Found by Simon Trimmer <simon@veritas.com>: shmem_file_write
failure path duplicates kunmap, causing oops holding kmap_lock.

Trivial: vfs_link in 2.5 checks S_ISDIR first, shmem_link
need not repeat it, but test crept back in at some stage.

Martin Schwidefsky <schwidefsky@de.ibm.com> reported "Bug with shared
memory" to LKML 14 May: hang due to schedule in truncate_list_pages
called from .... shm_destroy holding shm_lock spinlock.  shm_destroy
needs that lock for shm_rmid, but it can be safely unlocked once link
from id to shp has been removed.

Here is a quick fix.  I would like to synchronize with the irq handler
changes as well.  Becouse right now I know that preemption is killing
the disk subsystem when moving data between disks using different
request queues...  In esp.  It get's me in to do_request() with a queue
in unplugged state.  (Not everything is my fault, after all :-).

 - init thread needs to have preempt_count of 1 until sched_init().
   (William Lee Irwin III)
 - clean up the irq-mask macros. (Linus)
 - add barrier() to irq_enter() and irq_exit(). (based on Oleg Nesterov's
   comment.)
 - move the irqs-off check into preempt_schedule() and remove
   CONFIG_DEBUG_IRQ_SCHEDULE.
 - remove spin_unlock_no_resched() and comment the affected places more
   agressively.
 - slab.c needs to spin_unlock_no_resched(), instead of spin_unlock(). (It
   also has to check for preemption in the right spot.) This should fix
   the memory corruption.
 - irq_exit() needs to run softirqs if interrupts not active - in the
   previous patch it ran them when preempt_count() was 0, which is
   incorrect.
 - spinlock macros are updated to enable preemption after enabling
   interrupts. Besides avoiding false positive warnings, this also
 - fork.c has to call scheduler_tick() with preemption disabled -
   otherwise scheduler_tick()'s spin_unlock can preempt!
 - irqs_disabled() macro introduced.
 - [ all other local_irq_enable() or sti instances conditional on
     CONFIG_DEBUG_IRQ_SCHEDULE are to fix false positive warnings. ]
 - fix buggy in_softirq(). Fortunately the bug made the test broader,
   which didnt result in algorithmical breakage, just suboptimal
   performance.
 - move do_softirq() processing into irq_exit() => this also fixes the
   softirq processing bugs present in apic.c IRQ handlers that did not
   test for softirqs after irq_exit().
 - simplify local_bh_enable().

Amiga Zorro bus updates
 - Bring APUS Zorro bus announcement in sync with the m68k version
 - Fix typo in Zorro device generator name
 - Kill duplicate ] in Zorro device ID database
 - Include <asm/io.h> instead of duplicating definitions from <asm/io.h>

console.wait_key was removed

M68k VT updates:
  - Include <linux/vt_kern.h> instead of copying the kd_mksound() prototype
  - Lots of stuff depends on CONFIG_VT:
      o SYSRQ_KEY
      o kd_mksound()
      o mach_keyb_init()
      o mach_kbdrate()
      o mach_kbd_leds()
      o mach_kbd_translate()
      o mach_sysrq_xlate()
      o Atari keyboard and joystick drivers
  - There's no need to provide dummy routines for kbd{rate,_leds}()
  - dummy_con depends on CONFIG_DUMMY_CONSOLE

MVME serial driver updates (from 2.4.x)
  - Add missing save of interrupt state
  - _tty_name() got renamed to tty_name()
  - Add devfs support
  - Simplify baud logic
  - Add missing MOD_DEC_USE_COUNT
  - Add missing cli()
  - block_til_ready() got renamed to gs_block_til_ready()

MVME SCSI updates
 - Fix return type in prototype of ncr53c7xx_init()

M68k fault handling updates
  - Fix incorrect page alignment
  - Merge related inline asm statements to avoid .chip confusion
  - Add Sun-3 bus mapping test code
  - Kill warning

M68k TLB updates

M68k task and thread updates

Add new syscalls to m68k
  - Add syscall entries for sys_pivot_root, extended attributes, and futex
  - Fix formatting

Sun-3/3x frame buffer device updates
  - Fix breakage from recent fbdev changes
  - Add support for CG3 graphics on Sun-3/3x

Atari STRAM fixes
  - Add missing include
  - Fix parameter passing to atari_stram_reserve_pages()

Spelling fix

SCSI scatterlist updates for m68k
  - Fix printk() format
  - Update for page/offset in scatterlist

Q40/Q60 keyboard updates
  - Ignore illegal scancodes
  - Remove unused q40kbd_leds()
  - Fix return type and return value of q40kbd_init_hw()

Remove CPU information from /proc/hardware (already in /proc/cpuinfo)

Add PCI_DMA_BUS_IS_PHYS for m68k

M68k parport updates
 - Remove workaround code for parport bug (parport without FIFO)
 - ->dma_buf is used in FIFO mode only
 - Add missing prototype for parport_atari_init()
 - Initialize uninitialized variable

Mac/m68k Nubus updates
  - Use nubus_{read,write}[bwl]()
  - Fix /proc/bus/nubus
  - Fix type and linkage of nubus_init()
  - Use nubus_{read,write}[bwl]()
  - Kill some address assignment warnings
  - Include <asm/nubus.h>
  - nubus_init() is an initcall, hence static

Add definition for PROT_SEM on m68k

M68k core mm updates
  - Update fault handling
  - Remove superfluous flush_tlb_page()
  - Remove obsolete comment
  - empty_zero_page becomes a void *
  - Type and cast clean ups
  - Remove duplicated test for voff == 0
  - Move check_pgt_cache() to pgtable.h
  - Pte and pmd updates
  - Add additional parameter to {clear,copy}_user_page()
  - pfn updates
  - BUG() updates
  - Move check_pgt_cache() from motorola_pgalloc.h
  - Misc updates
  - Add __page_address() and page_to_phys()

Add Mac/m68k support to valkyriefb

Mac/m68k doesn't use the PowerMac NVRAM driver

Mac m68k Ethernet driver updates (from 2.4.x)
 - mace68k_probe() got renamed to mace_probe()
 - Add missing MODULE_LICENSE()
 - Add missing register definition for SONIC_DCR2

Misc Mac/m68k updates
 - Make sure we have one instance of mac_hw_present
 - Kill warning
 - Rename struct SCC to struct mac_SCC to avoid type conflict

Update ISA space code on m68k (from 2.4.x)
 - Add support for multiple types of ISA busses in one kernel image (e.g. an
   image for both Q40 and Amiga with PCMCIA)
 - Allow port operations on m68k if CONFIG_ISA is defined

M68k I/O core updates
 - Use memory mapped I/O access routines for non-ISA/PCI/PCMCIA m68k devices
 - Don't use writel() and friends on m68k memory mapped I/O
 - Fix typo in definition of writew()
 - Add missing definitions for {in,out}_le{16,32}()

The old form of designated initializers are obsolete: we need to replace them
with the ISO C forms before 2.6.  Gcc has always supported both forms anyway.

(from Rusty Russell <rusty@rustcorp.com.au>, with some mods by me)

M68k USER_HZ updates

HP300 LANCE driver updates
  - Update DIO probe code
  - Use out_8() instead of writeb() for non-PCI memory mapped I/O
  - Kill warnings

Merge remaining bits of core HP300 support

Fix Mac build if no fonts are selected

Add Sun-3x floppy support (from 2.4.x)