parameters, see the application note at:
+        http://www.intel.com/design/network/applnots/ap450.htm
+
+InterruptThrottleRate
+---------------------
+Valid Range:   0,1,3,4,100-100000 (0=off, 1=dynamic, 3=dynamic conservative,
+                                   4=simplified balancing)
+Default Value: 3
+
+The driver can limit the amount of interrupts per second that the adapter
+will generate for incoming packets. It does this by writing a value to the
+adapter that is based on the maximum amount of interrupts that the adapter
+will generate per second.
+
+Setting InterruptThrottleRate to a value greater or equal to 100
+will program the adapter to send out a maximum of that many interrupts
+per second, even if more packets have come in. This reduces interrupt
+load on the system and can lower CPU utilization under heavy load,
+but will increase latency as packets are not processed as quickly.
+
+The driver has two adaptive modes (setting 1 or 3) in which
+it dynamically adjusts the InterruptThrottleRate value based on the traffic
+that it receives. After determining the type of incoming traffic in the last
+timeframe, it will adjust the InterruptThrottleRate to an appropriate value
+for that traffic.
+
+The algorithm classifies the incoming traffic every interval into
+classes.  Once the class is determined, the InterruptThrottleRate value is
+adjusted to suit that traffic type the best. There are three classes defined:
+"Bulk traffic", for large amounts of packets of normal size; "Low latency",
+for small amounts of traffic and/or a significant percentage of small
+packets; and "Lowest latency", for almost completely small packets or
+minimal traffic.
+
+In dynamic conservative mode, the InterruptThrottleRate value is set to 4000
+for traffic that falls in class "Bulk traffic". If traffic falls in the "Low
+latency" or "Lowest latency" class, the InterruptThrottleRate is increased
+stepwise to 20000. This default mode is suitable for most applications.
+
+For situations where low latency is vital such as cluster or
+grid computing, the algorithm can reduce latency even more when
+InterruptThrottleRate is set to mode 1. In this mode, which operates
+the same as mode 3, the InterruptThrottleRate will be increased stepwise to
+70000 for traffic in class "Lowest latency".
+
+In simplified mode the interrupt rate is based on the ratio of Tx and
+Rx traffic.  If the bytes per second rate is approximately equal the
+interrupt rate will drop as low as 2000 interrupts per second.  If the
+traffic is mostly transmit or mostly receive, the interrupt rate could
+be as high as 8000.
+
+Setting InterruptThrottleRate to 0 turns off any interrupt moderation
+and may improve small packet latency, but is generally not suitable
+for bulk throughput traffic.
+
+NOTE:  InterruptThrottleRate takes precedence over the TxAbsIntDelay and
+       RxAbsIntDelay parameters.  In other words, minimizing the receive
+       and/or transmit absolute delays does not force the controller to
+       generate more interrupts than what the Interrupt Throttle Rate
+       allows.
+
+NOTE:  When e1000e is loaded with default settings and multiple adapters
+       are in use simultaneously, the CPU utilization may increase non-
+       linearly.  In order to limit the CPU utilization without impacting
+       the overall throughput, we recommend that you load the driver as
+       follows:
+
+           modprobe e1000e InterruptThrottleRate=3000,3000,3000
+
+       This sets the InterruptThrottleRate to 3000 interrupts/sec for
+       the first, second, and third instances of the driver.  The range
+       of 2000 to 3000 interrupts per second works on a majority of
+       systems and is a good starting point, but the optimal value will
+       be platform-specific.  If CPU utilization is not a concern, use
+       RX_POLLING (NAPI) and default driver settings.
+
+RxIntDelay
+----------
+Valid Range:   0-65535 (0=off)
+Default Value: 0
+
+This value delays the generation of receive interrupts in units of 1.024
+microseconds.  Receive interrupt reduction can improve CPU efficiency if
+properly tuned for specific network traffic.  Increasing this value adds
+extra latency to frame reception and can end up decreasing the throughput
+of TCP traffic.  If the system is reporting dropped receives, this value
+may be set too high, causing the driver to run out of available receive
+descriptors.
+
+CAUTION:  When setting RxIntDelay to a value other than 0, adapters may
+          hang (stop transmitting) under certain network conditions.  If
+          this occurs a NETDEV WATCHDOG message is logged in the system
+          event log.  In addition, the controller is automatically reset,
+          restoring the network connection.  To eliminate the potential
+          for the hang ensure that RxIntDelay is set to 0.
+
+RxAbsIntDelay
+-------------
+Valid Range:   0-65535 (0=off)
+Default Value: 8
+
+This value, in units of 1.024 microseconds, limits the delay in which a
+receive interrupt is generated.  Useful only if RxIntDelay is non-zero,
+this value ensures that an interrupt is generated after the initial
+packet is received within the set amount of time.  Proper tuning,
+along with RxIntDelay, may improve traffic throughput in specific network
+conditions.
+
+TxIntDelay
+----------
+Valid Range:   0-65535 (0=off)
+Default Value: 8
+
+This value delays the generation of transmit interrupts in units of
+1.024 microseconds.  Transmit interrupt reduction can improve CPU
+efficiency if properly tuned for specific network traffic.  If the
+system is reporting dropped transmits, this value may be set too high
+causing the driver to run out of available transmit descriptors.
+
+TxAbsIntDelay
+-------------
+Valid Range:   0-65535 (0=off)
+Default Value: 32
+
+This value, in units of 1.024 microseconds, limits the delay in which a
+transmit interrupt is generated.  Useful only if TxIntDelay is non-zero,
+this value ensures that an interrupt is generated after the initial
+packet is sent on the wire within the set amount of time.  Proper tuning,
+along with TxIntDelay, may improve traffic throughput in specific
+network conditions.
+
+Copybreak
+---------
+Valid Range:   0-xxxxxxx (0=off)
+Default Value: 256
+
+Driver copies all packets below or equaling this size to a fresh Rx
+buffer before handing it up the stack.
+
+This parameter is different than other parameters, in that it is a
+single (not 1,1,1 etc.) parameter applied to all driver instances and
+it is also available during runtime at
+/sys/module/e1000e/parameters/copybreak
+
+SmartPowerDownEnable
+--------------------
+Valid Range: 0-1
+Default Value:  0 (disabled)
+
+Allows PHY to turn off in lower power states. The user can set this parameter
+in supported chipsets.
+
+KumeranLockLoss
+---------------
+Valid Range: 0-1
+Default Value: 1 (enabled)
+
+This workaround skips resetting the PHY at shutdown for the initial
+silicon releases of ICH8 systems.
+
+IntMode
+-------
+Valid Range: 0-2 (0=legacy, 1=MSI, 2=MSI-X)
+Default Value: 2
+
+Allows changing the interrupt mode at module load time, without requiring a
+recompile. If the driver load fails to enable a specific interrupt mode, the
+driver will try other interrupt modes, from least to most compatible.  The
+interrupt order is MSI-X, MSI, Legacy.  If specifying MSI (IntMode=1)
+interrupts, only MSI and Legacy will be attempted.
+
+CrcStripping
+------------
+Valid Range: 0-1
+Default Value: 1 (enabled)
+
+Strip the CRC from received packets before sending up the network stack.  If
+you have a machine with a BMC enabled but cannot receive IPMI traffic after
+loading or enabling the driver, try disabling this feature.
+
+WriteProtectNVM
+---------------
+Valid Range: 0-1
+Default Value: 1 (enabled)
+
+Set the hardware to ignore all write/erase cycles to the GbE region in the
+ICHx NVM (non-volatile memory).  This feature can be disabled by the
+WriteProtectNVM module parameter (enabled by default) only after a hardware
+reset, but the machine must be power cycled before trying to enable writes.
+
+Note: the kernel boot option iomem=relaxed may need to be set if the kernel
+config option CONFIG_STRICT_DEVMEM=y, if the root user wants to write the
+NVM from user space via ethtool.
+
+Additional Configurations
+=========================
+
+  Jumbo Frames
+  ------------
+  Jumbo Frames support is enabled by changing the MTU to a value larger than
+  the default of 1500.  Use the ifconfig command to increase the MTU size.
+  For example:
+
+       ifconfig eth<x> mtu 9000 up
+
+  This setting is not saved across reboots.
+
+  Notes:
+
+  - The maximum MTU setting for Jumbo Frames is 9216.  This value coincides
+    with the maximum Jumbo Frames size of 9234 bytes.
+
+  - Using Jumbo Frames at 10 or 100 Mbps is not supported and may result in
+    poor performance or loss of link.
+
+  - Some adapters limit Jumbo Frames sized packets to a maximum of
+    4096 bytes and some adapters do not support Jumbo Frames.
+
+
+  Ethtool
+  -------
+  The driver utilizes the ethtool interface for driver configuration and
+  diagnostics, as well as displaying statistical information.  We
+  strongly recommend downloading the latest version of Ethtool at:
+
+  http://sourceforge.net/projects/gkernel.
+
+  Speed and Duplex
+  ----------------
+  Speed and Duplex are configured through the Ethtool* utility. For
+  instructions,  refer to the Ethtool man page.
+
+  Enabling Wake on LAN* (WoL)
+  ---------------------------
+  WoL is configured through the Ethtool* utility. For instructions on
+  enabling WoL with Ethtool, refer to the Ethtool man page.
+
+  WoL will be enabled on the system during the next shut down or reboot.
+  For this driver version, in order to enable WoL, the e1000e driver must be
+  loaded when shutting down or rebooting the system.
+
+  In most cases Wake On LAN is only supported on port A for multiple port
+  adapters. To verify if a port supports Wake on LAN run ethtool eth<X>.
+
+
+Support
+=======
+
+For general information, go to the Intel support website at:
+
+    www.intel.com/support/
+
+or the Intel Wired Networking project hosted by Sourceforge at:
+
+    http://sourceforge.net/projects/e1000
+
+If an issue is identified with the released source code on the supported
+kernel with a supported adapter, email the specific information related
+to the issue to e1000-devel@lists.sf.net
old mode 100755 (executable)
new mode 100644 (file)
index 19015de..21dd5d1
@@ -1,19 +1,16 @@
 Linux* Base Driver for Intel(R) Network Connection
 ==================================================
 
-November 24, 2009
+Intel Gigabit Linux driver.
+Copyright(c) 1999 - 2010 Intel Corporation.
 
 Contents
 ========
 
-- In This Release
 - Identifying Your Adapter
 - Known Issues/Troubleshooting
 - Support
 
-In This Release
-===============
-
 This file describes the ixgbevf Linux* Base Driver for Intel Network
 Connection.
 
@@ -33,7 +30,7 @@ Identifying Your Adapter
 For more information on how to identify your adapter, go to the Adapter &
 Driver ID Guide at:
 
-    http://support.intel.com/support/network/sb/CS-008441.htm
+    http://support.intel.com/support/go/network/adapter/idguide.htm
 
 Known Issues/Troubleshooting
 ============================
@@ -57,34 +54,3 @@ or the Intel Wired Networking project hosted by Sourceforge at:
 If an issue is identified with the released source code on the supported
 kernel with a supported adapter, email the specific information related
 to the issue to e1000-devel@lists.sf.net
-
-License
-=======
-
-Intel 10 Gigabit Linux driver.
-Copyright(c) 1999 - 2009 Intel Corporation.
-
-This program is free software; you can redistribute it and/or modify it
-under the terms and conditions of the GNU General Public License,
-version 2, as published by the Free Software Foundation.
-
-This program is distributed in the hope it will be useful, but WITHOUT
-ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
-FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
-more details.
-
-You should have received a copy of the GNU General Public License along with
-this program; if not, write to the Free Software Foundation, Inc.,
-51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
-
-The full GNU General Public License is included in this distribution in
-the file called "COPYING".
-
-Trademarks
-==========
-
-Intel, Itanium, and Pentium are trademarks or registered trademarks of
-Intel Corporation or its subsidiaries in the United States and other
-countries.
-
-* Other names and brands may be claimed as the property of others.
index ccd951f..cc96ee2 100644 (file)
@@ -478,7 +478,7 @@ static void prepare_hwpoison_fd(void)
        }
 
        if (opt_unpoison && !hwpoison_forget_fd) {
-               sprintf(buf, "%s/renew-pfn", hwpoison_debug_fs);
+               sprintf(buf, "%s/unpoison-pfn", hwpoison_debug_fs);
                hwpoison_forget_fd = checked_open(buf, O_WRONLY);
        }
 }
index f46d8e6..3d4179f 100644 (file)
@@ -969,6 +969,16 @@ L: linux-samsung-soc@vger.kernel.org (moderated for non-subscribers)
 S:     Maintained
 F:     arch/arm/mach-s5p*/
 
+ARM/SAMSUNG S5P SERIES FIMC SUPPORT
+M:     Kyungmin Park <kyungmin.park@samsung.com>
+M:     Sylwester Nawrocki <s.nawrocki@samsung.com>
+L:     linux-arm-kernel@lists.infradead.org
+L:     linux-media@vger.kernel.org
+S:     Maintained
+F:     arch/arm/plat-s5p/dev-fimc*
+F:     arch/arm/plat-samsung/include/plat/*fimc*
+F:     drivers/media/video/s5p-fimc/
+
 ARM/SHMOBILE ARM ARCHITECTURE
 M:     Paul Mundt <lethal@linux-sh.org>
 M:     Magnus Damm <magnus.damm@gmail.com>
@@ -1517,6 +1527,8 @@ T:        git git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client.git
 S:     Supported
 F:     Documentation/filesystems/ceph.txt
 F:     fs/ceph
+F:     net/ceph
+F:     include/linux/ceph
 
 CERTIFIED WIRELESS USB (WUSB) SUBSYSTEM:
 M:     David Vrabel <david.vrabel@csr.com>
@@ -2535,7 +2547,7 @@ S:        Supported
 F:     drivers/scsi/gdt*
 
 GENERIC GPIO I2C DRIVER
-M:     Haavard Skinnemoen <hskinnemoen@atmel.com>
+M:     Haavard Skinnemoen <hskinnemoen@gmail.com>
 S:     Supported
 F:     drivers/i2c/busses/i2c-gpio.c
 F:     include/linux/i2c-gpio.h
@@ -3063,16 +3075,27 @@ L:      netdev@vger.kernel.org
 S:     Maintained
 F:     drivers/net/ixp2000/
 
-INTEL ETHERNET DRIVERS (e100/e1000/e1000e/igb/igbvf/ixgb/ixgbe)
+INTEL ETHERNET DRIVERS (e100/e1000/e1000e/igb/igbvf/ixgb/ixgbe/ixgbevf)
 M:     Jeff Kirsher <jeffrey.t.kirsher@intel.com>
 M:     Jesse Brandeburg <jesse.brandeburg@intel.com>
 M:     Bruce Allan <bruce.w.allan@intel.com>
-M:     Alex Duyck <alexander.h.duyck@intel.com>
+M:     Carolyn Wyborny <carolyn.wyborny@intel.com>
+M:     Don Skidmore <donald.c.skidmore@intel.com>
+M:     Greg Rose <gregory.v.rose@intel.com>
 M:     PJ Waskiewicz <peter.p.waskiewicz.jr@intel.com>
+M:     Alex Duyck <alexander.h.duyck@intel.com>
 M:     John Ronciak <john.ronciak@intel.com>
 L:     e1000-devel@lists.sourceforge.net
 W:     http://e1000.sourceforge.net/
 S:     Supported
+F:     Documentation/networking/e100.txt
+F:     Documentation/networking/e1000.txt
+F:     Documentation/networking/e1000e.txt
+F:     Documentation/networking/igb.txt
+F:     Documentation/networking/igbvf.txt
+F:     Documentation/networking/ixgb.txt
+F:     Documentation/networking/ixgbe.txt
+F:     Documentation/networking/ixgbevf.txt
 F:     drivers/net/e100.c
 F:     drivers/net/e1000/
 F:     drivers/net/e1000e/
@@ -3080,6 +3103,7 @@ F:        drivers/net/igb/
 F:     drivers/net/igbvf/
 F:     drivers/net/ixgb/
 F:     drivers/net/ixgbe/
+F:     drivers/net/ixgbevf/
 
 INTEL PRO/WIRELESS 2100 NETWORK CONNECTION SUPPORT
 L:     linux-wireless@vger.kernel.org
@@ -3140,7 +3164,7 @@ F:        drivers/net/ioc3-eth.c
 
 IOC3 SERIAL DRIVER
 M:     Pat Gefre <pfg@sgi.com>
-L:     linux-mips@linux-mips.org
+L:     linux-serial@vger.kernel.org
 S:     Maintained
 F:     drivers/serial/ioc3_serial.c
 
@@ -4783,6 +4807,15 @@ F:       fs/qnx4/
 F:     include/linux/qnx4_fs.h
 F:     include/linux/qnxtypes.h
 
+RADOS BLOCK DEVICE (RBD)
+F:     include/linux/qnxtypes.h
+M:     Yehuda Sadeh <yehuda@hq.newdream.net>
+M:     Sage Weil <sage@newdream.net>
+M:     ceph-devel@vger.kernel.org
+S:     Supported
+F:     drivers/block/rbd.c
+F:     drivers/block/rbd_types.h
+
 RADEON FRAMEBUFFER DISPLAY DRIVER
 M:     Benjamin Herrenschmidt <benh@kernel.crashing.org>
 L:     linux-fbdev@vger.kernel.org
@@ -5008,6 +5041,12 @@ F:       drivers/media/common/saa7146*
 F:     drivers/media/video/*7146*
 F:     include/media/*7146*
 
+SAMSUNG AUDIO (ASoC) DRIVERS
+M:     Jassi Brar <jassi.brar@samsung.com>
+L:     alsa-devel@alsa-project.org (moderated for non-subscribers)
+S:     Supported
+F:     sound/soc/s3c24xx
+
 TLG2300 VIDEO4LINUX-2 DRIVER
 M:     Huang Shijie <shijie8@gmail.com>
 M:     Kang Yong <kangyong@telegent.com>
@@ -6450,8 +6489,10 @@ F:       include/linux/wm97xx.h
 WOLFSON MICROELECTRONICS DRIVERS
 M:     Mark Brown <broonie@opensource.wolfsonmicro.com>
 M:     Ian Lartey <ian@opensource.wolfsonmicro.com>
+M:     Dimitris Papastamos <dp@opensource.wolfsonmicro.com>
+T:     git git://opensource.wolfsonmicro.com/linux-2.6-asoc
 T:     git git://opensource.wolfsonmicro.com/linux-2.6-audioplus
-W:     http://opensource.wolfsonmicro.com/node/8
+W:     http://opensource.wolfsonmicro.com/content/linux-drivers-wolfson-devices
 S:     Supported
 F:     Documentation/hwmon/wm83??
 F:     drivers/leds/leds-wm83*.c
index 77b5c6e..d3c1071 100644 (file)
--- a/Makefile
+++ b/Makefile
@@ -1,8 +1,8 @@
 VERSION = 2
 PATCHLEVEL = 6
 SUBLEVEL = 36
-EXTRAVERSION = -rc7
-NAME = Sheep on Meth
+EXTRAVERSION =
+NAME = Flesh-Eating Bats with Fangs
 
 # *DOCUMENTATION*
 # To see a list of typical targets execute "make help"
@@ -568,6 +568,12 @@ endif
 
 ifdef CONFIG_FUNCTION_TRACER
 KBUILD_CFLAGS  += -pg
+ifdef CONFIG_DYNAMIC_FTRACE
+       ifdef CONFIG_HAVE_C_RECORDMCOUNT
+               BUILD_C_RECORDMCOUNT := y
+               export BUILD_C_RECORDMCOUNT
+       endif
+endif
 endif
 
 # We trigger additional mismatches with less inlining
@@ -591,6 +597,11 @@ KBUILD_CFLAGS      += $(call cc-option,-fno-strict-overflow)
 # conserve stack if available
 KBUILD_CFLAGS   += $(call cc-option,-fconserve-stack)
 
+# check for 'asm goto'
+ifeq ($(shell $(CONFIG_SHELL) $(srctree)/scripts/gcc-goto.sh $(CC)), y)
+       KBUILD_CFLAGS += -DCC_HAVE_ASM_GOTO
+endif
+
 # Add user supplied CPPFLAGS, AFLAGS and CFLAGS as the last assignments
 # But warn user when we do so
 warn-assign = \
index fe48fc7..53d7f61 100644 (file)
@@ -158,4 +158,7 @@ config HAVE_PERF_EVENTS_NMI
          subsystem.  Also has support for calculating CPU cycle events
          to determine how many clock cycles in a given period.
 
+config HAVE_ARCH_JUMP_LABEL
+       bool
+
 source "kernel/gcov/Kconfig"
index b9647bb..d04ccd7 100644 (file)
@@ -9,6 +9,7 @@ config ALPHA
        select HAVE_IDE
        select HAVE_OPROFILE
        select HAVE_SYSCALL_WRAPPERS
+       select HAVE_IRQ_WORK
        select HAVE_PERF_EVENTS
        select HAVE_DMA_ATTRS
        help
index 4157cd3..fe792ca 100644 (file)
@@ -1,11 +1,6 @@
 #ifndef __ASM_ALPHA_PERF_EVENT_H
 #define __ASM_ALPHA_PERF_EVENT_H
 
-/* Alpha only supports software events through this interface. */
-extern void set_perf_event_pending(void);
-
-#define PERF_EVENT_INDEX_OFFSET 0
-
 #ifdef CONFIG_PERF_EVENTS
 extern void init_hw_perf_events(void);
 #else
index 85d8e4f..1cc4968 100644 (file)
@@ -307,7 +307,7 @@ again:
                             new_raw_count) != prev_raw_count)
                goto again;
 
-       delta = (new_raw_count  - (prev_raw_count & alpha_pmu->pmc_count_mask[idx])) + ovf;
+       delta = (new_raw_count - (prev_raw_count & alpha_pmu->pmc_count_mask[idx])) + ovf;
 
        /* It is possible on very rare occasions that the PMC has overflowed
         * but the interrupt is yet to come.  Detect and fix this situation.
@@ -402,14 +402,13 @@ static void maybe_change_configuration(struct cpu_hw_events *cpuc)
                struct hw_perf_event *hwc = &pe->hw;
                int idx = hwc->idx;
 
-               if (cpuc->current_idx[j] != PMC_NO_INDEX) {
-                       cpuc->idx_mask |= (1<<cpuc->current_idx[j]);
-                       continue;
+               if (cpuc->current_idx[j] == PMC_NO_INDEX) {
+                       alpha_perf_event_set_period(pe, hwc, idx);
+                       cpuc->current_idx[j] = idx;
                }
 
-               alpha_perf_event_set_period(pe, hwc, idx);
-               cpuc->current_idx[j] = idx;
-               cpuc->idx_mask |= (1<<cpuc->current_idx[j]);
+               if (!(hwc->state & PERF_HES_STOPPED))
+                       cpuc->idx_mask |= (1<<cpuc->current_idx[j]);
        }
        cpuc->config = cpuc->event[0]->hw.config_base;
 }
@@ -420,12 +419,13 @@ static void maybe_change_configuration(struct cpu_hw_events *cpuc)
  *  - this function is called from outside this module via the pmu struct
  *    returned from perf event initialisation.
  */
-static int alpha_pmu_enable(struct perf_event *event)
+static int alpha_pmu_add(struct perf_event *event, int flags)
 {
        struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
+       struct hw_perf_event *hwc = &event->hw;
        int n0;
        int ret;
-       unsigned long flags;
+       unsigned long irq_flags;
 
        /*
         * The Sparc code has the IRQ disable first followed by the perf
@@ -435,8 +435,8 @@ static int alpha_pmu_enable(struct perf_event *event)
         * nevertheless we disable the PMCs first to enable a potential
         * final PMI to occur before we disable interrupts.
         */
-       perf_disable();
-       local_irq_save(flags);
+       perf_pmu_disable(event->pmu);
+       local_irq_save(irq_flags);
 
        /* Default to error to be returned */
        ret = -EAGAIN;
@@ -455,8 +455,12 @@ static int alpha_pmu_enable(struct perf_event *event)
                }
        }
 
-       local_irq_restore(flags);
-       perf_enable();
+       hwc->state = PERF_HES_UPTODATE;
+       if (!(flags & PERF_EF_START))
+               hwc->state |= PERF_HES_STOPPED;
+
+       local_irq_restore(irq_flags);
+       perf_pmu_enable(event->pmu);
 
        return ret;
 }
@@ -467,15 +471,15 @@ static int alpha_pmu_enable(struct perf_event *event)
  *  - this function is called from outside this module via the pmu struct
  *    returned from perf event initialisation.
  */
-static void alpha_pmu_disable(struct perf_event *event)
+static void alpha_pmu_del(struct perf_event *event, int flags)
 {
        struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
        struct hw_perf_event *hwc = &event->hw;
-       unsigned long flags;
+       unsigned long irq_flags;
        int j;
 
-       perf_disable();
-       local_irq_save(flags);
+       perf_pmu_disable(event->pmu);
+       local_irq_save(irq_flags);
 
        for (j = 0; j < cpuc->n_events; j++) {
                if (event == cpuc->event[j]) {
@@ -501,8 +505,8 @@ static void alpha_pmu_disable(struct perf_event *event)
                }
        }
 
-       local_irq_restore(flags);
-       perf_enable();
+       local_irq_restore(irq_flags);
+       perf_pmu_enable(event->pmu);
 }
 
 
@@ -514,13 +518,44 @@ static void alpha_pmu_read(struct perf_event *event)
 }
 
 
-static void alpha_pmu_unthrottle(struct perf_event *event)
+static void alpha_pmu_stop(struct perf_event *event, int flags)
+{
+       struct hw_perf_event *hwc = &event->hw;
+       struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
+
+       if (!(hwc->state & PERF_HES_STOPPED)) {
+               cpuc->idx_mask &= ~(1UL<<hwc->idx);
+               hwc->state |= PERF_HES_STOPPED;
+       }
+
+       if ((flags & PERF_EF_UPDATE) && !(hwc->state & PERF_HES_UPTODATE)) {
+               alpha_perf_event_update(event, hwc, hwc->idx, 0);
+               hwc->state |= PERF_HES_UPTODATE;
+       }
+
+       if (cpuc->enabled)
+               wrperfmon(PERFMON_CMD_DISABLE, (1UL<<hwc->idx));
+}
+
+
+static void alpha_pmu_start(struct perf_event *event, int flags)
 {
        struct hw_perf_event *hwc = &event->hw;
        struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
 
+       if (WARN_ON_ONCE(!(hwc->state & PERF_HES_STOPPED)))
+               return;
+
+       if (flags & PERF_EF_RELOAD) {
+               WARN_ON_ONCE(!(hwc->state & PERF_HES_UPTODATE));
+               alpha_perf_event_set_period(event, hwc, hwc->idx);
+       }
+
+       hwc->state = 0;
+
        cpuc->idx_mask |= 1UL<<hwc->idx;
-       wrperfmon(PERFMON_CMD_ENABLE, (1UL<<hwc->idx));
+       if (cpuc->enabled)
+               wrperfmon(PERFMON_CMD_ENABLE, (1UL<<hwc->idx));
 }
 
 
@@ -642,39 +677,36 @@ static int __hw_perf_event_init(struct perf_event *event)
        return 0;
 }
 
-static const struct pmu pmu = {
-       .enable         = alpha_pmu_enable,
-       .disable        = alpha_pmu_disable,
-       .read           = alpha_pmu_read,
-       .unthrottle     = alpha_pmu_unthrottle,
-};
-
-
 /*
  * Main entry point to initialise a HW performance event.
  */
-const struct pmu *hw_perf_event_init(struct perf_event *event)
+static int alpha_pmu_event_init(struct perf_event *event)
 {
        int err;
 
+       switch (event->attr.type) {
+       case PERF_TYPE_RAW:
+       case PERF_TYPE_HARDWARE:
+       case PERF_TYPE_HW_CACHE:
+               break;
+
+       default:
+               return -ENOENT;
+       }
+
        if (!alpha_pmu)
-               return ERR_PTR(-ENODEV);
+               return -ENODEV;
 
        /* Do the real initialisation work. */
        err = __hw_perf_event_init(event);
 
-       if (err)
-               return ERR_PTR(err);
-
-       return &pmu;
+       return err;
 }
 
-
-
 /*
  * Main entry point - enable HW performance counters.
  */
-void hw_perf_enable(void)
+static void alpha_pmu_enable(struct pmu *pmu)
 {
        struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
 
@@ -700,7 +732,7 @@ void hw_perf_enable(void)
  * Main entry point - disable HW performance counters.
  */
 
-void hw_perf_disable(void)
+static void alpha_pmu_disable(struct pmu *pmu)
 {
        struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
 
@@ -713,6 +745,17 @@ void hw_perf_disable(void)
        wrperfmon(PERFMON_CMD_DISABLE, cpuc->idx_mask);
 }
 
+static struct pmu pmu = {
+       .pmu_enable     = alpha_pmu_enable,
+       .pmu_disable    = alpha_pmu_disable,
+       .event_init     = alpha_pmu_event_init,
+       .add            = alpha_pmu_add,
+       .del            = alpha_pmu_del,
+       .start          = alpha_pmu_start,
+       .stop           = alpha_pmu_stop,
+       .read           = alpha_pmu_read,
+};
+
 
 /*
  * Main entry point - don't know when this is called but it
@@ -766,7 +809,7 @@ static void alpha_perf_event_irq_handler(unsigned long la_ptr,
        wrperfmon(PERFMON_CMD_DISABLE, cpuc->idx_mask);
 
        /* la_ptr is the counter that overflowed. */
-       if (unlikely(la_ptr >= perf_max_events)) {
+       if (unlikely(la_ptr >= alpha_pmu->num_pmcs)) {
                /* This should never occur! */
                irq_err_count++;
                pr_warning("PMI: silly index %ld\n", la_ptr);
@@ -807,7 +850,7 @@ static void alpha_perf_event_irq_handler(unsigned long la_ptr,
                        /* Interrupts coming too quickly; "throttle" the
                         * counter, i.e., disable it for a little while.
                         */
-                       cpuc->idx_mask &= ~(1UL<<idx);
+                       alpha_pmu_stop(event, 0);
                }
        }
        wrperfmon(PERFMON_CMD_ENABLE, cpuc->idx_mask);
@@ -837,6 +880,7 @@ void __init init_hw_perf_events(void)
 
        /* And set up PMU specification */
        alpha_pmu = &ev67_pmu;
-       perf_max_events = alpha_pmu->num_pmcs;
+
+       perf_pmu_register(&pmu);
 }
 
index 396af17..0f1d849 100644 (file)
@@ -41,7 +41,7 @@
 #include <linux/init.h>
 #include <linux/bcd.h>
 #include <linux/profile.h>
-#include <linux/perf_event.h>
+#include <linux/irq_work.h>
 
 #include <asm/uaccess.h>
 #include <asm/io.h>
@@ -83,25 +83,25 @@ static struct {
 
 unsigned long est_cycle_freq;
 
-#ifdef CONFIG_PERF_EVENTS
+#ifdef CONFIG_IRQ_WORK
 
-DEFINE_PER_CPU(u8, perf_event_pending);
+DEFINE_PER_CPU(u8, irq_work_pending);
 
-#define set_perf_event_pending_flag()  __get_cpu_var(perf_event_pending) = 1
-#define test_perf_event_pending()      __get_cpu_var(perf_event_pending)
-#define clear_perf_event_pending()     __get_cpu_var(perf_event_pending) = 0
+#define set_irq_work_pending_flag()  __get_cpu_var(irq_work_pending) = 1
+#define test_irq_work_pending()      __get_cpu_var(irq_work_pending)
+#define clear_irq_work_pending()     __get_cpu_var(irq_work_pending) = 0
 
-void set_perf_event_pending(void)
+void set_irq_work_pending(void)
 {
-       set_perf_event_pending_flag();
+       set_irq_work_pending_flag();
 }
 
-#else  /* CONFIG_PERF_EVENTS */
+#else  /* CONFIG_IRQ_WORK */
 
-#define test_perf_event_pending()      0
-#define clear_perf_event_pending()
+#define test_irq_work_pending()      0
+#define clear_irq_work_pending()
 
-#endif /* CONFIG_PERF_EVENTS */
+#endif /* CONFIG_IRQ_WORK */
 
 
 static inline __u32 rpcc(void)
@@ -191,9 +191,9 @@ irqreturn_t timer_interrupt(int irq, void *dev)
 
        write_sequnlock(&xtime_lock);
 
-       if (test_perf_event_pending()) {
-               clear_perf_event_pending();
-               perf_event_do_pending();
+       if (test_irq_work_pending()) {
+               clear_irq_work_pending();
+               irq_work_run();
        }
 
 #ifndef CONFIG_SMP
index 88c97bc..9103904 100644 (file)
@@ -23,6 +23,7 @@ config ARM
        select HAVE_KERNEL_GZIP
        select HAVE_KERNEL_LZO
        select HAVE_KERNEL_LZMA
+       select HAVE_IRQ_WORK
        select HAVE_PERF_EVENTS
        select PERF_USE_VMALLOC
        select HAVE_REGS_AND_STACK_ACCESS_API
@@ -1101,6 +1102,20 @@ config ARM_ERRATA_720789
          invalidated are not, resulting in an incoherency in the system page
          tables. The workaround changes the TLB flushing routines to invalidate
          entries regardless of the ASID.
+
+config ARM_ERRATA_743622
+       bool "ARM errata: Faulty hazard checking in the Store Buffer may lead to data corruption"
+       depends on CPU_V7
+       help
+         This option enables the workaround for the 743622 Cortex-A9
+         (r2p0..r2p2) erratum. Under very rare conditions, a faulty
+         optimisation in the Cortex-A9 Store Buffer may lead to data
+         corruption. This workaround sets a specific bit in the diagnostic
+         register of the Cortex-A9 which disables the Store Buffer
+         optimisation, preventing the defect from occurring. This has no
+         visible impact on the overall performance or power consumption of the
+         processor.
+
 endmenu
 
 source "arch/arm/common/Kconfig"
index b5799a3..c4aa4e8 100644 (file)
 #ifndef __ARM_PERF_EVENT_H__
 #define __ARM_PERF_EVENT_H__
 
-/*
- * NOP: on *most* (read: all supported) ARM platforms, the performance
- * counter interrupts are regular interrupts and not an NMI. This
- * means that when we receive the interrupt we can call
- * perf_event_do_pending() that handles all of the work with
- * interrupts disabled.
- */
-static inline void
-set_perf_event_pending(void)
-{
-}
-
 /* ARM performance counters start from 1 (in the cp15 accesses) so use the
  * same indexes here for consistency. */
 #define PERF_EVENT_INDEX_OFFSET 1
index 8bccbfa..2c1f005 100644 (file)
@@ -1162,11 +1162,12 @@ space_cccc_001x(kprobe_opcode_t insn, struct arch_specific_insn *asi)
 {
        /*
         * MSR   : cccc 0011 0x10 xxxx xxxx xxxx xxxx xxxx
-        * Undef : cccc 0011 0x00 xxxx xxxx xxxx xxxx xxxx
+        * Undef : cccc 0011 0100 xxxx xxxx xxxx xxxx xxxx
         * ALU op with S bit and Rd == 15 :
         *         cccc 001x xxx1 xxxx 1111 xxxx xxxx xxxx
         */
-       if ((insn & 0x0f900000) == 0x03200000 ||        /* MSR & Undef */
+       if ((insn & 0x0fb00000) == 0x03200000 ||        /* MSR */
+           (insn & 0x0ff00000) == 0x03400000 ||        /* Undef */
            (insn & 0x0e10f000) == 0x0210f000)          /* ALU s-bit, R15  */
                return INSN_REJECTED;
 
@@ -1177,7 +1178,7 @@ space_cccc_001x(kprobe_opcode_t insn, struct arch_specific_insn *asi)
         * *S (bit 20) updates condition codes
         * ADC/SBC/RSC reads the C flag
         */
-       insn &= 0xfff00fff;     /* Rn = r0, Rd = r0 */
+       insn &= 0xffff0fff;     /* Rd = r0 */
        asi->insn[0] = insn;
        asi->insn_handler = (insn & (1 << 20)) ?  /* S-bit */
                        emulate_alu_imm_rwflags : emulate_alu_imm_rflags;
index ecbb028..49643b1 100644 (file)
@@ -123,6 +123,12 @@ armpmu_get_max_events(void)
 }
 EXPORT_SYMBOL_GPL(armpmu_get_max_events);
 
+int perf_num_counters(void)
+{
+       return armpmu_get_max_events();
+}
+EXPORT_SYMBOL_GPL(perf_num_counters);
+
 #define HW_OP_UNSUPPORTED              0xFFFF
 
 #define C(_x) \
@@ -221,46 +227,56 @@ again:
 }
 
 static void
-armpmu_disable(struct perf_event *event)
+armpmu_read(struct perf_event *event)
 {
-       struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
        struct hw_perf_event *hwc = &event->hw;
-       int idx = hwc->idx;
-
-       WARN_ON(idx < 0);
-
-       clear_bit(idx, cpuc->active_mask);
-       armpmu->disable(hwc, idx);
-
-       barrier();
 
-       armpmu_event_update(event, hwc, idx);
-       cpuc->events[idx] = NULL;
-       clear_bit(idx, cpuc->used_mask);
+       /* Don't read disabled counters! */
+       if (hwc->idx < 0)
+               return;
 
-       perf_event_update_userpage(event);
+       armpmu_event_update(event, hwc, hwc->idx);
 }
 
 static void
-armpmu_read(struct perf_event *event)
+armpmu_stop(struct perf_event *event, int flags)
 {
        struct hw_perf_event *hwc = &event->hw;
 
-       /* Don't read disabled counters! */
-       if (hwc->idx < 0)
+       if (!armpmu)
                return;
 
-       armpmu_event_update(event, hwc, hwc->idx);
+       /*
+        * ARM pmu always has to update the counter, so ignore
+        * PERF_EF_UPDATE, see comments in armpmu_start().
+        */
+       if (!(hwc->state & PERF_HES_STOPPED)) {
+               armpmu->disable(hwc, hwc->idx);
+               barrier(); /* why? */
+               armpmu_event_update(event, hwc, hwc->idx);
+               hwc->state |= PERF_HES_STOPPED | PERF_HES_UPTODATE;
+       }
 }
 
 static void
-armpmu_unthrottle(struct perf_event *event)
+armpmu_start(struct perf_event *event, int flags)
 {
        struct hw_perf_event *hwc = &event->hw;
 
+       if (!armpmu)
+               return;
+
+       /*
+        * ARM pmu always has to reprogram the period, so ignore
+        * PERF_EF_RELOAD, see the comment below.
+        */
+       if (flags & PERF_EF_RELOAD)
+               WARN_ON_ONCE(!(hwc->state & PERF_HES_UPTODATE));
+
+       hwc->state = 0;
        /*
         * Set the period again. Some counters can't be stopped, so when we
-        * were throttled we simply disabled the IRQ source and the counter
+        * were stopped we simply disabled the IRQ source and the counter
         * may have been left counting. If we don't do this step then we may
         * get an interrupt too soon or *way* too late if the overflow has
         * happened since disabling.
@@ -269,14 +285,33 @@ armpmu_unthrottle(struct perf_event *event)
        armpmu->enable(hwc, hwc->idx);
 }
 
+static void
+armpmu_del(struct perf_event *event, int flags)
+{
+       struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
+       struct hw_perf_event *hwc = &event->hw;
+       int idx = hwc->idx;
+
+       WARN_ON(idx < 0);
+
+       clear_bit(idx, cpuc->active_mask);
+       armpmu_stop(event, PERF_EF_UPDATE);
+       cpuc->events[idx] = NULL;
+       clear_bit(idx, cpuc->used_mask);
+
+       perf_event_update_userpage(event);
+}
+
 static int
-armpmu_enable(struct perf_event *event)
+armpmu_add(struct perf_event *event, int flags)
 {
        struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
        struct hw_perf_event *hwc = &event->hw;
        int idx;
        int err = 0;
 
+       perf_pmu_disable(event->pmu);
+
        /* If we don't have a space for the counter then finish early. */
        idx = armpmu->get_event_idx(cpuc, hwc);
        if (idx < 0) {
@@ -293,25 +328,19 @@ armpmu_enable(struct perf_event *event)
        cpuc->events[idx] = event;
        set_bit(idx, cpuc->active_mask);
 
-       /* Set the period for the event. */
-       armpmu_event_set_period(event, hwc, idx);
-
-       /* Enable the event. */
-       armpmu->enable(hwc, idx);
+       hwc->state = PERF_HES_STOPPED | PERF_HES_UPTODATE;
+       if (flags & PERF_EF_START)
+               armpmu_start(event, PERF_EF_RELOAD);
 
        /* Propagate our changes to the userspace mapping. */
        perf_event_update_userpage(event);
 
 out:
+       perf_pmu_enable(event->pmu);
        return err;
 }
 
-static struct pmu pmu = {
-       .enable     = armpmu_enable,
-       .disable    = armpmu_disable,
-       .unthrottle = armpmu_unthrottle,
-       .read       = armpmu_read,
-};
+static struct pmu pmu;
 
 static int
 validate_event(struct cpu_hw_events *cpuc,
@@ -491,20 +520,29 @@ __hw_perf_event_init(struct perf_event *event)
        return err;
 }
 
-const struct pmu *
-hw_perf_event_init(struct perf_event *event)
+static int armpmu_event_init(struct perf_event *event)
 {
        int err = 0;
 
+       switch (event->attr.type) {
+       case PERF_TYPE_RAW:
+       case PERF_TYPE_HARDWARE:
+       case PERF_TYPE_HW_CACHE:
+               break;
+
+       default:
+               return -ENOENT;
+       }
+
        if (!armpmu)
-               return ERR_PTR(-ENODEV);
+               return -ENODEV;
 
        event->destroy = hw_perf_event_destroy;
 
        if (!atomic_inc_not_zero(&active_events)) {
-               if (atomic_read(&active_events) > perf_max_events) {
+               if (atomic_read(&active_events) > armpmu->num_events) {
                        atomic_dec(&active_events);
-                       return ERR_PTR(-ENOSPC);
+                       return -ENOSPC;
                }
 
                mutex_lock(&pmu_reserve_mutex);
@@ -518,17 +556,16 @@ hw_perf_event_init(struct perf_event *event)
        }
 
        if (err)
-               return ERR_PTR(err);
+               return err;
 
        err = __hw_perf_event_init(event);
        if (err)
                hw_perf_event_destroy(event);
 
-       return err ? ERR_PTR(err) : &pmu;
+       return err;
 }
 
-void
-hw_perf_enable(void)
+static void armpmu_enable(struct pmu *pmu)
 {
        /* Enable all of the perf events on hardware. */
        int idx;
@@ -549,13 +586,23 @@ hw_perf_enable(void)
        armpmu->start();
 }
 
-void
-hw_perf_disable(void)
+static void armpmu_disable(struct pmu *pmu)
 {
        if (armpmu)
                armpmu->stop();
 }
 
+static struct pmu pmu = {
+       .pmu_enable     = armpmu_enable,
+       .pmu_disable    = armpmu_disable,
+       .event_init     = armpmu_event_init,
+       .add            = armpmu_add,
+       .del            = armpmu_del,
+       .start          = armpmu_start,
+       .stop           = armpmu_stop,
+       .read           = armpmu_read,
+};
+
 /*
  * ARMv6 Performance counter handling code.
  *
@@ -1045,7 +1092,7 @@ armv6pmu_handle_irq(int irq_num,
         * platforms that can have the PMU interrupts raised as an NMI, this
         * will not work.
         */
-       perf_event_do_pending();
+       irq_work_run();
 
        return IRQ_HANDLED;
 }
@@ -2021,7 +2068,7 @@ static irqreturn_t armv7pmu_handle_irq(int irq_num, void *dev)
         * platforms that can have the PMU interrupts raised as an NMI, this
         * will not work.
         */
-       perf_event_do_pending();
+       irq_work_run();
 
        return IRQ_HANDLED;
 }
@@ -2389,7 +2436,7 @@ xscale1pmu_handle_irq(int irq_num, void *dev)
                        armpmu->disable(hwc, idx);
        }
 
-       perf_event_do_pending();
+       irq_work_run();
 
        /*
         * Re-enable the PMU.
@@ -2716,7 +2763,7 @@ xscale2pmu_handle_irq(int irq_num, void *dev)
                        armpmu->disable(hwc, idx);
        }
 
-       perf_event_do_pending();
+       irq_work_run();
 
        /*
         * Re-enable the PMU.
@@ -2933,14 +2980,12 @@ init_hw_perf_events(void)
                        armpmu = &armv6pmu;
                        memcpy(armpmu_perf_cache_map, armv6_perf_cache_map,
                                        sizeof(armv6_perf_cache_map));
-                       perf_max_events = armv6pmu.num_events;
                        break;
                case 0xB020:    /* ARM11mpcore */
                        armpmu = &armv6mpcore_pmu;
                        memcpy(armpmu_perf_cache_map,
                               armv6mpcore_perf_cache_map,
                               sizeof(armv6mpcore_perf_cache_map));
-                       perf_max_events = armv6mpcore_pmu.num_events;
                        break;
                case 0xC080:    /* Cortex-A8 */
                        armv7pmu.id = ARM_PERF_PMU_ID_CA8;
@@ -2952,7 +2997,6 @@ init_hw_perf_events(void)
                        /* Reset PMNC and read the nb of CNTx counters
                            supported */
                        armv7pmu.num_events = armv7_reset_read_pmnc();
-                       perf_max_events = armv7pmu.num_events;
                        break;
                case 0xC090:    /* Cortex-A9 */
                        armv7pmu.id = ARM_PERF_PMU_ID_CA9;
@@ -2964,7 +3008,6 @@ init_hw_perf_events(void)
                        /* Reset PMNC and read the nb of CNTx counters
                            supported */
                        armv7pmu.num_events = armv7_reset_read_pmnc();
-                       perf_max_events = armv7pmu.num_events;
                        break;
                }
        /* Intel CPUs [xscale]. */
@@ -2975,13 +3018,11 @@ init_hw_perf_events(void)
                        armpmu = &xscale1pmu;
                        memcpy(armpmu_perf_cache_map, xscale_perf_cache_map,
                                        sizeof(xscale_perf_cache_map));
-                       perf_max_events = xscale1pmu.num_events;
                        break;
                case 2:
                        armpmu = &xscale2pmu;
                        memcpy(armpmu_perf_cache_map, xscale_perf_cache_map,
                                        sizeof(xscale_perf_cache_map));
-                       perf_max_events = xscale2pmu.num_events;
                        break;
                }
        }
@@ -2991,9 +3032,10 @@ init_hw_perf_events(void)
                                arm_pmu_names[armpmu->id], armpmu->num_events);
        } else {
                pr_info("no hardware support available\n");
-               perf_max_events = -1;
        }
 
+       perf_pmu_register(&pmu);
+
        return 0;
 }
 arch_initcall(init_hw_perf_events);
@@ -3001,13 +3043,6 @@ arch_initcall(init_hw_perf_events);
 /*
  * Callchain handling code.
  */
-static inline void
-callchain_store(struct perf_callchain_entry *entry,
-               u64 ip)
-{
-       if (entry->nr < PERF_MAX_STACK_DEPTH)
-               entry->ip[entry->nr++] = ip;
-}
 
 /*
  * The registers we're interested in are at the end of the variable
@@ -3039,7 +3074,7 @@ user_backtrace(struct frame_tail *tail,
        if (__copy_from_user_inatomic(&buftail, tail, sizeof(buftail)))
                return NULL;
 
-       callchain_store(entry, buftail.lr);
+       perf_callchain_store(entry, buftail.lr);
 
        /*
         * Frame pointers should strictly progress back up the stack
@@ -3051,16 +3086,11 @@ user_backtrace(struct frame_tail *tail,
        return buftail.fp - 1;
 }
 
-static void
-perf_callchain_user(struct pt_regs *regs,
-                   struct perf_callchain_entry *entry)
+void
+perf_callchain_user(struct perf_callchain_entry *entry, struct pt_regs *regs)
 {
        struct frame_tail *tail;
 
-       callchain_store(entry, PERF_CONTEXT_USER);
-
-       if (!user_mode(regs))
-               regs = task_pt_regs(current);
 
        tail = (struct frame_tail *)regs->ARM_fp - 1;
 
@@ -3078,56 +3108,18 @@ callchain_trace(struct stackframe *fr,
                void *data)
 {
        struct perf_callchain_entry *entry = data;
-       callchain_store(entry, fr->pc);
+       perf_callchain_store(entry, fr->pc);
        return 0;
 }
 
-static void
-perf_callchain_kernel(struct pt_regs *regs,
-                     struct perf_callchain_entry *entry)
+void
+perf_callchain_kernel(struct perf_callchain_entry *entry, struct pt_regs *regs)
 {
        struct stackframe fr;
 
-       callchain_store(entry, PERF_CONTEXT_KERNEL);
        fr.fp = regs->ARM_fp;
        fr.sp = regs->ARM_sp;
        fr.lr = regs->ARM_lr;
        fr.pc = regs->ARM_pc;
        walk_stackframe(&fr, callchain_trace, entry);
 }
-
-static void
-perf_do_callchain(struct pt_regs *regs,
-                 struct perf_callchain_entry *entry)
-{
-       int is_user;
-
-       if (!regs)
-               return;
-
-       is_user = user_mode(regs);
-
-       if (!current || !current->pid)
-               return;
-
-       if (is_user && current->state != TASK_RUNNING)
-               return;
-
-       if (!is_user)
-               perf_callchain_kernel(regs, entry);
-
-       if (current->mm)
-               perf_callchain_user(regs, entry);
-}
-
-static DEFINE_PER_CPU(struct perf_callchain_entry, pmc_irq_entry);
-
-struct perf_callchain_entry *
-perf_callchain(struct pt_regs *regs)
-{
-       struct perf_callchain_entry *entry = &__get_cpu_var(pmc_irq_entry);
-
-       entry->nr = 0;
-       perf_do_callchain(regs, entry);
-       return entry;
-}
index c80e090..ee8db15 100644 (file)
 
 static inline void arch_idle(void)
 {
-#ifndef CONFIG_DEBUG_KERNEL
        /*
         * Disable the processor clock.  The processor will be automatically
         * re-enabled by an interrupt or by a reset.
         */
        at91_sys_write(AT91_PMC_SCDR, AT91_PMC_PCK);
-#else
+#ifndef CONFIG_CPU_ARM920T
        /*
         * Set the processor (CP15) into 'Wait for Interrupt' mode.
-        * Unlike disabling the processor clock via the PMC (above)
-        *  this allows the processor to be woken via JTAG.
+        * Post-RM9200 processors need this in conjunction with the above
+        * to save power when idle.
         */
        cpu_do_idle();
 #endif
index 29c0a91..77eb35c 100644 (file)
@@ -691,7 +691,7 @@ int dma_init(void)
 
        memset(&gDMA, 0, sizeof(gDMA));
 
-       init_MUTEX_LOCKED(&gDMA.lock);
+       sema_init(&gDMA.lock, 0);
        init_waitqueue_head(&gDMA.freeChannelQ);
 
        /* Initialize the Hardware */
@@ -1574,7 +1574,7 @@ int dma_init_mem_map(DMA_MemMap_t *memMap)
 {
        memset(memMap, 0, sizeof(*memMap));
 
-       init_MUTEX(&memMap->lock);
+       sema_init(&memMap->lock, 1);
 
        return 0;
 }
index 8904ca4..a696d35 100644 (file)
@@ -276,7 +276,7 @@ static void channel_disable(struct m2p_channel *ch)
        v &= ~(M2P_CONTROL_STALL_IRQ_EN | M2P_CONTROL_NFB_IRQ_EN);
        m2p_set_control(ch, v);
 
-       while (m2p_channel_state(ch) == STATE_ON)
+       while (m2p_channel_state(ch) >= STATE_ON)
                cpu_relax();
 
        m2p_set_control(ch, 0x0);
index c5c0369..2f7e272 100644 (file)
@@ -122,6 +122,7 @@ config MACH_CPUIMX27
        select IMX_HAVE_PLATFORM_IMX_I2C
        select IMX_HAVE_PLATFORM_IMX_UART
        select IMX_HAVE_PLATFORM_MXC_NAND
+       select MXC_ULPI if USB_ULPI
        help
          Include support for Eukrea CPUIMX27 platform. This includes
          specific configurations for the module and its peripherals.
index 339150a..6830afd 100644 (file)
@@ -259,7 +259,7 @@ static void __init eukrea_cpuimx27_init(void)
        i2c_register_board_info(0, eukrea_cpuimx27_i2c_devices,
                                ARRAY_SIZE(eukrea_cpuimx27_i2c_devices));
 
-       imx27_add_i2c_imx1(&cpuimx27_i2c1_data);
+       imx27_add_i2c_imx0(&cpuimx27_i2c1_data);
 
        platform_add_devices(platform_devices, ARRAY_SIZE(platform_devices));
 
index 526f33a..ec592e8 100644 (file)
@@ -19,6 +19,7 @@
 #include <linux/sysdev.h>
 #include <linux/serial_core.h>
 #include <linux/platform_device.h>
+#include <linux/sched.h>
 
 #include <asm/mach/arch.h>
 #include <asm/mach/map.h>
index a48fb55..70ac681 100644 (file)
@@ -19,6 +19,7 @@
 #include <linux/sysdev.h>
 #include <linux/serial_core.h>
 #include <linux/platform_device.h>
+#include <linux/sched.h>
 
 #include <asm/mach/arch.h>
 #include <asm/mach/map.h>
index 251c92a..cd1afbc 100644 (file)
@@ -21,6 +21,7 @@
 #include <linux/sysdev.h>
 #include <linux/serial_core.h>
 #include <linux/platform_device.h>
+#include <linux/sched.h>
 
 #include <asm/mach/arch.h>
 #include <asm/mach/map.h>
index cfecd70..d562670 100644 (file)
@@ -173,11 +173,6 @@ static int s5pv210_clk_ip3_ctrl(struct clk *clk, int enable)
        return s5p_gatectrl(S5P_CLKGATE_IP3, clk, enable);
 }
 
-static int s5pv210_clk_ip4_ctrl(struct clk *clk, int enable)
-{
-       return s5p_gatectrl(S5P_CLKGATE_IP4, clk, enable);
-}
-
 static int s5pv210_clk_mask0_ctrl(struct clk *clk, int enable)
 {
        return s5p_gatectrl(S5P_CLK_SRC_MASK0, clk, enable);
index 77f456c..245b82b 100644 (file)
@@ -19,6 +19,7 @@
 #include <linux/io.h>
 #include <linux/sysdev.h>
 #include <linux/platform_device.h>
+#include <linux/sched.h>
 
 #include <asm/mach/arch.h>
 #include <asm/mach/map.h>
index efb1270..71fb173 100644 (file)
@@ -68,7 +68,7 @@ static void __init ct_ca9x4_init_irq(void)
 }
 
 #if 0
-static void ct_ca9x4_timer_init(void)
+static void __init ct_ca9x4_timer_init(void)
 {
        writel(0, MMIO_P2V(CT_CA9X4_TIMER0) + TIMER_CTRL);
        writel(0, MMIO_P2V(CT_CA9X4_TIMER1) + TIMER_CTRL);
@@ -222,7 +222,7 @@ static struct platform_device pmu_device = {
        .resource       = pmu_resources,
 };
 
-static void ct_ca9x4_init(void)
+static void __init ct_ca9x4_init(void)
 {
        int i;
 
index 817f0ad..7eaa232 100644 (file)
@@ -48,7 +48,7 @@ void __init v2m_map_io(struct map_desc *tile, size_t num)
 }
 
 
-static void v2m_timer_init(void)
+static void __init v2m_timer_init(void)
 {
        writel(0, MMIO_P2V(V2M_TIMER0) + TIMER_CTRL);
        writel(0, MMIO_P2V(V2M_TIMER1) + TIMER_CTRL);
index ab50627..17e7b0b 100644 (file)
@@ -204,8 +204,12 @@ void __iomem * __arm_ioremap_pfn_caller(unsigned long pfn,
        /*
         * Don't allow RAM to be mapped - this causes problems with ARMv6+
         */
-       if (WARN_ON(pfn_valid(pfn)))
-               return NULL;
+       if (pfn_valid(pfn)) {
+               printk(KERN_WARNING "BUG: Your driver calls ioremap() on system memory.  This leads\n"
+                      KERN_WARNING "to architecturally unpredictable behaviour on ARMv6+, and ioremap()\n"
+                      KERN_WARNING "will fail in the next kernel release.  Please fix your driver.\n");
+               WARN_ON(1);
+       }
 
        type = get_mem_type(mtype);
        if (!type)
index 6a3a2d0..e8ed9dc 100644 (file)
@@ -248,7 +248,7 @@ static struct mem_type mem_types[] = {
        },
        [MT_MEMORY] = {
                .prot_pte  = L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_DIRTY |
-                               L_PTE_USER | L_PTE_EXEC,
+                               L_PTE_WRITE | L_PTE_EXEC,
                .prot_l1   = PMD_TYPE_TABLE,
                .prot_sect = PMD_TYPE_SECT | PMD_SECT_AP_WRITE,
                .domain    = DOMAIN_KERNEL,
@@ -259,7 +259,7 @@ static struct mem_type mem_types[] = {
        },
        [MT_MEMORY_NONCACHED] = {
                .prot_pte  = L_PTE_PRESENT | L_PTE_YOUNG | L_PTE_DIRTY |
-                               L_PTE_USER | L_PTE_EXEC | L_PTE_MT_BUFFERABLE,
+                               L_PTE_WRITE | L_PTE_EXEC | L_PTE_MT_BUFFERABLE,
                .prot_l1   = PMD_TYPE_TABLE,
                .prot_sect = PMD_TYPE_SECT | PMD_SECT_AP_WRITE,
                .domain    = DOMAIN_KERNEL,
index 7563ff0..197f21b 100644 (file)
@@ -253,6 +253,14 @@ __v7_setup:
        orreq   r10, r10, #1 << 22              @ set bit #22
        mcreq   p15, 0, r10, c15, c0, 1         @ write diagnostic register
 #endif
+#ifdef CONFIG_ARM_ERRATA_743622
+       teq     r6, #0x20                       @ present in r2p0
+       teqne   r6, #0x21                       @ present in r2p1
+       teqne   r6, #0x22                       @ present in r2p2
+       mrceq   p15, 0, r10, c15, c0, 1         @ read diagnostic register
+       orreq   r10, r10, #1 << 6               @ set bit #6
+       mcreq   p15, 0, r10, c15, c0, 1         @ write diagnostic register
+#endif
 
 3:     mov     r10, #0
 #ifdef HARVARD_CACHE
@@ -365,7 +373,7 @@ __v7_ca9mp_proc_info:
        b       __v7_ca9mp_setup
        .long   cpu_arch_name
        .long   cpu_elf_name
-       .long   HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
+       .long   HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP|HWCAP_TLS
        .long   cpu_v7_name
        .long   v7_processor_functions
        .long   v7wbi_tlb_fns
index e666eaf..b2215c6 100644 (file)
@@ -6,4 +6,8 @@ DRIVER_OBJS = $(addprefix ../../../drivers/oprofile/, \
                oprofilefs.o oprofile_stats.o \
                timer_int.o )
 
+ifeq ($(CONFIG_HW_PERF_EVENTS),y)
+DRIVER_OBJS += $(addprefix ../../../drivers/oprofile/, oprofile_perf.o)
+endif
+
 oprofile-y                             := $(DRIVER_OBJS) common.o
index 72e09eb..8aa9744 100644 (file)
 #include <asm/ptrace.h>
 
 #ifdef CONFIG_HW_PERF_EVENTS
-/*
- * Per performance monitor configuration as set via oprofilefs.
- */
-struct op_counter_config {
-       unsigned long count;
-       unsigned long enabled;
-       unsigned long event;
-       unsigned long unit_mask;
-       unsigned long kernel;
-       unsigned long user;
-       struct perf_event_attr attr;
-};
-
-static int op_arm_enabled;
-static DEFINE_MUTEX(op_arm_mutex);
-
-static struct op_counter_config *counter_config;
-static struct perf_event **perf_events[nr_cpumask_bits];
-static int perf_num_counters;
-
-/*
- * Overflow callback for oprofile.
- */
-static void op_overflow_handler(struct perf_event *event, int unused,
-                       struct perf_sample_data *data, struct pt_regs *regs)
+char *op_name_from_perf_id(void)
 {
-       int id;
-       u32 cpu = smp_processor_id();
-
-       for (id = 0; id < perf_num_counters; ++id)
-               if (perf_events[cpu][id] == event)
-                       break;
-
-       if (id != perf_num_counters)
-               oprofile_add_sample(regs, id);
-       else
-               pr_warning("oprofile: ignoring spurious overflow "
-                               "on cpu %u\n", cpu);
-}
-
-/*
- * Called by op_arm_setup to create perf attributes to mirror the oprofile
- * settings in counter_config. Attributes are created as `pinned' events and
- * so are permanently scheduled on the PMU.
- */
-static void op_perf_setup(void)
-{
-       int i;
-       u32 size = sizeof(struct perf_event_attr);
-       struct perf_event_attr *attr;
-
-       for (i = 0; i < perf_num_counters; ++i) {
-               attr = &counter_config[i].attr;
-               memset(attr, 0, size);
-               attr->type              = PERF_TYPE_RAW;
-               attr->size              = size;
-               attr->config            = counter_config[i].event;
-               attr->sample_period     = counter_config[i].count;
-               attr->pinned            = 1;
-       }
-}
-
-static int op_create_counter(int cpu, int event)
-{
-       int ret = 0;
-       struct perf_event *pevent;
-
-       if (!counter_config[event].enabled || (perf_events[cpu][event] != NULL))
-               return ret;
-
-       pevent = perf_event_create_kernel_counter(&counter_config[event].attr,
-                                                 cpu, -1,
-                                                 op_overflow_handler);
-
-       if (IS_ERR(pevent)) {
-               ret = PTR_ERR(pevent);
-       } else if (pevent->state != PERF_EVENT_STATE_ACTIVE) {
-               perf_event_release_kernel(pevent);
-               pr_warning("oprofile: failed to enable event %d "
-                               "on CPU %d\n", event, cpu);
-               ret = -EBUSY;
-       } else {
-               perf_events[cpu][event] = pevent;
-       }
-
-       return ret;
-}
+       enum arm_perf_pmu_ids id = armpmu_get_pmu_id();
 
-static void op_destroy_counter(int cpu, int event)
-{
-       struct perf_event *pevent = perf_events[cpu][event];
-
-       if (pevent) {
-               perf_event_release_kernel(pevent);
-               perf_events[cpu][event] = NULL;
-       }
-}
-
-/*
- * Called by op_arm_start to create active perf events based on the
- * perviously configured attributes.
- */
-static int op_perf_start(void)
-{
-       int cpu, event, ret = 0;
-
-       for_each_online_cpu(cpu) {
-               for (event = 0; event < perf_num_counters; ++event) {
-                       ret = op_create_counter(cpu, event);
-                       if (ret)
-                               goto out;
-               }
-       }
-
-out:
-       return ret;
-}
-
-/*
- * Called by op_arm_stop at the end of a profiling run.
- */
-static void op_perf_stop(void)
-{
-       int cpu, event;
-
-       for_each_online_cpu(cpu)
-               for (event = 0; event < perf_num_counters; ++event)
-                       op_destroy_counter(cpu, event);
-}
-
-
-static char *op_name_from_perf_id(enum arm_perf_pmu_ids id)
-{
        switch (id) {
        case ARM_PERF_PMU_ID_XSCALE1:
                return "arm/xscale1";
@@ -176,116 +47,6 @@ static char *op_name_from_perf_id(enum arm_perf_pmu_ids id)
        }
 }
 
-static int op_arm_create_files(struct super_block *sb, struct dentry *root)
-{
-       unsigned int i;
-
-       for (i = 0; i < perf_num_counters; i++) {
-               struct dentry *dir;
-               char buf[4];
-
-               snprintf(buf, sizeof buf, "%d", i);
-               dir = oprofilefs_mkdir(sb, root, buf);
-               oprofilefs_create_ulong(sb, dir, "enabled", &counter_config[i].enabled);
-               oprofilefs_create_ulong(sb, dir, "event", &counter_config[i].event);
-               oprofilefs_create_ulong(sb, dir, "count", &counter_config[i].count);
-               oprofilefs_create_ulong(sb, dir, "unit_mask", &counter_config[i].unit_mask);
-               oprofilefs_create_ulong(sb, dir, "kernel", &counter_config[i].kernel);
-               oprofilefs_create_ulong(sb, dir, "user", &counter_config[i].user);
-       }
-
-       return 0;
-}
-
-static int op_arm_setup(void)
-{
-       spin_lock(&oprofilefs_lock);
-       op_perf_setup();
-       spin_unlock(&oprofilefs_lock);
-       return 0;
-}
-
-static int op_arm_start(void)
-{
-       int ret = -EBUSY;
-
-       mutex_lock(&op_arm_mutex);
-       if (!op_arm_enabled) {
-               ret = 0;
-               op_perf_start();
-               op_arm_enabled = 1;
-       }
-       mutex_unlock(&op_arm_mutex);
-       return ret;
-}
-
-static void op_arm_stop(void)
-{
-       mutex_lock(&op_arm_mutex);
-       if (op_arm_enabled)
-               op_perf_stop();
-       op_arm_enabled = 0;
-       mutex_unlock(&op_arm_mutex);
-}
-
-#ifdef CONFIG_PM
-static int op_arm_suspend(struct platform_device *dev, pm_message_t state)
-{
-       mutex_lock(&op_arm_mutex);
-       if (op_arm_enabled)
-               op_perf_stop();
-       mutex_unlock(&op_arm_mutex);
-       return 0;
-}
-
-static int op_arm_resume(struct platform_device *dev)
-{
-       mutex_lock(&op_arm_mutex);
-       if (op_arm_enabled && op_perf_start())
-               op_arm_enabled = 0;
-       mutex_unlock(&op_arm_mutex);
-       return 0;
-}
-
-static struct platform_driver oprofile_driver = {
-       .driver         = {
-               .name           = "arm-oprofile",
-       },
-       .resume         = op_arm_resume,
-       .suspend        = op_arm_suspend,
-};
-
-static struct platform_device *oprofile_pdev;
-
-static int __init init_driverfs(void)
-{
-       int ret;
-
-       ret = platform_driver_register(&oprofile_driver);
-       if (ret)
-               goto out;
-
-       oprofile_pdev = platform_device_register_simple(
-                               oprofile_driver.driver.name, 0, NULL, 0);
-       if (IS_ERR(oprofile_pdev)) {
-               ret = PTR_ERR(oprofile_pdev);
-               platform_driver_unregister(&oprofile_driver);
-       }
-
-out:
-       return ret;
-}
-
-static void  exit_driverfs(void)
-{
-       platform_device_unregister(oprofile_pdev);
-       platform_driver_unregister(&oprofile_driver);
-}
-#else
-static int __init init_driverfs(void) { return 0; }
-#define exit_driverfs() do { } while (0)
-#endif /* CONFIG_PM */
-
 static int report_trace(struct stackframe *frame, void *d)
 {
        unsigned int *depth = d;
@@ -350,74 +111,14 @@ static void arm_backtrace(struct pt_regs * const regs, unsigned int depth)
 
 int __init oprofile_arch_init(struct oprofile_operations *ops)
 {
-       int cpu, ret = 0;
-
-       perf_num_counters = armpmu_get_max_events();
-
-       counter_config = kcalloc(perf_num_counters,
-                       sizeof(struct op_counter_config), GFP_KERNEL);
-
-       if (!counter_config) {
-               pr_info("oprofile: failed to allocate %d "
-                               "counters\n", perf_num_counters);
-               return -ENOMEM;
-       }
-
-       ret = init_driverfs();
-       if (ret) {
-               kfree(counter_config);
-               counter_config = NULL;
-               return ret;
-       }
-
-       for_each_possible_cpu(cpu) {
-               perf_events[cpu] = kcalloc(perf_num_counters,
-                               sizeof(struct perf_event *), GFP_KERNEL);
-               if (!perf_events[cpu]) {
-                       pr_info("oprofile: failed to allocate %d perf events "
-                                       "for cpu %d\n", perf_num_counters, cpu);
-                       while (--cpu >= 0)
-                               kfree(perf_events[cpu]);
-                       return -ENOMEM;
-               }
-       }
-
        ops->backtrace          = arm_backtrace;
-       ops->create_files       = op_arm_create_files;
-       ops->setup              = op_arm_setup;
-       ops->start              = op_arm_start;
-       ops->stop               = op_arm_stop;
-       ops->shutdown           = op_arm_stop;
-       ops->cpu_type           = op_name_from_perf_id(armpmu_get_pmu_id());
-
-       if (!ops->cpu_type)
-               ret = -ENODEV;
-       else
-               pr_info("oprofile: using %s\n", ops->cpu_type);
 
-       return ret;
+       return oprofile_perf_init(ops);
 }
 
-void oprofile_arch_exit(void)
+void __exit oprofile_arch_exit(void)
 {
-       int cpu, id;
-       struct perf_event *event;
-
-       if (*perf_events) {
-               for_each_possible_cpu(cpu) {
-                       for (id = 0; id < perf_num_counters; ++id) {
-                               event = perf_events[cpu][id];
-                               if (event != NULL)
-                                       perf_event_release_kernel(event);
-                       }
-                       kfree(perf_events[cpu]);
-               }
-       }
-
-       if (counter_config) {
-               kfree(counter_config);
-               exit_driverfs();
-       }
+       oprofile_perf_exit();
 }
 #else
 int __init oprofile_arch_init(struct oprofile_operations *ops)
@@ -425,5 +126,5 @@ int __init oprofile_arch_init(struct oprofile_operations *ops)
        pr_info("oprofile: hardware counters not available\n");
        return -ENODEV;
 }
-void oprofile_arch_exit(void) {}
+void __exit oprofile_arch_exit(void) {}
 #endif /* CONFIG_HW_PERF_EVENTS */
index a202a2c..6cd151b 100644 (file)
@@ -320,6 +320,7 @@ void flush_iotlb_page(struct iommu *obj, u32 da)
                if ((start <= da) && (da < start + bytes)) {
                        dev_dbg(obj->dev, "%s: %08x<=%08x(%x)\n",
                                __func__, start, da, bytes);
+                       iotlb_load_cr(obj, &cr);
                        iommu_write_reg(obj, 1, MMU_FLUSH_ENTRY);
                }
        }
index 04d9521..e8f2be2 100644 (file)
@@ -435,7 +435,6 @@ static int s3c_adc_suspend(struct platform_device *pdev, pm_message_t state)
 static int s3c_adc_resume(struct platform_device *pdev)
 {
        struct adc_device *adc = platform_get_drvdata(pdev);
-       unsigned long flags;
 
        clk_enable(adc->clk);
        enable_irq(adc->irq);
index 90a2051..e8d20b0 100644 (file)
@@ -48,6 +48,9 @@
 #include <plat/clock.h>
 #include <plat/cpu.h>
 
+#include <linux/serial_core.h>
+#include <plat/regs-serial.h> /* for s3c24xx_uart_devs */
+
 /* clock information */
 
 static LIST_HEAD(clocks);
@@ -65,6 +68,28 @@ static int clk_null_enable(struct clk *clk, int enable)
        return 0;
 }
 
+static int dev_is_s3c_uart(struct device *dev)
+{
+       struct platform_device **pdev = s3c24xx_uart_devs;
+       int i;
+       for (i = 0; i < ARRAY_SIZE(s3c24xx_uart_devs); i++, pdev++)
+               if (*pdev && dev == &(*pdev)->dev)
+                       return 1;
+       return 0;
+}
+
+/*
+ * Serial drivers call get_clock() very early, before platform bus
+ * has been set up, this requires a special check to let them get
+ * a proper clock
+ */
+
+static int dev_is_platform_device(struct device *dev)
+{
+       return dev->bus == &platform_bus_type ||
+              (dev->bus == NULL && dev_is_s3c_uart(dev));
+}
+
 /* Clock API calls */
 
 struct clk *clk_get(struct device *dev, const char *id)
@@ -73,7 +98,7 @@ struct clk *clk_get(struct device *dev, const char *id)
        struct clk *clk = ERR_PTR(-ENOENT);
        int idno;
 
-       if (dev == NULL || dev->bus != &platform_bus_type)
+       if (dev == NULL || !dev_is_platform_device(dev))
                idno = -1;
        else
                idno = to_platform_device(dev)->id;
index 16399bd..0f2417d 100644 (file)
@@ -7,6 +7,7 @@ config FRV
        default y
        select HAVE_IDE
        select HAVE_ARCH_TRACEHOOK
+       select HAVE_IRQ_WORK
        select HAVE_PERF_EVENTS
 
 config ZONE_DMA
index f470975..4ff2fb1 100644 (file)
@@ -5,4 +5,4 @@
 lib-y := \
        __ashldi3.o __lshrdi3.o __muldi3.o __ashrdi3.o __negdi2.o __ucmpdi2.o \
        checksum.o memcpy.o memset.o atomic-ops.o atomic64-ops.o \
-       outsl_ns.o outsl_sw.o insl_ns.o insl_sw.o cache.o perf_event.o
+       outsl_ns.o outsl_sw.o insl_ns.o insl_sw.o cache.o
diff --git a/arch/frv/lib/perf_event.c b/arch/frv/lib/perf_event.c
deleted file mode 100644 (file)
index 9ac5acf..0000000
+++ /dev/null
@@ -1,19 +0,0 @@
-/* Performance event handling
- *
- * Copyright (C) 2009 Red Hat, Inc. All Rights Reserved.
- * Written by David Howells (dhowells@redhat.com)
- *
- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public Licence
- * as published by the Free Software Foundation; either version
- * 2 of the Licence, or (at your option) any later version.
- */
-
-#include <linux/perf_event.h>
-
-/*
- * mark the performance event as pending
- */
-void set_perf_event_pending(void)
-{
-}
index d514cd9..8fb7d33 100644 (file)
@@ -6,12 +6,6 @@
  *     David Mosberger-Tang <davidm@hpl.hp.com>
  */
 
-
-#include <linux/threads.h>
-#include <linux/irq.h>
-
-#include <asm/processor.h>
-
 /*
  * No irq_cpustat_t for IA-64.  The data is held in the per-CPU data structure.
  */
 
 #define local_softirq_pending()                (local_cpu_data->softirq_pending)
 
+#include <linux/threads.h>
+#include <linux/irq.h>
+
+#include <asm/processor.h>
+
 extern void __iomem *ipi_base_addr;
 
 void ack_bad_irq(unsigned int irq);
index 9f342a5..dd028f2 100644 (file)
@@ -272,10 +272,6 @@ void cpu_idle_wait(void);
 
 void default_idle(void);
 
-#ifdef CONFIG_VIRT_CPU_ACCOUNTING
-extern void account_system_vtime(struct task_struct *);
-#endif
-
 #endif /* __KERNEL__ */
 
 #endif /* __ASSEMBLY__ */
index 2f85412..b8da7d0 100644 (file)
@@ -82,9 +82,9 @@ typedef elf_fpreg_t elf_fpregset_t;
  * These are used to set parameters in the core dumps.
  */
 #define ELF_CLASS      ELFCLASS32
-#if defined(__LITTLE_ENDIAN)
+#if defined(__LITTLE_ENDIAN__)
 #define ELF_DATA       ELFDATA2LSB
-#elif defined(__BIG_ENDIAN)
+#elif defined(__BIG_ENDIAN__)
 #define ELF_DATA       ELFDATA2MSB
 #else
 #error no endian defined
diff --git a/arch/m32r/kernel/.gitignore b/arch/m32r/kernel/.gitignore
new file mode 100644 (file)
index 0000000..c5f676c
--- /dev/null
@@ -0,0 +1 @@
+vmlinux.lds
index 7bbe386..a08697f 100644 (file)
@@ -28,6 +28,8 @@
 
 #define DEBUG_SIG 0
 
+#define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP)))
+
 asmlinkage int
 sys_sigaltstack(const stack_t __user *uss, stack_t __user *uoss,
                unsigned long r2, unsigned long r3, unsigned long r4,
@@ -254,7 +256,7 @@ give_sigsegv:
 static int prev_insn(struct pt_regs *regs)
 {
        u16 inst;
-       if (get_user(&inst, (u16 __user *)(regs->bpc - 2)))
+       if (get_user(inst, (u16 __user *)(regs->bpc - 2)))
                return -EFAULT;
        if ((inst & 0xfff0) == 0x10f0)  /* trap ? */
                regs->bpc -= 2;
index e322d65..7dd65cf 100644 (file)
@@ -7,6 +7,10 @@ subdir-ccflags-y := -Werror
 include arch/mips/Kbuild.platforms
 obj-y := $(platform-y)
 
+# make clean traverses $(obj-) without having included .config, so
+# everything ends up here
+obj- := $(platform-)
+
 # mips object files
 # The object files are linked as core-y files would be linked
 
index 5526faa..4c9f402 100644 (file)
@@ -881,11 +881,15 @@ config NO_IOPORT
 config GENERIC_ISA_DMA
        bool
        select ZONE_DMA if GENERIC_ISA_DMA_SUPPORT_BROKEN=n
+       select ISA_DMA_API
 
 config GENERIC_ISA_DMA_SUPPORT_BROKEN
        bool
        select GENERIC_ISA_DMA
 
+config ISA_DMA_API
+       bool
+
 config GENERIC_GPIO
        bool
 
index 5fd7f7a..5042d51 100644 (file)
@@ -105,4 +105,4 @@ OBJCOPYFLAGS_vmlinuz.srec := $(OBJCOPYFLAGS) -S -O srec
 vmlinuz.srec: vmlinuz
        $(call cmd,objcopy)
 
-clean-files := $(objtree)/vmlinuz.*
+clean-files := $(objtree)/vmlinuz $(objtree)/vmlinuz.{32,ecoff,bin,srec}
index 3adbcbd..cf55a6f 100644 (file)
@@ -1,7 +1,7 @@
 #
 # DECstation family
 #
-platform-$(CONFIG_MACH_DECSTATION)     = dec/
+platform-$(CONFIG_MACH_DECSTATION)     += dec/
 cflags-$(CONFIG_MACH_DECSTATION)       += \
                        -I$(srctree)/arch/mips/include/asm/mach-dec
 libs-$(CONFIG_MACH_DECSTATION)         += arch/mips/dec/prom/
index e482fe9..75edded 100644 (file)
@@ -56,6 +56,7 @@
  */
 
 #ifdef CONFIG_32BIT
+#include <linux/types.h>
 
 struct flock {
        short   l_type;
index 96e28f1..1ca64b4 100644 (file)
@@ -88,6 +88,7 @@ typedef struct siginfo {
 #ifdef __ARCH_SI_TRAPNO
                        int _trapno;    /* TRAP # which caused the signal */
 #endif
+                       short _addr_lsb;
                } _sigfault;
 
                /* SIGPOLL, SIGXFSZ (To do ...)  */
index 6a97230..ba91be9 100644 (file)
@@ -1,3 +1,3 @@
-core-$(CONFIG_MACH_JZ4740)     += arch/mips/jz4740/
+platform-$(CONFIG_MACH_JZ4740) += jz4740/
 cflags-$(CONFIG_MACH_JZ4740)   += -I$(srctree)/arch/mips/include/asm/mach-jz4740
 load-$(CONFIG_MACH_JZ4740)     += 0xffffffff80010000
index 0176ed0..32103cc 100644 (file)
@@ -40,7 +40,6 @@ int __compute_return_epc(struct pt_regs *regs)
                return -EFAULT;
        }
 
-       regs->regs[0] = 0;
        switch (insn.i_format.opcode) {
        /*
         * jr and jalr are in r_format format.
index 2340f11..9a526ba 100644 (file)
@@ -103,7 +103,7 @@ asmlinkage long mipsmt_sys_sched_setaffinity(pid_t pid, unsigned int len,
        if (!check_same_owner(p) && !capable(CAP_SYS_NICE))
                goto out_unlock;
 
-       retval = security_task_setscheduler(p, 0, NULL);
+       retval = security_task_setscheduler(p)
        if (retval)
                goto out_unlock;
 
index c51b95f..c877733 100644 (file)
@@ -536,7 +536,7 @@ asmlinkage void do_syscall_trace(struct pt_regs *regs, int entryexit)
 {
        /* do the secure computing check first */
        if (!entryexit)
-               secure_computing(regs->regs[0]);
+               secure_computing(regs->regs[2]);
 
        if (unlikely(current->audit_context) && entryexit)
                audit_syscall_exit(AUDITSC_RESULT(regs->regs[2]),
@@ -565,7 +565,7 @@ asmlinkage void do_syscall_trace(struct pt_regs *regs, int entryexit)
 
 out:
        if (unlikely(current->audit_context) && !entryexit)
-               audit_syscall_entry(audit_arch(), regs->regs[0],
+               audit_syscall_entry(audit_arch(), regs->regs[2],
                                    regs->regs[4], regs->regs[5],
            &nbs