* [PATCH -mm 1/5] Blackfin: blackfin architecture patch update
@ 2007-03-01 4:14 Wu, Bryan
2007-03-03 20:38 ` Arnd Bergmann
` (2 more replies)
0 siblings, 3 replies; 24+ messages in thread
From: Wu, Bryan @ 2007-03-01 4:14 UTC (permalink / raw)
To: Andrew Morton, linux-kernel
Hi folks,
Here is the update version of blackfin-arch.patch in -mm tree.
simply add support to utrace and it was tested on blackfin STAMP board
as well as other following patches.
The whole patch is located at URL:
https://blackfin.uclinux.org/gf/download/frsrelease/39/2583/blackfin-arch.patch
The incremental patch is located at URL:
https://blackfin.uclinux.org/gf/download/frsrelease/39/2584/blackfin-arch-mm2-update.patch
[PATCH] Blackfin Architecture
This adds support for the Analog Devices Blackfin processor
architecture, and currently supports the BF533, BF532, BF531, BF537,
BF536, BF534, and BF561 (Dual Core) devices, with a variety of
development platforms including those avaliable from Analog Devices
(BF533-EZKit, BF533-STAMP, BF537-STAMP, BF561-EZKIT), and Bluetechnix!
Tinyboards.
The Blackfin architecture was jointly developed by Intel and Analog
Devices Inc. (ADI) as the Micro Signal Architecture (MSA) core and
introduced it in December of 2000. Since then ADI has put this core into
it’s Blackfin processor family of devices. The Blackfin core has the
advantages of a clean, orthogonal,RISC-like microprocessor instruction
set. It combines a dual‑MAC (Multiply/Accumulate), state‑of‑the‑art
signal processing engine and single-instruction, multiple‑data (SIMD)
multimedia capabilities into a single instruction-set architecture.
The Blackfin architecture, including the instruction set, is described
by the ADSP-BF53x/BF56x Blackfin® Processor Programming Reference
http://blackfin.uclinux.org/gf/download/frsrelease/29/2549/Blackfin_PRM.pdf
The Blackfin processor is already supported by major releases of gcc,
and there are binary and source rpms/tarballs for many architectures at:
http://blackfin.uclinux.org/gf/project/toolchain/frs
There is complete documentation, including "getting started" guides
available at:
http://docs.blackfin.uclinux.org/
which provides links to the sources and patches you will need in order
to set up a cross-compiling environment for bfin-linux-uclibc
This patch, as well as the other patches (toolchain, distribution,
uClibc) are actively supported by Analog Devices Inc, at:
http://blackfin.uclinux.org/
We have tested this on LTP, and our test plan (including pass/fails) can
be found at:
http://docs.blackfin.uclinux.org/doku.php?id=testing_the_linux_kernel
Signed-off-by: Bryan Wu <bryan.wu@analog.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---
arch/blackfin/Kconfig | 989 ++++++++
arch/blackfin/Makefile | 81
arch/blackfin/boot/Makefile | 27
arch/blackfin/defconfig | 1314 +++++++++++
arch/blackfin/kernel/Makefile | 13
arch/blackfin/kernel/asm-offsets.c | 140 +
arch/blackfin/kernel/bfin_dma_5xx.c | 747 ++++++
arch/blackfin/kernel/bfin_gpio.c | 654 +++++
arch/blackfin/kernel/bfin_ksyms.c | 119
arch/blackfin/kernel/dma-mapping.c | 174 +
arch/blackfin/kernel/dualcore_test.c | 51
arch/blackfin/kernel/entry.S | 96
arch/blackfin/kernel/init_task.c | 63
arch/blackfin/kernel/irqchip.c | 149 +
arch/blackfin/kernel/module.c | 431 +++
arch/blackfin/kernel/process.c | 398 +++
arch/blackfin/kernel/ptrace.c | 421 +++
arch/blackfin/kernel/setup.c | 921 +++++++
arch/blackfin/kernel/signal.c | 436 +++
arch/blackfin/kernel/sys_bfin.c | 133 +
arch/blackfin/kernel/time.c | 330 ++
arch/blackfin/kernel/traps.c | 666 +++++
arch/blackfin/kernel/vmlinux.lds.S | 221 +
arch/blackfin/lib/Makefile | 11
arch/blackfin/lib/ashldi3.c | 60
arch/blackfin/lib/ashrdi3.c | 61
arch/blackfin/lib/checksum.c | 142 +
arch/blackfin/lib/divsi3.S | 217 +
arch/blackfin/lib/gcclib.h | 49
arch/blackfin/lib/ins.S | 71
arch/blackfin/lib/lshrdi3.c | 74
arch/blackfin/lib/memchr.S | 65
arch/blackfin/lib/memcmp.S | 110
arch/blackfin/lib/memcpy.S | 135 +
arch/blackfin/lib/memmove.S | 103
arch/blackfin/lib/memset.S | 109
arch/blackfin/lib/modsi3.S | 81
arch/blackfin/lib/muldi3.c | 101
arch/blackfin/lib/outs.S | 64
arch/blackfin/lib/smulsi3_highpart.S | 30
arch/blackfin/lib/strcmp.c | 11
arch/blackfin/lib/strcpy.c | 11
arch/blackfin/lib/strncmp.c | 11
arch/blackfin/lib/strncpy.c | 11
arch/blackfin/lib/udivsi3.S | 300 ++
arch/blackfin/lib/umodsi3.S | 68
arch/blackfin/lib/umulsi3_highpart.S | 23
arch/blackfin/mach-bf533/Kconfig | 92
arch/blackfin/mach-bf533/Makefile | 9
arch/blackfin/mach-bf533/boards/Makefile | 8
arch/blackfin/mach-bf533/boards/cm_bf533.c | 271 ++
arch/blackfin/mach-bf533/boards/ezkit.c | 231 +
arch/blackfin/mach-bf533/boards/generic_board.c | 97
arch/blackfin/mach-bf533/boards/stamp.c | 328 ++
arch/blackfin/mach-bf533/cpu.c | 163 +
arch/blackfin/mach-bf533/head.S | 776 ++++++
arch/blackfin/mach-bf533/ints-priority.c | 67
arch/blackfin/mach-bf537/Kconfig | 141 +
arch/blackfin/mach-bf537/Makefile | 9
arch/blackfin/mach-bf537/boards/Makefile | 8
arch/blackfin/mach-bf537/boards/cm_bf537.c | 379 +++
arch/blackfin/mach-bf537/boards/generic_board.c | 464 +++
arch/blackfin/mach-bf537/boards/led.S | 183 +
arch/blackfin/mach-bf537/boards/pnav10.c | 541 ++++
arch/blackfin/mach-bf537/boards/stamp.c | 624 +++++
arch/blackfin/mach-bf537/cpu.c | 163 +
arch/blackfin/mach-bf537/head.S | 604 +++++
arch/blackfin/mach-bf537/ints-priority.c | 76
arch/blackfin/mach-bf561/Kconfig | 222 +
arch/blackfin/mach-bf561/Makefile | 9
arch/blackfin/mach-bf561/boards/Makefile | 6
arch/blackfin/mach-bf561/boards/cm_bf561.c | 295 ++
arch/blackfin/mach-bf561/boards/ezkit.c | 151 +
arch/blackfin/mach-bf561/boards/generic_board.c | 84
arch/blackfin/mach-bf561/coreb.c | 404 +++
arch/blackfin/mach-bf561/head.S | 514 ++++
arch/blackfin/mach-bf561/ints-priority.c | 110
arch/blackfin/mach-common/Makefile | 12
arch/blackfin/mach-common/cache.S | 255 ++
arch/blackfin/mach-common/cacheinit.S | 139 +
arch/blackfin/mach-common/cplbhdlr.S | 132 +
arch/blackfin/mach-common/cplbinfo.c | 213 +
arch/blackfin/mach-common/cplbmgr.S | 609 +++++
arch/blackfin/mach-common/dpmc.S | 420 +++
arch/blackfin/mach-common/entry.S | 1209 ++++++++++
arch/blackfin/mach-common/interrupt.S | 255 ++
arch/blackfin/mach-common/ints-priority-dc.c | 473 +++
arch/blackfin/mach-common/ints-priority-sc.c | 580 ++++
arch/blackfin/mach-common/irqpanic.c | 196 +
arch/blackfin/mach-common/lock.S | 206 +
arch/blackfin/mach-common/pm.c | 183 +
arch/blackfin/mm/Makefile | 5
arch/blackfin/mm/blackfin_sram.c | 537 ++++
arch/blackfin/mm/blackfin_sram.h | 40
arch/blackfin/mm/init.c | 211 +
arch/blackfin/mm/kmap.c | 84
arch/blackfin/oprofile/Kconfig | 29
arch/blackfin/oprofile/Makefile | 14
arch/blackfin/oprofile/common.c | 170 +
arch/blackfin/oprofile/op_blackfin.h | 100
arch/blackfin/oprofile/op_model_bf533.c | 163 +
arch/blackfin/oprofile/timer_int.c | 76
fs/Kconfig.binfmt | 2
include/asm-blackfin/Kbuild | 1
include/asm-blackfin/a.out.h | 25
include/asm-blackfin/atomic.h | 144 +
include/asm-blackfin/auxvec.h | 4
include/asm-blackfin/bf5xx_timers.h | 209 +
include/asm-blackfin/bfin-global.h | 120 +
include/asm-blackfin/bfin5xx_spi.h | 170 +
include/asm-blackfin/bfin_simple_timer.h | 13
include/asm-blackfin/bfin_spi_channel.h | 182 +
include/asm-blackfin/bfin_sport.h | 177 +
include/asm-blackfin/bitops.h | 213 +
include/asm-blackfin/blackfin.h | 56
include/asm-blackfin/bug.h | 15
include/asm-blackfin/bugs.h | 16
include/asm-blackfin/byteorder.h | 48
include/asm-blackfin/cache.h | 18
include/asm-blackfin/cacheflush.h | 91
include/asm-blackfin/checksum.h | 101
include/asm-blackfin/cplb.h | 51
include/asm-blackfin/cplbinit.h | 205 +
include/asm-blackfin/cpumask.h | 6
include/asm-blackfin/cputime.h | 6
include/asm-blackfin/current.h | 23
include/asm-blackfin/delay.h | 44
include/asm-blackfin/device.h | 7
include/asm-blackfin/div64.h | 1
include/asm-blackfin/dma-mapping.h | 73
include/asm-blackfin/dma.h | 203 +
include/asm-blackfin/dpmc.h | 66
include/asm-blackfin/elf.h | 127 +
include/asm-blackfin/emergency-restart.h | 6
include/asm-blackfin/entry.h | 61
include/asm-blackfin/errno.h | 6
include/asm-blackfin/fcntl.h | 13
include/asm-blackfin/flat.h | 130 +
include/asm-blackfin/futex.h | 6
include/asm-blackfin/gpio.h | 369 +++
include/asm-blackfin/hardirq.h | 41
include/asm-blackfin/hw_irq.h | 6
include/asm-blackfin/ide.h | 32
include/asm-blackfin/io.h | 152 +
include/asm-blackfin/ioctl.h | 1
include/asm-blackfin/ioctls.h | 82
include/asm-blackfin/ipc.h | 1
include/asm-blackfin/ipcbuf.h | 30
include/asm-blackfin/irq.h | 72
include/asm-blackfin/irq_regs.h | 1
include/asm-blackfin/kmap_types.h | 21
include/asm-blackfin/l1layout.h | 31
include/asm-blackfin/linkage.h | 7
include/asm-blackfin/local.h | 6
include/asm-blackfin/mach-bf533/anomaly.h | 176 +
include/asm-blackfin/mach-bf533/bf533.h | 308 ++
include/asm-blackfin/mach-bf533/bfin_serial_5xx.h | 108
include/asm-blackfin/mach-bf533/blackfin.h | 46
include/asm-blackfin/mach-bf533/cdefBF532.h | 707 +++++
include/asm-blackfin/mach-bf533/defBF532.h | 1178 +++++++++
include/asm-blackfin/mach-bf533/dma.h | 56
include/asm-blackfin/mach-bf533/irq.h | 178 +
include/asm-blackfin/mach-bf533/mem_init.h | 316 ++
include/asm-blackfin/mach-bf533/mem_map.h | 169 +
include/asm-blackfin/mach-bf535/bf535.h | 1277 ++++++++++
include/asm-blackfin/mach-bf535/bf535_serial.h | 109
include/asm-blackfin/mach-bf535/blackfin.h | 43
include/asm-blackfin/mach-bf535/cdefBF535.h | 121 +
include/asm-blackfin/mach-bf535/cdefblackfin.h | 69
include/asm-blackfin/mach-bf535/defBF535.h | 1818 +++++++++++++++
include/asm-blackfin/mach-bf535/defblackfin.h | 444 +++
include/asm-blackfin/mach-bf535/irq.h | 125 +
include/asm-blackfin/mach-bf537/anomaly.h | 120 +
include/asm-blackfin/mach-bf537/bf537.h | 289 ++
include/asm-blackfin/mach-bf537/bfin_serial_5xx.h | 147 +
include/asm-blackfin/mach-bf537/blackfin.h | 430 +++
include/asm-blackfin/mach-bf537/cdefBF534.h | 1823 +++++++++++++++
include/asm-blackfin/mach-bf537/cdefBF537.h | 209 +
include/asm-blackfin/mach-bf537/defBF534.h | 2502 +++++++++++++++++++++
include/asm-blackfin/mach-bf537/defBF537.h | 404 +++
include/asm-blackfin/mach-bf537/dma.h | 55
include/asm-blackfin/mach-bf537/irq.h | 219 +
include/asm-blackfin/mach-bf537/mem_init.h | 330 ++
include/asm-blackfin/mach-bf537/mem_map.h | 182 +
include/asm-blackfin/mach-bf561/anomaly.h | 185 +
include/asm-blackfin/mach-bf561/bf561.h | 410 +++
include/asm-blackfin/mach-bf561/bfin_serial_5xx.h | 108
include/asm-blackfin/mach-bf561/blackfin.h | 53
include/asm-blackfin/mach-bf561/cdefBF561.h | 1544 ++++++++++++
include/asm-blackfin/mach-bf561/defBF561.h | 1718 ++++++++++++++
include/asm-blackfin/mach-bf561/dma.h | 36
include/asm-blackfin/mach-bf561/irq.h | 451 +++
include/asm-blackfin/mach-bf561/mem_init.h | 322 ++
include/asm-blackfin/mach-bf561/mem_map.h | 77
include/asm-blackfin/mach-common/cdef_LPBlackfin.h | 474 +++
include/asm-blackfin/mach-common/context.S | 350 ++
include/asm-blackfin/mach-common/def_LPBlackfin.h | 694 +++++
include/asm-blackfin/macros.h | 95
include/asm-blackfin/mem_map.h | 12
include/asm-blackfin/mman.h | 45
include/asm-blackfin/mmu.h | 30
include/asm-blackfin/mmu_context.h | 131 +
include/asm-blackfin/module.h | 19
include/asm-blackfin/msgbuf.h | 31
include/asm-blackfin/mutex.h | 9
include/asm-blackfin/namei.h | 19
include/asm-blackfin/page.h | 89
include/asm-blackfin/page_offset.h | 6
include/asm-blackfin/param.h | 22
include/asm-blackfin/pci.h | 148 +
include/asm-blackfin/percpu.h | 6
include/asm-blackfin/pgalloc.h | 8
include/asm-blackfin/pgtable.h | 70
include/asm-blackfin/poll.h | 24
include/asm-blackfin/posix_types.h | 65
include/asm-blackfin/processor.h | 105
include/asm-blackfin/ptrace.h | 166 +
include/asm-blackfin/resource.h | 6
include/asm-blackfin/scatterlist.h | 26
include/asm-blackfin/sections.h | 7
include/asm-blackfin/segment.h | 7
include/asm-blackfin/semaphore-helper.h | 82
include/asm-blackfin/semaphore.h | 106
include/asm-blackfin/sembuf.h | 25
include/asm-blackfin/setup.h | 17
include/asm-blackfin/shmbuf.h | 42
include/asm-blackfin/shmparam.h | 6
include/asm-blackfin/sigcontext.h | 50
include/asm-blackfin/siginfo.h | 35
include/asm-blackfin/signal.h | 160 +
include/asm-blackfin/socket.h | 52
include/asm-blackfin/sockios.h | 12
include/asm-blackfin/spinlock.h | 6
include/asm-blackfin/stat.h | 63
include/asm-blackfin/statfs.h | 6
include/asm-blackfin/string.h | 104
include/asm-blackfin/system.h | 253 ++
include/asm-blackfin/termbits.h | 184 +
include/asm-blackfin/termios.h | 106
include/asm-blackfin/thread_info.h | 146 +
include/asm-blackfin/timex.h | 18
include/asm-blackfin/tlb.h | 16
include/asm-blackfin/tlbflush.h | 62
include/asm-blackfin/topology.h | 6
include/asm-blackfin/tracehook.h | 79
include/asm-blackfin/traps.h | 75
include/asm-blackfin/types.h | 66
include/asm-blackfin/uaccess.h | 254 ++
include/asm-blackfin/ucontext.h | 30
include/asm-blackfin/unaligned.h | 6
include/asm-blackfin/unistd.h | 382 +++
include/asm-blackfin/user.h | 89
include/asm-blackfin/xor.h | 1
include/linux/elf-em.h | 1
include/linux/spi/ad7877.h | 24
include/linux/usb_sl811.h | 26
init/Kconfig | 2
lib/Kconfig.debug | 2
scripts/genksyms/genksyms.c | 3
scripts/mod/mk_elfconfig.c | 3
260 files changed, 51444 insertions(+), 5 deletions(-)
https://blackfin.uclinux.org/gf/download/frsrelease/39/2583/blackfin-arch.patch
---
Thanks
-Bryan Wu
^ permalink raw reply [flat|nested] 24+ messages in thread* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-01 4:14 [PATCH -mm 1/5] Blackfin: blackfin architecture patch update Wu, Bryan @ 2007-03-03 20:38 ` Arnd Bergmann 2007-03-05 7:13 ` Wu, Bryan 2007-03-03 22:30 ` Arnd Bergmann 2007-03-05 9:23 ` Paul Mundt 2 siblings, 1 reply; 24+ messages in thread From: Arnd Bergmann @ 2007-03-03 20:38 UTC (permalink / raw) To: bryan.wu; +Cc: Andrew Morton, linux-kernel On Thursday 01 March 2007 05:14:40 Wu, Bryan wrote: > The whole patch is located at URL: > https://blackfin.uclinux.org/gf/download/frsrelease/39/2583/blackfin-arch.p >atch The incremental patch is located at URL: > https://blackfin.uclinux.org/gf/download/frsrelease/39/2584/blackfin-arch-m >m2-update.patch I'm not sure if that was intentional, but the second patch does not apply on top of the -mm kernel but rather patch the the patch old itself. This basically makes it impossible to review just that part, so better provide the diff between the kernel with the old patch and the kernel with the new patch next time. OTOH, from what I could see from the contents, the changes themselves look pretty good, I'd probably add my 'Acked-by' if I could read that patch more easily ;-) Arnd <>< ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-03 20:38 ` Arnd Bergmann @ 2007-03-05 7:13 ` Wu, Bryan 0 siblings, 0 replies; 24+ messages in thread From: Wu, Bryan @ 2007-03-05 7:13 UTC (permalink / raw) To: Arnd Bergmann; +Cc: Andrew Morton, bert.hubert, linux-kernel On Sat, 2007-03-03 at 15:38 -0500, Arnd Bergmann wrote: > On Thursday 01 March 2007 05:14:40 Wu, Bryan wrote: > > The whole patch is located at URL: > > > https://blackfin.uclinux.org/gf/download/frsrelease/39/2583/blackfin-arch.p > >atch The incremental patch is located at URL: > > > https://blackfin.uclinux.org/gf/download/frsrelease/39/2584/blackfin-arch-m > >m2-update.patch > > I'm not sure if that was intentional, but the second patch does not > apply > on top of the -mm kernel but rather patch the the patch old itself. > This basically makes it impossible to review just that part, so > better > provide the diff between the kernel with the old patch and the kernel > with the new patch next time. > Sorry for messing up this incremental patch. Here is the updated incremental patch Signed-off-by: Bryan Wu <bryan.wu@analog.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> --- arch/blackfin/Kconfig | 122 ++++++- arch/blackfin/kernel/entry.S | 4 arch/blackfin/kernel/irqchip.c | 5 arch/blackfin/kernel/process.c | 48 ++ arch/blackfin/kernel/setup.c | 41 +- arch/blackfin/kernel/signal.c | 5 arch/blackfin/kernel/sys_bfin.c | 16 arch/blackfin/kernel/time.c | 6 arch/blackfin/kernel/traps.c | 4 arch/blackfin/lib/ashldi3.c | 4 arch/blackfin/lib/ashrdi3.c | 4 arch/blackfin/lib/checksum.c | 4 arch/blackfin/lib/divsi3.S | 8 arch/blackfin/lib/lshrdi3.c | 4 arch/blackfin/lib/memcpy.S | 6 arch/blackfin/lib/memset.S | 6 arch/blackfin/lib/modsi3.S | 7 arch/blackfin/lib/muldi3.c | 4 arch/blackfin/lib/smulsi3_highpart.S | 6 arch/blackfin/lib/udivsi3.S | 9 arch/blackfin/lib/umodsi3.S | 5 arch/blackfin/lib/umulsi3_highpart.S | 6 arch/blackfin/mach-bf537/boards/stamp.c | 49 ++ arch/blackfin/mach-bf537/head.S | 1 arch/blackfin/mach-bf561/boards/generic_board.c | 2 arch/blackfin/mach-common/cache.S | 2 arch/blackfin/mach-common/cplbhdlr.S | 4 arch/blackfin/mach-common/cplbmgr.S | 4 arch/blackfin/mach-common/entry.S | 17 - arch/blackfin/mach-common/ints-priority-dc.c | 4 arch/blackfin/mach-common/ints-priority-sc.c | 4 include/asm-blackfin/cplbinit.h | 397 ++++++++++++------------ include/asm-blackfin/delay.h | 5 include/asm-blackfin/dma-mapping.h | 5 include/asm-blackfin/mach-bf533/bf533.h | 11 include/asm-blackfin/mach-bf533/mem_map.h | 62 ++- include/asm-blackfin/mach-bf537/bf537.h | 11 include/asm-blackfin/mach-bf537/defBF534.h | 2 include/asm-blackfin/mach-bf537/mem_map.h | 77 +++- include/asm-blackfin/mach-bf561/bf561.h | 5 include/asm-blackfin/mach-bf561/mem_map.h | 28 + include/asm-blackfin/pgtable.h | 8 include/asm-blackfin/uaccess.h | 41 -- include/asm-blackfin/unistd.h | 2 include/asm-blackfin/xor.h | 1 init/Kconfig | 3 46 files changed, 725 insertions(+), 344 deletions(-) diff -uprN linux-2.6-orig/arch/blackfin/Kconfig linux-2.6/arch/blackfin/Kconfig --- linux-2.6-orig/arch/blackfin/Kconfig 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/Kconfig 2007-03-05 11:57:23.000000000 +0800 @@ -9,6 +9,10 @@ config MMU bool default n +config ZONE_DMA + bool + default y + config FPU bool default n @@ -21,6 +25,10 @@ config RWSEM_XCHGADD_ALGORITHM bool default n +config BLACKFIN + bool + default y + config BFIN bool default y @@ -140,21 +148,14 @@ config BF_REV_0_5 endchoice -config BLACKFIN +config BFIN_DUAL_CORE bool - depends on (BF537 || BF536 \ - || BF534 || BF561 || BF535 || BF533 || BF532 || BF531) + depends on (BF561) default y config BFIN_SINGLE_CORE bool - depends on (BF537 || BF536 \ - || BF534 || BF535 || BF533 || BF532 || BF531) - default y - -config BFIN_DUAL_CORE - bool - depends on (BF561) + depends on !BFIN_DUAL_CORE default y choice @@ -427,11 +428,104 @@ comment "Memory Optimizations" config I_ENTRY_L1 bool "Locate interrupt entry code in L1 Memory" - default n + default y help If enabled interrupt entry code (STORE/RESTORE CONTEXT) is linked into L1 instruction memory.(less latency) +config EXCPT_IRQ_SYSC_L1 + bool "Locate entire ASM lowlevel excepetion / interrupt - Syscall and CPLB handler code in L1 Memory" + default y + help + If enabled entire ASM lowlevel exception and interrupt entry code (STORE/RESTORE CONTEXT) is linked + into L1 instruction memory.(less latency) + +config DO_IRQ_L1 + bool "Locate frequently called do_irq dispatcher function in L1 Memory" + default y + help + If enabled frequently called do_irq dispatcher function is linked + into L1 instruction memory.(less latency) + +config CORE_TIMER_IRQ_L1 + bool "Locate frequently called timer_interrupt() function in L1 Memory" + default y + help + If enabled frequently called timer_interrupt() function is linked + into L1 instruction memory.(less latency) + +config IDLE_L1 + bool "Locate frequently idle function in L1 Memory" + default y + help + If enabled frequently called idle function is linked + into L1 instruction memory.(less latency) + +config SCHEDULE_L1 + bool "Locate kernel schedule function in L1 Memory" + default y + help + If enabled frequently called kernel schedule is linked + into L1 instruction memory.(less latency) + +config ARITHMETIC_OPS_L1 + bool "Locate kernel owned arithmetic functions in L1 Memory" + default y + help + If enabled arithmetic functions are linked + into L1 instruction memory.(less latency) + +config ACCESS_OK_L1 + bool "Locate access_ok function in L1 Memory" + default y + help + If enabled access_ok function is linked + into L1 instruction memory.(less latency) + +config MEMSET_L1 + bool "Locate memset function in L1 Memory" + default y + help + If enabled memset function is linked + into L1 instruction memory.(less latency) + +config MEMCPY_L1 + bool "Locate memcpy function in L1 Memory" + default y + help + If enabled memcpy function is linked + into L1 instruction memory.(less latency) + +config SYS_BFIN_SPINLOCK_L1 + bool "Locate sys_bfin_spinlock function in L1 Memory" + default y + help + If enabled sys_bfin_spinlock function is linked + into L1 instruction memory.(less latency) + +config IP_CHECKSUM_L1 + bool "Locate IP Checksum function in L1 Memory" + default n + help + If enabled IP Checksum function is linked + into L1 instruction memory.(less latency) + +config SYSCALL_TAB_L1 + bool "Locate Syscall Table L1 Data Memory" + default n + depends on !BF531 + help + If enabled the Syscall LUT is linked + into L1 data memory.(less latency) + +config CPLB_SWITCH_TAB_L1 + bool "Locate CPLB Switch Tables L1 Data Memory" + default n + depends on !BF531 + help + If enabled the CPLB Switch Tables are linked + into L1 data memory.(less latency) + endmenu @@ -488,6 +582,10 @@ config BLKFIN_CACHE bool "Enable ICACHE" config BLKFIN_DCACHE bool "Enable DCACHE" +config BLKFIN_DCACHE_BANKA + bool "Enable only 16k BankA DCACHE - BankB is SRAM" + depends on BLKFIN_DCACHE && !BF531 + default n config BLKFIN_CACHE_LOCK bool "Enable Cache Locking" @@ -817,7 +915,7 @@ config DEBUG_KERNEL_START config DEBUG_SERIAL_EARLY_INIT bool "Initialize serial driver early" default n - depends on SERIAL_BLACKFIN + depends on SERIAL_BFIN help Say Y here if you want to get kernel output early when kernel crashes before the normal console initialization. If this option diff -uprN linux-2.6-orig/arch/blackfin/kernel/entry.S linux-2.6/arch/blackfin/kernel/entry.S --- linux-2.6-orig/arch/blackfin/kernel/entry.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/kernel/entry.S 2007-03-05 11:57:23.000000000 +0800 @@ -36,7 +36,11 @@ #include <asm/mach-common/context.S> +#ifdef CONFIG_EXCPT_IRQ_SYSC_L1 +.section .l1.text +#else .text +#endif ENTRY(_ret_from_fork) SP += -12; diff -uprN linux-2.6-orig/arch/blackfin/kernel/irqchip.c linux-2.6/arch/blackfin/kernel/irqchip.c --- linux-2.6-orig/arch/blackfin/kernel/irqchip.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/kernel/irqchip.c 2007-03-05 13:54:10.000000000 +0800 @@ -97,6 +97,11 @@ int show_interrupts(struct seq_file *p, * come via this function. Instead, they should provide their * own 'handler' */ + +#ifdef CONFIG_DO_IRQ_L1 +asmlinkage void asm_do_IRQ(unsigned int irq, struct pt_regs *regs)__attribute__((l1_text)); +#endif + asmlinkage void asm_do_IRQ(unsigned int irq, struct pt_regs *regs) { struct pt_regs *old_regs; diff -uprN linux-2.6-orig/arch/blackfin/kernel/process.c linux-2.6/arch/blackfin/kernel/process.c --- linux-2.6-orig/arch/blackfin/kernel/process.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/kernel/process.c 2007-03-05 13:54:10.000000000 +0800 @@ -97,6 +97,11 @@ static inline void leds_switch(int flag) /* * The idle loop on BFIN */ +#ifdef CONFIG_IDLE_L1 +static inline void default_idle(void)__attribute__((l1_text)); +void cpu_idle(void)__attribute__((l1_text)); +#endif + static inline void default_idle(void) { while (!need_resched()) { @@ -346,3 +351,46 @@ unsigned long get_wchan(struct task_stru while (count++ < 16); return 0; } + +#if !defined(CONFIG_NO_ACCESS_CHECK) +int _access_ok(unsigned long addr, unsigned long size) +{ + + if (addr > (addr + size)) + return 0; + if (segment_eq(get_fs(),KERNEL_DS)) + return 1; +#ifdef CONFIG_MTD_UCLINUX + if (addr >= memory_start && (addr + size) <= memory_end) + return 1; + if (addr >= memory_mtd_end && (addr + size) <= physical_mem_end) + return 1; +#else + if (addr >= memory_start && (addr + size) <= physical_mem_end) + return 1; +#endif + if (addr >= (unsigned long)__init_begin && + addr + size <= (unsigned long)__init_end) + return 1; + if (addr >= L1_SCRATCH_START + && addr + size <= L1_SCRATCH_START + L1_SCRATCH_LENGTH) + return 1; +#if L1_CODE_LENGTH != 0 + if (addr >= L1_CODE_START + (_etext_l1 - _stext_l1) + && addr + size <= L1_CODE_START + L1_CODE_LENGTH) + return 1; +#endif +#if L1_DATA_A_LENGTH != 0 + if (addr >= L1_DATA_A_START + (_ebss_l1 - _sdata_l1) + && addr + size <= L1_DATA_A_START + L1_DATA_A_LENGTH) + return 1; +#endif +#if L1_DATA_B_LENGTH != 0 + if (addr >= L1_DATA_B_START + && addr + size <= L1_DATA_B_START + L1_DATA_B_LENGTH) + return 1; +#endif + return 0; +} +EXPORT_SYMBOL(_access_ok); +#endif diff -uprN linux-2.6-orig/arch/blackfin/kernel/setup.c linux-2.6/arch/blackfin/kernel/setup.c --- linux-2.6-orig/arch/blackfin/kernel/setup.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/kernel/setup.c 2007-03-05 11:57:23.000000000 +0800 @@ -309,6 +309,8 @@ void __init setup_arch(char **cmdline_p) printk(KERN_NOTICE "Warning: limiting memory to %liMB due to hardware anomaly 05000263\n", memory_end >> 20); #endif /* ANOMALY_05000263 */ + memory_end -= SIZE_4K; /*In case there is no valid CPLB behind memory_end make sure we don't get to close*/ + init_mm.start_code = (unsigned long)_stext; init_mm.end_code = (unsigned long)_etext; init_mm.end_data = (unsigned long)_edata; @@ -447,8 +449,8 @@ subsys_initcall(topology_init); #if defined(CONFIG_BLKFIN_DCACHE) || defined(CONFIG_BLKFIN_CACHE) u16 lock_kernel_check(u32 start, u32 end) { - if ((start <= (u32) _stext && end >= (u32) __bss_stop) - || (start >= (u32) _stext && end <= (u32) __bss_stop)) + if ((start <= (u32) _stext && end >= (u32) _end) + || (start >= (u32) _stext && end <= (u32) _end)) return IN_KERNEL; return 0; } @@ -763,16 +765,12 @@ static int show_cpuinfo(struct seq_file #endif u_long cclk = 0, sclk = 0; - u_int id; + u_int id, dcache_size = 0, dsup_banks = 0; cpu = CPU; mmu = "none"; fpu = "none"; - if (&bfin_board_name) { - name = bfin_board_name; - } else { - name = "Unknown"; - } + name = bfin_board_name; cclk = get_cclk(); sclk = get_sclk(); @@ -808,13 +806,36 @@ static int show_cpuinfo(struct seq_file "\n"); else seq_printf(m, "D-CACHE:\tOFF\n"); + + + switch(bfin_read_DMEM_CONTROL() & (1 << DMC0_P | 1 << DMC1_P)) { + case ACACHE_BSRAM: + seq_printf(m, "DBANK-A:\tCACHE\n" "DBANK-B:\tSRAM\n"); + dcache_size = 16; + dsup_banks = 1; + break; + case ACACHE_BCACHE: + seq_printf(m, "DBANK-A:\tCACHE\n" "DBANK-B:\tCACHE\n"); + dcache_size = 32; + dsup_banks = 2; + break; + case ASRAM_BSRAM: + seq_printf(m, "DBANK-A:\tSRAM\n" "DBANK-B:\tSRAM\n"); + dcache_size = 0; + dsup_banks = 0; + break; + default: + break; + } + + seq_printf(m, "I-CACHE Size:\t%dKB\n", BLKFIN_ICACHESIZE / 1024); - seq_printf(m, "D-CACHE Size:\t%dKB\n", BLKFIN_DCACHESIZE / 1024); + seq_printf(m, "D-CACHE Size:\t%dKB\n", dcache_size); seq_printf(m, "I-CACHE Setup:\t%d Sub-banks/%d Ways, %d Lines/Way\n", BLKFIN_ISUBBANKS, BLKFIN_IWAYS, BLKFIN_ILINES); seq_printf(m, "D-CACHE Setup:\t%d Super-banks/%d Sub-banks/%d Ways, %d Lines/Way\n", - BLKFIN_DSUPBANKS, BLKFIN_DSUBBANKS, BLKFIN_DWAYS, + dsup_banks, BLKFIN_DSUBBANKS, BLKFIN_DWAYS, BLKFIN_DLINES); #ifdef CONFIG_BLKFIN_CACHE_LOCK lock = read_iloc(); diff -uprN linux-2.6-orig/arch/blackfin/kernel/signal.c linux-2.6/arch/blackfin/kernel/signal.c --- linux-2.6-orig/arch/blackfin/kernel/signal.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/kernel/signal.c 2007-03-05 11:57:23.000000000 +0800 @@ -66,6 +66,11 @@ struct rt_sigframe { struct ucontext uc; }; +asmlinkage int sys_sigaltstack(const stack_t * uss, stack_t * uoss) +{ + return do_sigaltstack(uss, uoss, rdusp()); +} + static inline int rt_restore_ucontext(struct pt_regs *regs, struct ucontext *uc, int *pr0) { diff -uprN linux-2.6-orig/arch/blackfin/kernel/sys_bfin.c linux-2.6/arch/blackfin/kernel/sys_bfin.c --- linux-2.6-orig/arch/blackfin/kernel/sys_bfin.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/kernel/sys_bfin.c 2007-03-05 13:39:47.000000000 +0800 @@ -115,19 +115,3 @@ asmlinkage void *sys_dma_memcpy(void *de { return safe_dma_memcpy(dest, src, len); } - -/* - * Do a system call from kernel instead of calling sys_execve so we - * end up with proper pt_regs. - */ -int kernel_execve(const char *filename, char *const argv[], char *const envp[]) -{ - register long __res asm ("P0") = __NR_execve; - register long __a asm ("R0") = (long)(filename); - register long __b asm ("R1") = (long)(argv); - register long __c asm ("R2") = (long)(envp); - asm volatile ("raise 15" : "+d" (__a) - : "d" (__res), "d" (__b), "d" (__c)); - return __a; -} - diff -uprN linux-2.6-orig/arch/blackfin/kernel/time.c linux-2.6/arch/blackfin/kernel/time.c --- linux-2.6-orig/arch/blackfin/kernel/time.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/kernel/time.c 2007-03-05 11:57:23.000000000 +0800 @@ -197,6 +197,10 @@ static inline int set_rtc_mmss(unsigned * timer_interrupt() needs to keep up the real-time clock, * as well as call the "do_timer()" routine every clocktick */ +#ifdef CONFIG_CORE_TIMER_IRQ_L1 +irqreturn_t timer_interrupt(int irq, void *dummy)__attribute__((l1_text)); +#endif + irqreturn_t timer_interrupt(int irq, void *dummy) { /* last time the cmos clock got updated */ @@ -243,7 +247,7 @@ void __init time_init(void) * userspace to have to deal with time values greater than * 2^31 seconds (which uClibc cannot cope with yet) */ - if (bfin_read_RTC_STAT() & 0x60000000) { + if ((bfin_read_RTC_STAT() & 0xC0000000) == 0xC0000000) { printk(KERN_NOTICE "bfin-rtc: invalid date; resetting\n"); bfin_write_RTC_STAT(0); } diff -uprN linux-2.6-orig/arch/blackfin/kernel/traps.c linux-2.6/arch/blackfin/kernel/traps.c --- linux-2.6-orig/arch/blackfin/kernel/traps.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/kernel/traps.c 2007-03-05 13:54:10.000000000 +0800 @@ -622,6 +622,10 @@ void dump_bfin_regs(struct pt_regs *fp, printk("\n\n"); } +#ifdef CONFIG_SYS_BFIN_SPINLOCK_L1 +asmlinkage int sys_bfin_spinlock(int *spinlock)__attribute__((l1_text)); +#endif + asmlinkage int sys_bfin_spinlock(int *spinlock) { int ret = 0; diff -uprN linux-2.6-orig/arch/blackfin/lib/ashldi3.c linux-2.6/arch/blackfin/lib/ashldi3.c --- linux-2.6-orig/arch/blackfin/lib/ashldi3.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/ashldi3.c 2007-03-05 11:57:23.000000000 +0800 @@ -31,6 +31,10 @@ #include "gcclib.h" +#ifdef CONFIG_ARITHMETIC_OPS_L1 +DItype __ashldi3(DItype u, word_type b)__attribute__((l1_text)); +#endif + DItype __ashldi3(DItype u, word_type b) { DIunion w; diff -uprN linux-2.6-orig/arch/blackfin/lib/ashrdi3.c linux-2.6/arch/blackfin/lib/ashrdi3.c --- linux-2.6-orig/arch/blackfin/lib/ashrdi3.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/ashrdi3.c 2007-03-05 11:57:23.000000000 +0800 @@ -31,6 +31,10 @@ #include "gcclib.h" +#ifdef CONFIG_ARITHMETIC_OPS_L1 +DItype __ashrdi3(DItype u, word_type b)__attribute__((l1_text)); +#endif + DItype __ashrdi3(DItype u, word_type b) { DIunion w; diff -uprN linux-2.6-orig/arch/blackfin/lib/checksum.c linux-2.6/arch/blackfin/lib/checksum.c --- linux-2.6-orig/arch/blackfin/lib/checksum.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/checksum.c 2007-03-05 11:57:23.000000000 +0800 @@ -34,6 +34,10 @@ #include <net/checksum.h> #include <asm/checksum.h> +#ifdef CONFIG_IP_CHECKSUM_L1 +static unsigned short do_csum(const unsigned char *buff, int len)__attribute__((l1_text)); +#endif + static unsigned short do_csum(const unsigned char *buff, int len) { register unsigned long sum = 0; diff -uprN linux-2.6-orig/arch/blackfin/lib/divsi3.S linux-2.6/arch/blackfin/lib/divsi3.S --- linux-2.6-orig/arch/blackfin/lib/divsi3.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/divsi3.S 2007-03-05 13:54:10.000000000 +0800 @@ -46,7 +46,13 @@ .global ___divsi3; -.align 8; +#ifdef CONFIG_ARITHMETIC_OPS_L1 +.section .l1.text +#else +.text +#endif + +.align 2; ___divsi3 : diff -uprN linux-2.6-orig/arch/blackfin/lib/lshrdi3.c linux-2.6/arch/blackfin/lib/lshrdi3.c --- linux-2.6-orig/arch/blackfin/lib/lshrdi3.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/lshrdi3.c 2007-03-05 11:57:23.000000000 +0800 @@ -45,6 +45,10 @@ typedef union { DItype ll; } DIunion; +#ifdef CONFIG_ARITHMETIC_OPS_L1 +DItype __lshrdi3(DItype u, word_type b)__attribute__((l1_text)); +#endif + DItype __lshrdi3(DItype u, word_type b) { DIunion w; diff -uprN linux-2.6-orig/arch/blackfin/lib/memcpy.S linux-2.6/arch/blackfin/lib/memcpy.S --- linux-2.6-orig/arch/blackfin/lib/memcpy.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/memcpy.S 2007-03-05 11:57:23.000000000 +0800 @@ -36,6 +36,12 @@ #include <linux/linkage.h> +#ifdef CONFIG_MEMCPY_L1 +.section .l1.text +#else +.text +#endif + .align 2 ENTRY(_memcpy) diff -uprN linux-2.6-orig/arch/blackfin/lib/memset.S linux-2.6/arch/blackfin/lib/memset.S --- linux-2.6-orig/arch/blackfin/lib/memset.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/memset.S 2007-03-05 11:57:23.000000000 +0800 @@ -33,6 +33,12 @@ .align 2 +#ifdef CONFIG_MEMSET_L1 +.section .l1.text +#else +.text +#endif + /* * C Library function MEMSET * R0 = address (leave unchanged to form result) diff -uprN linux-2.6-orig/arch/blackfin/lib/modsi3.S linux-2.6/arch/blackfin/lib/modsi3.S --- linux-2.6-orig/arch/blackfin/lib/modsi3.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/modsi3.S 2007-03-05 11:57:23.000000000 +0800 @@ -39,6 +39,13 @@ .type ___modsi3, STT_FUNC; .extern ___divsi3; .type ___divsi3, STT_FUNC; + +#ifdef CONFIG_ARITHMETIC_OPS_L1 +.section .l1.text +#else +.text +#endif + ___modsi3: CC=R0==0; diff -uprN linux-2.6-orig/arch/blackfin/lib/muldi3.c linux-2.6/arch/blackfin/lib/muldi3.c --- linux-2.6-orig/arch/blackfin/lib/muldi3.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/muldi3.c 2007-03-05 11:57:23.000000000 +0800 @@ -83,6 +83,10 @@ typedef union { ditype ll; } diunion; +#ifdef CONFIG_ARITHMETIC_OPS_L1 +ditype __muldi3(ditype u, ditype v)__attribute__((l1_text)); +#endif + ditype __muldi3(ditype u, ditype v) { diunion w; diff -uprN linux-2.6-orig/arch/blackfin/lib/smulsi3_highpart.S linux-2.6/arch/blackfin/lib/smulsi3_highpart.S --- linux-2.6-orig/arch/blackfin/lib/smulsi3_highpart.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/smulsi3_highpart.S 2007-03-05 11:57:23.000000000 +0800 @@ -2,6 +2,12 @@ .global ___smulsi3_highpart; .type ___smulsi3_highpart, STT_FUNC; +#ifdef CONFIG_ARITHMETIC_OPS_L1 +.section .l1.text +#else +.text +#endif + ___smulsi3_highpart: R2 = R1.L * R0.L (FU); R3 = R1.H * R0.L (IS,M); diff -uprN linux-2.6-orig/arch/blackfin/lib/udivsi3.S linux-2.6/arch/blackfin/lib/udivsi3.S --- linux-2.6-orig/arch/blackfin/lib/udivsi3.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/udivsi3.S 2007-03-05 14:33:30.000000000 +0800 @@ -33,6 +33,13 @@ #define CARRY AC0 +#ifdef CONFIG_ARITHMETIC_OPS_L1 +.section .l1.text +#else +.text +#endif + + ENTRY(___udivsi3) CC = R0 < R1 (IU); /* If X < Y, always return 0 */ @@ -105,7 +112,7 @@ ENTRY(___udivsi3) ** with some post-adjustment */ R3 = R1 >> 1; /* Pre-scaled divisor for primitive case */ - R2 = R0 >> 16 ; + R2 = R0 >> 16; R2 = R3 - R2; /* shifted divisor < upper 16 bits of dividend */ CC &= CARRY; diff -uprN linux-2.6-orig/arch/blackfin/lib/umodsi3.S linux-2.6/arch/blackfin/lib/umodsi3.S --- linux-2.6-orig/arch/blackfin/lib/umodsi3.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/umodsi3.S 2007-03-05 11:57:23.000000000 +0800 @@ -29,7 +29,12 @@ * 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ +#ifdef CONFIG_ARITHMETIC_OPS_L1 +.section .l1.text +#else .text +#endif + .extern ___udivsi3; .globl ___umodsi3 ___umodsi3: diff -uprN linux-2.6-orig/arch/blackfin/lib/umulsi3_highpart.S linux-2.6/arch/blackfin/lib/umulsi3_highpart.S --- linux-2.6-orig/arch/blackfin/lib/umulsi3_highpart.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/lib/umulsi3_highpart.S 2007-03-05 11:57:23.000000000 +0800 @@ -2,6 +2,12 @@ .global ___umulsi3_highpart; .type ___umulsi3_highpart, STT_FUNC; +#ifdef CONFIG_ARITHMETIC_OPS_L1 +.section .l1.text +#else +.text +#endif + ___umulsi3_highpart: R2 = R1.H * R0.H, R3 = R1.L * R0.H (FU); R0 = R1.L * R0.L, R1 = R1.H * R0.L (FU); diff -uprN linux-2.6-orig/arch/blackfin/mach-bf537/boards/stamp.c linux-2.6/arch/blackfin/mach-bf537/boards/stamp.c --- linux-2.6-orig/arch/blackfin/mach-bf537/boards/stamp.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/mach-bf537/boards/stamp.c 2007-03-05 13:54:10.000000000 +0800 @@ -40,6 +40,8 @@ #include <linux/usb_isp1362.h> #endif #include <asm/irq.h> +#include <linux/irq.h> +#include <linux/interrupt.h> #include <asm/bfin5xx_spi.h> #include <linux/usb_sl811.h> @@ -54,6 +56,53 @@ char *bfin_board_name = "ADDS-BF537-STAM * Driver needs to know address, irq and flag pin. */ +#define ISP1761_BASE 0x203C0000 +#define ISP1761_IRQ IRQ_PF7 + +#if defined(CONFIG_USB_ISP1760_HCD) || defined(CONFIG_USB_ISP1760_HCD_MODULE) +static struct resource bfin_isp1761_resources[] = { + [0] = { + .name = "isp1761-regs", + .start = ISP1761_BASE + 0x00000000, + .end = ISP1761_BASE + 0x000fffff, + .flags = IORESOURCE_MEM, + }, + [1] = { + .start = ISP1761_IRQ, + .end = ISP1761_IRQ, + .flags = IORESOURCE_IRQ | IORESOURCE_IRQ_LOWLEVEL, + }, +}; + +static struct platform_device bfin_isp1761_device = { + .name = "isp1761", + .id = 0, + .num_resources = ARRAY_SIZE(bfin_isp1761_resources), + .resource = bfin_isp1761_resources, +}; + +static struct platform_device *bfin_isp1761_devices[] = { + &bfin_isp1761_device, +}; + +int __init bfin_isp1761_init(void) +{ + unsigned int num_devices=ARRAY_SIZE(bfin_isp1761_devices); + + printk(KERN_INFO "%s(): registering device resources\n", __FUNCTION__); + set_irq_type(ISP1761_IRQ, IRQF_TRIGGER_FALLING); + + return platform_add_devices(bfin_isp1761_devices, num_devices); +} + +void __exit bfin_isp1761_exit(void) +{ + platform_device_unregister(&bfin_isp1761_device); +} + +arch_initcall(bfin_isp1761_init); +#endif + #if defined(CONFIG_BFIN_CFPCMCIA) || defined(CONFIG_BFIN_CFPCMCIA_MODULE) static struct resource bfin_pcmcia_cf_resources[] = { { diff -uprN linux-2.6-orig/arch/blackfin/mach-bf537/head.S linux-2.6/arch/blackfin/mach-bf537/head.S --- linux-2.6-orig/arch/blackfin/mach-bf537/head.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/mach-bf537/head.S 2007-03-05 11:57:23.000000000 +0800 @@ -240,7 +240,6 @@ ENTRY(__stext) r0 = AMGCTLVAL; w[p2] = r0; ssync; - call _real_start; /* This section keeps the processor in supervisor mode * during kernel boot. Switches to user mode at end of boot. diff -uprN linux-2.6-orig/arch/blackfin/mach-bf561/boards/generic_board.c linux-2.6/arch/blackfin/mach-bf561/boards/generic_board.c --- linux-2.6-orig/arch/blackfin/mach-bf561/boards/generic_board.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/mach-bf561/boards/generic_board.c 2007-03-05 11:57:23.000000000 +0800 @@ -34,6 +34,8 @@ #include <linux/platform_device.h> #include <asm/irq.h> +char *bfin_board_name = "UNKNOWN BOARD"; + /* * Driver needs to know address, irq and flag pin. */ diff -uprN linux-2.6-orig/arch/blackfin/mach-common/cache.S linux-2.6/arch/blackfin/mach-common/cache.S --- linux-2.6-orig/arch/blackfin/mach-common/cache.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/mach-common/cache.S 2007-03-05 11:57:23.000000000 +0800 @@ -211,7 +211,7 @@ ENTRY(_dcache_invalidate) /* Configures the data cache again */ - R6 = (ACACHE_BCACHE | ENDCPLB | PORT_PREF0); + R6 = DMEM_CNTR; R7 = R7 | R6; CLI R6; diff -uprN linux-2.6-orig/arch/blackfin/mach-common/cplbhdlr.S linux-2.6/arch/blackfin/mach-common/cplbhdlr.S --- linux-2.6-orig/arch/blackfin/mach-common/cplbhdlr.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/mach-common/cplbhdlr.S 2007-03-05 11:57:23.000000000 +0800 @@ -33,7 +33,11 @@ #include <asm/cplb.h> #include <asm/entry.h> +#ifdef CONFIG_EXCPT_IRQ_SYSC_L1 +.section .l1.text +#else .text +#endif .type _cplb_mgr, STT_FUNC; .type _panic_cplb_error, STT_FUNC; diff -uprN linux-2.6-orig/arch/blackfin/mach-common/cplbmgr.S linux-2.6/arch/blackfin/mach-common/cplbmgr.S --- linux-2.6-orig/arch/blackfin/mach-common/cplbmgr.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/mach-common/cplbmgr.S 2007-03-05 13:54:10.000000000 +0800 @@ -55,7 +55,11 @@ #include <asm/blackfin.h> #include <asm/cplb.h> +#ifdef CONFIG_EXCPT_IRQ_SYSC_L1 +.section .l1.text +#else .text +#endif .align 2; ENTRY(_cplb_mgr) diff -uprN linux-2.6-orig/arch/blackfin/mach-common/entry.S linux-2.6/arch/blackfin/mach-common/entry.S --- linux-2.6-orig/arch/blackfin/mach-common/entry.S 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/mach-common/entry.S 2007-03-05 13:39:47.000000000 +0800 @@ -79,13 +79,18 @@ # define DEBUG_STOP_HWTRACE #endif +#ifdef CONFIG_EXCPT_IRQ_SYSC_L1 +.section .l1.text +#else .text +#endif /* Slightly simplified and streamlined entry point for CPLB misses. * This one does not lower the level to IRQ5, and thus can be used to * patch up CPLB misses on the kernel stack. */ ENTRY(_ex_dcplb) +#if defined(ANOMALY_05000261) /* * Work around an anomaly: if we see a new DCPLB fault, return * without doing anything. Then, if we get the same fault again, @@ -99,6 +104,7 @@ ENTRY(_ex_dcplb) cc = r6 == r7; if !cc jump _return_from_exception; /* fall through */ +#endif ENTRY(_ex_icplb) (R7:6,P5:4) = [sp++]; @@ -333,7 +339,7 @@ ENTRY(_trap) /* Exception: 4th entry int [sp + PT_R0] = r7; /* return value from system call */ jump .Lsyscall_really_exit; -ENTRY(_execve) +ENTRY(_kernel_execve) link SIZEOF_PTREGS; p0 = sp; r3 = SIZEOF_PTREGS / 4; @@ -749,7 +755,11 @@ ENTRY(_init_exception_buff) * a CPLB. This is needed to ensure we don't get double fault conditions */ +#ifdef CONFIG_SYSCALL_TAB_L1 +.section .l1.data +#else .data +#endif ALIGN _extable: /* entry for each EXCAUSE[5:0] @@ -1010,7 +1020,7 @@ ENTRY(_sys_call_table) .long _sys_getcwd .long _sys_capget .long _sys_capset /* 185 */ - .long _sys_ni_syscall /* old sys_sigaltstack */ + .long _sys_sigaltstack .long _sys_sendfile .long _sys_ni_syscall /* streams1 */ .long _sys_ni_syscall /* streams2 */ @@ -1178,10 +1188,11 @@ _exception_stack: .endr _exception_stack_top: +#if defined(ANOMALY_05000261) /* Used by the assembly entry point to work around an anomaly. */ _last_cplb_fault_retx: .long 0; - +#endif /* * Single instructions can have multiple faults, which need to be * handled by traps.c, in irq5. We store the exception cause to ensure diff -uprN linux-2.6-orig/arch/blackfin/mach-common/ints-priority-dc.c linux-2.6/arch/blackfin/mach-common/ints-priority-dc.c --- linux-2.6-orig/arch/blackfin/mach-common/ints-priority-dc.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/mach-common/ints-priority-dc.c 2007-03-05 13:54:10.000000000 +0800 @@ -438,6 +438,10 @@ int __init init_arch_irq(void) return 0; } +#ifdef CONFIG_DO_IRQ_L1 +void do_irq(int vec, struct pt_regs *fp)__attribute__((l1_text)); +#endif + void do_irq(int vec, struct pt_regs *fp) { if (vec == EVT_IVTMR_P) { diff -uprN linux-2.6-orig/arch/blackfin/mach-common/ints-priority-sc.c linux-2.6/arch/blackfin/mach-common/ints-priority-sc.c --- linux-2.6-orig/arch/blackfin/mach-common/ints-priority-sc.c 2007-03-05 11:50:21.000000000 +0800 +++ linux-2.6/arch/blackfin/mach-common/ints-priority-sc.c 2007-03-05 11:57:23.000000000 +0800 @@ -547,6 +547,10 @@ int __init init_arch_irq(void) return 0; } +#ifdef CONFIG_DO_IRQ_L1 +void do_irq(int vec, struct pt_regs *fp)__attribute__((l1_text)); +#endif + void do_irq(int vec, struct pt_regs *fp) { if (vec == EVT_IVTMR_P) { diff -uprN linux-2.6-orig/include/asm-blackfin/cplbinit.h linux-2.6/include/asm-blackfin/cplbinit.h --- linux-2.6-orig/include/asm-blackfin/cplbinit.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/cplbinit.h 2007-03-05 11:57:23.000000000 +0800 @@ -1,192 +1,205 @@ -/* - * File: include/asm-blackfin/cplbinit.h - * Based on: - * Author: - * - * Created: - * Description: - * - * Rev: $Id: cplbinit.h,v 1.3 2007/01/02 03:11:59 adamliyi Exp $ - * - * Modified: - * Copyright 2004-2006 Analog Devices Inc. - * - * Bugs: Enter bugs at http://blackfin.uclinux.org/ - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. - * - * This program is distributed in the hope that 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, see the file COPYING, or write - * to the Free Software Foundation, Inc., - * 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA - */ - -#include <asm/blackfin.h> -#include <asm/cplb.h> - -#define INITIAL_T 0x1 -#define SWITCH_T 0x2 -#define I_CPLB 0x4 -#define D_CPLB 0x8 - -#define IN_KERNEL 1 - -enum -{ZERO_P, L1I_MEM, L1D_MEM, SDRAM_KERN , SDRAM_RAM_MTD, SDRAM_DMAZ, RES_MEM, ASYNC_MEM, L2_MEM}; - -struct cplb_desc { - u32 start; /* start address */ - u32 end; /* end address */ - u32 psize; /* prefered size if any otherwise 1MB or 4MB*/ - u16 attr;/* attributes */ - u16 i_conf;/* I-CPLB DATA */ - u16 d_conf;/* D-CPLB DATA */ - u16 valid;/* valid */ - const s8 name[30];/* name */ -}; - -struct cplb_tab { - u_long *tab; - u16 pos; - u16 size; -}; - -u_long icplb_table[MAX_CPLBS+1]; -u_long dcplb_table[MAX_CPLBS+1]; - -/* Till here we are discussing about the static memory management model. - * However, the operating envoronments commonly define more CPLB - * descriptors to cover the entire addressable memory than will fit into - * the available on-chip 16 CPLB MMRs. When this happens, the below table - * will be used which will hold all the potentially required CPLB descriptors - * - * This is how Page descriptor Table is implemented in uClinux/Blackfin. - */ - -u_long ipdt_table[MAX_SWITCH_I_CPLBS+1]; -u_long dpdt_table[MAX_SWITCH_D_CPLBS+1]; - -#ifdef CONFIG_CPLB_INFO -u_long ipdt_swapcount_table[MAX_SWITCH_I_CPLBS]; -u_long dpdt_swapcount_table[MAX_SWITCH_D_CPLBS]; -#endif - -struct s_cplb { - struct cplb_tab init_i; - struct cplb_tab init_d; - struct cplb_tab switch_i; - struct cplb_tab switch_d; -}; - -#if defined(CONFIG_BLKFIN_DCACHE) || defined(CONFIG_BLKFIN_CACHE) -static struct cplb_desc cplb_data[] = { - { - .start = 0, - .end = SIZE_4K, - .psize = SIZE_4K, - .attr = INITIAL_T | SWITCH_T | I_CPLB | D_CPLB, - .i_conf = SDRAM_OOPS, - .d_conf = SDRAM_OOPS, -#if defined(CONFIG_DEBUG_HUNT_FOR_ZERO) - .valid = 1, -#else - .valid = 0, -#endif - .name = "ZERO Pointer Saveguard", - }, - { - .start = L1_CODE_START, - .end = L1_CODE_START + L1_CODE_LENGTH, - .psize = SIZE_4M, - .attr = INITIAL_T | SWITCH_T | I_CPLB, - .i_conf = L1_IMEMORY, - .d_conf = 0, - .valid = 1, - .name = "L1 I-Memory", - }, - { - .start = L1_DATA_A_START, - .end = L1_DATA_B_START + L1_DATA_B_LENGTH, - .psize = SIZE_4M, - .attr = INITIAL_T | SWITCH_T | D_CPLB, - .i_conf = 0, - .d_conf = L1_DMEMORY, -#if defined(CONFIG_BF536) - .valid = 0, -#else - .valid = 1, -#endif - .name = "L1 D-Memory", - }, - { - .start = 0, - .end = 0, /* dynamic */ - .psize = 0, - .attr = INITIAL_T | SWITCH_T | I_CPLB | D_CPLB, - .i_conf = SDRAM_IGENERIC, - .d_conf = SDRAM_DGENERIC, - .valid = 1, - .name = "SDRAM Kernel", - }, - { - .start = 0, /* dynamic */ - .end = 0, /* dynamic */ - .psize = 0, - .attr = INITIAL_T | SWITCH_T | D_CPLB, - .i_conf = SDRAM_IGENERIC, - .d_conf = SDRAM_DNON_CHBL, - .valid = 1, - .name = "SDRAM RAM MTD", - }, - { - .start = 0, /* dynamic */ - .end = 0, /* dynamic */ - .psize = SIZE_1M, - .attr = INITIAL_T | SWITCH_T | D_CPLB, - .d_conf = SDRAM_DNON_CHBL, - .valid = 1,//(DMA_UNCACHED_REGION > 0), - .name = "SDRAM Uncached DMA ZONE", - }, - { - .start = 0, /* dynamic */ - .end = 0, /* dynamic */ - .psize = 0, - .attr = SWITCH_T | D_CPLB, - .i_conf = 0, /* dynamic */ - .d_conf = 0, /* dynamic */ - .valid = 1, - .name = "SDRAM Reserved Memory", - }, - { - .start = ASYNC_BANK0_BASE, - .end = ASYNC_BANK3_BASE + ASYNC_BANK3_SIZE, - .psize = 0, - .attr = SWITCH_T | D_CPLB, - .d_conf = SDRAM_EBIU, - .valid = 1, - .name = "ASYNC Memory", - }, - { -#if defined(CONFIG_BF561) - .start = L2_SRAM, - .end = L2_SRAM_END, - .psize = SIZE_1M, - .attr = SWITCH_T | D_CPLB, - .i_conf = L2_MEMORY, - .d_conf = L2_MEMORY, - .valid = 1, -#else - .valid = 0, -#endif - .name = "L2 Memory", - } -}; -#endif +/* + * File: include/asm-blackfin/cplbinit.h + * Based on: + * Author: + * + * Created: + * Description: + * + * Rev: $Id: cplbinit.h,v 1.3 2007/01/02 03:11:59 adamliyi Exp $ + * + * Modified: + * Copyright 2004-2006 Analog Devices Inc. + * + * Bugs: Enter bugs at http://blackfin.uclinux.org/ + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that 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, see the file COPYING, or write + * to the Free Software Foundation, Inc., + * 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include <asm/blackfin.h> +#include <asm/cplb.h> + +#define INITIAL_T 0x1 +#define SWITCH_T 0x2 +#define I_CPLB 0x4 +#define D_CPLB 0x8 + +#define IN_KERNEL 1 + +enum +{ZERO_P, L1I_MEM, L1D_MEM, SDRAM_KERN , SDRAM_RAM_MTD, SDRAM_DMAZ, RES_MEM, ASYNC_MEM, L2_MEM}; + +struct cplb_desc { + u32 start; /* start address */ + u32 end; /* end address */ + u32 psize; /* prefered size if any otherwise 1MB or 4MB*/ + u16 attr;/* attributes */ + u16 i_conf;/* I-CPLB DATA */ + u16 d_conf;/* D-CPLB DATA */ + u16 valid;/* valid */ + const s8 name[30];/* name */ +}; + +struct cplb_tab { + u_long *tab; + u16 pos; + u16 size; +}; + +u_long icplb_table[MAX_CPLBS+1]; +u_long dcplb_table[MAX_CPLBS+1]; + +/* Till here we are discussing about the static memory management model. + * However, the operating envoronments commonly define more CPLB + * descriptors to cover the entire addressable memory than will fit into + * the available on-chip 16 CPLB MMRs. When this happens, the below table + * will be used which will hold all the potentially required CPLB descriptors + * + * This is how Page descriptor Table is implemented in uClinux/Blackfin. + */ + +#ifdef CONFIG_CPLB_SWITCH_TAB_L1 +u_long ipdt_table[MAX_SWITCH_I_CPLBS+1]__attribute__((l1_data)); +u_long dpdt_table[MAX_SWITCH_D_CPLBS+1]__attribute__((l1_data)); + +#ifdef CONFIG_CPLB_INFO +u_long ipdt_swapcount_table[MAX_SWITCH_I_CPLBS]__attribute__((l1_data)); +u_long dpdt_swapcount_table[MAX_SWITCH_D_CPLBS]__attribute__((l1_data)); +#endif /* CONFIG_CPLB_INFO */ + +#else + +u_long ipdt_table[MAX_SWITCH_I_CPLBS+1]; +u_long dpdt_table[MAX_SWITCH_D_CPLBS+1]; + +#ifdef CONFIG_CPLB_INFO +u_long ipdt_swapcount_table[MAX_SWITCH_I_CPLBS]; +u_long dpdt_swapcount_table[MAX_SWITCH_D_CPLBS]; +#endif /* CONFIG_CPLB_INFO */ + +#endif /*CONFIG_CPLB_SWITCH_TAB_L1*/ + +struct s_cplb { + struct cplb_tab init_i; + struct cplb_tab init_d; + struct cplb_tab switch_i; + struct cplb_tab switch_d; +}; + +#if defined(CONFIG_BLKFIN_DCACHE) || defined(CONFIG_BLKFIN_CACHE) +static struct cplb_desc cplb_data[] = { + { + .start = 0, + .end = SIZE_4K, + .psize = SIZE_4K, + .attr = INITIAL_T | SWITCH_T | I_CPLB | D_CPLB, + .i_conf = SDRAM_OOPS, + .d_conf = SDRAM_OOPS, +#if defined(CONFIG_DEBUG_HUNT_FOR_ZERO) + .valid = 1, +#else + .valid = 0, +#endif + .name = "ZERO Pointer Saveguard", + }, + { + .start = L1_CODE_START, + .end = L1_CODE_START + L1_CODE_LENGTH, + .psize = SIZE_4M, + .attr = INITIAL_T | SWITCH_T | I_CPLB, + .i_conf = L1_IMEMORY, + .d_conf = 0, + .valid = 1, + .name = "L1 I-Memory", + }, + { + .start = L1_DATA_A_START, + .end = L1_DATA_B_START + L1_DATA_B_LENGTH, + .psize = SIZE_4M, + .attr = INITIAL_T | SWITCH_T | D_CPLB, + .i_conf = 0, + .d_conf = L1_DMEMORY, +#if ((L1_DATA_A_LENGTH > 0) || (L1_DATA_B_LENGTH > 0)) + .valid = 1, +#else + .valid = 0, +#endif + .name = "L1 D-Memory", + }, + { + .start = 0, + .end = 0, /* dynamic */ + .psize = 0, + .attr = INITIAL_T | SWITCH_T | I_CPLB | D_CPLB, + .i_conf = SDRAM_IGENERIC, + .d_conf = SDRAM_DGENERIC, + .valid = 1, + .name = "SDRAM Kernel", + }, + { + .start = 0, /* dynamic */ + .end = 0, /* dynamic */ + .psize = 0, + .attr = INITIAL_T | SWITCH_T | D_CPLB, + .i_conf = SDRAM_IGENERIC, + .d_conf = SDRAM_DNON_CHBL, + .valid = 1, + .name = "SDRAM RAM MTD", + }, + { + .start = 0, /* dynamic */ + .end = 0, /* dynamic */ + .psize = SIZE_1M, + .attr = INITIAL_T | SWITCH_T | D_CPLB, + .d_conf = SDRAM_DNON_CHBL, + .valid = 1,//(DMA_UNCACHED_REGION > 0), + .name = "SDRAM Uncached DMA ZONE", + }, + { + .start = 0, /* dynamic */ + .end = 0, /* dynamic */ + .psize = 0, + .attr = SWITCH_T | D_CPLB, + .i_conf = 0, /* dynamic */ + .d_conf = 0, /* dynamic */ + .valid = 1, + .name = "SDRAM Reserved Memory", + }, + { + .start = ASYNC_BANK0_BASE, + .end = ASYNC_BANK3_BASE + ASYNC_BANK3_SIZE, + .psize = 0, + .attr = SWITCH_T | D_CPLB, + .d_conf = SDRAM_EBIU, + .valid = 1, + .name = "ASYNC Memory", + }, + { +#if defined(CONFIG_BF561) + .start = L2_SRAM, + .end = L2_SRAM_END, + .psize = SIZE_1M, + .attr = SWITCH_T | D_CPLB, + .i_conf = L2_MEMORY, + .d_conf = L2_MEMORY, + .valid = 1, +#else + .valid = 0, +#endif + .name = "L2 Memory", + } +}; +#endif diff -uprN linux-2.6-orig/include/asm-blackfin/delay.h linux-2.6/include/asm-blackfin/delay.h --- linux-2.6-orig/include/asm-blackfin/delay.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/delay.h 2007-03-05 11:57:23.000000000 +0800 @@ -7,10 +7,11 @@ static inline void __delay(unsigned long /* FIXME: Currently the assembler doesn't recognize Loop Register Clobbers, uncomment this as soon those are implemented */ /* - __asm__ __volatile__ ( "\t LSETUP (1,1) LC0= %0\n\t" - "l:\t NOP;\n\t" + __asm__ __volatile__ ( "\t LSETUP (1f,1f) LC0= %0\n\t" + "1:\t NOP;\n\t" : :"a" (loops) : "LT0","LB0","LC0"); + */ __asm__ __volatile__("[--SP] = LC0;\n\t" diff -uprN linux-2.6-orig/include/asm-blackfin/dma-mapping.h linux-2.6/include/asm-blackfin/dma-mapping.h --- linux-2.6-orig/include/asm-blackfin/dma-mapping.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/dma-mapping.h 2007-03-05 11:57:23.000000000 +0800 @@ -65,4 +65,9 @@ static inline BUG_ON(direction == DMA_NONE); } +/* Now for the API extensions over the pci_ one */ + +#define dma_alloc_noncoherent(d, s, h, f) dma_alloc_coherent(d, s, h, f) +#define dma_free_noncoherent(d, s, v, h) dma_free_coherent(d, s, v, h) + #endif /* _BLACKFIN_DMA_MAPPING_H */ diff -uprN linux-2.6-orig/include/asm-blackfin/mach-bf533/bf533.h linux-2.6/include/asm-blackfin/mach-bf533/bf533.h --- linux-2.6-orig/include/asm-blackfin/mach-bf533/bf533.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/mach-bf533/bf533.h 2007-03-05 13:54:10.000000000 +0800 @@ -53,15 +53,6 @@ /***************************/ -#define BLKFIN_ICACHESIZE (16*1024) - -#if defined(CONFIG_BF533) || defined(CONFIG_BF532) -#define BLKFIN_DCACHESIZE (32*1024) -#define BLKFIN_DSUPBANKS 2 -#else -#define BLKFIN_DCACHESIZE (16*1024) -#define BLKFIN_DSUPBANKS 1 -#endif #define BLKFIN_DSUBBANKS 4 #define BLKFIN_DWAYS 2 @@ -280,7 +271,7 @@ | CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND) #endif -#define L1_DMEMORY (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_DIRTY) +#define L1_DMEMORY (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_LOCK | CPLB_DIRTY) #define SDRAM_DNON_CHBL (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_DIRTY) #define SDRAM_EBIU (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_DIRTY) #define SDRAM_OOPS (CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_LOCK | CPLB_DIRTY) diff -uprN linux-2.6-orig/include/asm-blackfin/mach-bf533/mem_map.h linux-2.6/include/asm-blackfin/mach-bf533/mem_map.h --- linux-2.6-orig/include/asm-blackfin/mach-bf533/mem_map.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/mach-bf533/mem_map.h 2007-03-05 11:57:23.000000000 +0800 @@ -52,6 +52,12 @@ /* Level 1 Memory */ +#ifdef CONFIG_BLKFIN_CACHE +#define BLKFIN_ICACHESIZE (16*1024) +#else +#define BLKFIN_ICACHESIZE (0*1024) +#endif + /* Memory Map for ADSP-BF533 processors */ #ifdef CONFIG_BF533 @@ -66,18 +72,28 @@ #endif #ifdef CONFIG_BLKFIN_DCACHE -#define DMEM_CNTR (ACACHE_BCACHE | ENDCPLB | PORT_PREF0) + +#ifdef CONFIG_BLKFIN_DCACHE_BANKA +#define DMEM_CNTR (ACACHE_BSRAM | ENDCPLB | PORT_PREF0) #define L1_DATA_A_LENGTH (0x8000 - 0x4000) +#define L1_DATA_B_LENGTH 0x8000 +#define BLKFIN_DCACHESIZE (16*1024) +#define BLKFIN_DSUPBANKS 1 #else -#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) -#define L1_DATA_A_LENGTH 0x8000 +#define DMEM_CNTR (ACACHE_BCACHE | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH (0x8000 - 0x4000) +#define L1_DATA_B_LENGTH (0x8000 - 0x4000) +#define BLKFIN_DCACHESIZE (32*1024) +#define BLKFIN_DSUPBANKS 2 #endif -#ifdef CONFIG_BLKFIN_DCACHE -#define L1_DATA_B_LENGTH (0x8000 - 0x4000) #else +#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH 0x8000 #define L1_DATA_B_LENGTH 0x8000 -#endif +#define BLKFIN_DCACHESIZE (0*1024) +#define BLKFIN_DSUPBANKS 0 +#endif /*CONFIG_BLKFIN_DCACHE*/ #endif /* Memory Map for ADSP-BF532 processors */ @@ -94,18 +110,29 @@ #endif #ifdef CONFIG_BLKFIN_DCACHE -#define DMEM_CNTR (ACACHE_BCACHE | ENDCPLB | PORT_PREF0) + +#ifdef CONFIG_BLKFIN_DCACHE_BANKA +#define DMEM_CNTR (ACACHE_BSRAM | ENDCPLB | PORT_PREF0) #define L1_DATA_A_LENGTH (0x4000 - 0x4000) +#define L1_DATA_B_LENGTH 0x4000 +#define BLKFIN_DCACHESIZE (16*1024) +#define BLKFIN_DSUPBANKS 1 + #else -#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) -#define L1_DATA_A_LENGTH 0x4000 +#define DMEM_CNTR (ACACHE_BCACHE | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH (0x4000 - 0x4000) +#define L1_DATA_B_LENGTH (0x4000 - 0x4000) +#define BLKFIN_DCACHESIZE (32*1024) +#define BLKFIN_DSUPBANKS 2 #endif -#ifdef CONFIG_BLKFIN_DCACHE -#define L1_DATA_B_LENGTH (0x4000 - 0x4000) #else +#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH 0x4000 #define L1_DATA_B_LENGTH 0x4000 -#endif +#define BLKFIN_DCACHESIZE (0*1024) +#define BLKFIN_DSUPBANKS 0 +#endif /*CONFIG_BLKFIN_DCACHE*/ #endif /* Memory Map for ADSP-BF531 processors */ @@ -116,13 +143,20 @@ #define L1_DATA_B_START 0xFF904000 #define L1_CODE_LENGTH 0x4000 #define L1_DATA_B_LENGTH 0x0000 + + #ifdef CONFIG_BLKFIN_DCACHE -#define DMEM_CNTR (ACACHE_BSRAM | ENDCPLB) +#define DMEM_CNTR (ACACHE_BSRAM | ENDCPLB | PORT_PREF0) #define L1_DATA_A_LENGTH (0x4000 - 0x4000) +#define BLKFIN_DCACHESIZE (16*1024) +#define BLKFIN_DSUPBANKS 1 #else -#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB) +#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) #define L1_DATA_A_LENGTH 0x4000 +#define BLKFIN_DCACHESIZE (0*1024) +#define BLKFIN_DSUPBANKS 0 #endif + #endif /* Scratch Pad Memory */ diff -uprN linux-2.6-orig/include/asm-blackfin/mach-bf537/bf537.h linux-2.6/include/asm-blackfin/mach-bf537/bf537.h --- linux-2.6-orig/include/asm-blackfin/mach-bf537/bf537.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/mach-bf537/bf537.h 2007-03-05 11:57:23.000000000 +0800 @@ -63,15 +63,6 @@ /***************************/ -#define BLKFIN_ICACHESIZE (16*1024) - -#if defined(CONFIG_BF537) || defined(CONFIG_BF534) -#define BLKFIN_DCACHESIZE (32*1024) -#define BLKFIN_DSUPBANKS 2 -#else -#define BLKFIN_DCACHESIZE (16*1024) -#define BLKFIN_DSUPBANKS 1 -#endif #define BLKFIN_DSUBBANKS 4 #define BLKFIN_DWAYS 2 @@ -261,7 +252,7 @@ #endif -#define L1_DMEMORY (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_DIRTY ) +#define L1_DMEMORY (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_LOCK | CPLB_DIRTY ) #define SDRAM_DNON_CHBL (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_DIRTY ) #define SDRAM_EBIU (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_DIRTY ) #define SDRAM_OOPS (CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_LOCK | CPLB_DIRTY ) diff -uprN linux-2.6-orig/include/asm-blackfin/mach-bf537/defBF534.h linux-2.6/include/asm-blackfin/mach-bf537/defBF534.h --- linux-2.6-orig/include/asm-blackfin/mach-bf537/defBF534.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/mach-bf537/defBF534.h 2007-03-05 11:57:23.000000000 +0800 @@ -1171,7 +1171,7 @@ #define GM 0x0008 /* Get More (When RDBR Full, Overwrite/Discard*) */ #define PSSE 0x0010 /* Slave-Select Input Enable */ #define EMISO 0x0020 /* Enable MISO As Output */ -#define SIZE 0x0100 /* Size of Words (16/8* Bits) */ +#define SPI_SIZE 0x0100 /* Size of Words (16/8* Bits) */ #define LSBF 0x0200 /* LSB First */ #define CPHA 0x0400 /* Clock Phase */ #define CPOL 0x0800 /* Clock Polarity */ diff -uprN linux-2.6-orig/include/asm-blackfin/mach-bf537/mem_map.h linux-2.6/include/asm-blackfin/mach-bf537/mem_map.h --- linux-2.6-orig/include/asm-blackfin/mach-bf537/mem_map.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/mach-bf537/mem_map.h 2007-03-05 11:57:23.000000000 +0800 @@ -59,6 +59,13 @@ /* Memory Map for ADSP-BF537 processors */ +#ifdef CONFIG_BLKFIN_CACHE +#define BLKFIN_ICACHESIZE (16*1024) +#else +#define BLKFIN_ICACHESIZE (0*1024) +#endif + + #ifdef CONFIG_BF537 #define L1_CODE_START 0xFFA00000 #define L1_DATA_A_START 0xFF800000 @@ -67,20 +74,30 @@ #define L1_CODE_LENGTH 0xC000 #ifdef CONFIG_BLKFIN_DCACHE -#define DMEM_CNTR (ACACHE_BCACHE | ENDCPLB | PORT_PREF0) + +#ifdef CONFIG_BLKFIN_DCACHE_BANKA +#define DMEM_CNTR (ACACHE_BSRAM | ENDCPLB | PORT_PREF0) #define L1_DATA_A_LENGTH (0x8000 - 0x4000) +#define L1_DATA_B_LENGTH 0x8000 +#define BLKFIN_DCACHESIZE (16*1024) +#define BLKFIN_DSUPBANKS 1 #else -#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) -#define L1_DATA_A_LENGTH 0x8000 +#define DMEM_CNTR (ACACHE_BCACHE | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH (0x8000 - 0x4000) +#define L1_DATA_B_LENGTH (0x8000 - 0x4000) +#define BLKFIN_DCACHESIZE (32*1024) +#define BLKFIN_DSUPBANKS 2 #endif -#ifdef CONFIG_BLKFIN_DCACHE -#define L1_DATA_B_LENGTH (0x8000 - 0x4000) #else +#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH 0x8000 #define L1_DATA_B_LENGTH 0x8000 -#endif +#define BLKFIN_DCACHESIZE (0*1024) +#define BLKFIN_DSUPBANKS 0 +#endif /*CONFIG_BLKFIN_DCACHE*/ -#endif +#endif /*CONFIG_BF537*/ /* Memory Map for ADSP-BF536 processors */ @@ -91,19 +108,31 @@ #define L1_CODE_LENGTH 0xC000 + #ifdef CONFIG_BLKFIN_DCACHE -#define DMEM_CNTR (ACACHE_BSRAM | ENDCPLB) + +#ifdef CONFIG_BLKFIN_DCACHE_BANKA +#define DMEM_CNTR (ACACHE_BSRAM | ENDCPLB | PORT_PREF0) #define L1_DATA_A_LENGTH (0x4000 - 0x4000) +#define L1_DATA_B_LENGTH 0x4000 +#define BLKFIN_DCACHESIZE (16*1024) +#define BLKFIN_DSUPBANKS 1 + #else -#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB) -#define L1_DATA_A_LENGTH 0x4000 +#define DMEM_CNTR (ACACHE_BCACHE | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH (0x4000 - 0x4000) +#define L1_DATA_B_LENGTH (0x4000 - 0x4000) +#define BLKFIN_DCACHESIZE (32*1024) +#define BLKFIN_DSUPBANKS 2 #endif -#ifdef CONFIG_BLKFIN_DCACHE -#define L1_DATA_B_LENGTH (0x4000 - 0x4000) #else +#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH 0x4000 #define L1_DATA_B_LENGTH 0x4000 -#endif +#define BLKFIN_DCACHESIZE (0*1024) +#define BLKFIN_DSUPBANKS 0 +#endif /*CONFIG_BLKFIN_DCACHE*/ #endif @@ -118,18 +147,28 @@ #ifdef CONFIG_BLKFIN_DCACHE -#define DMEM_CNTR (ACACHE_BCACHE | ENDCPLB | PORT_PREF0) +#ifdef CONFIG_BLKFIN_DCACHE_BANKA +#define DMEM_CNTR (ACACHE_BSRAM | ENDCPLB | PORT_PREF0) #define L1_DATA_A_LENGTH (0x8000 - 0x4000) +#define L1_DATA_B_LENGTH 0x8000 +#define BLKFIN_DCACHESIZE (16*1024) +#define BLKFIN_DSUPBANKS 1 + #else -#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) -#define L1_DATA_A_LENGTH 0x8000 -#endif -#ifdef CONFIG_BLKFIN_DCACHE +#define DMEM_CNTR (ACACHE_BCACHE | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH (0x8000 - 0x4000) #define L1_DATA_B_LENGTH (0x8000 - 0x4000) +#define BLKFIN_DCACHESIZE (32*1024) +#define BLKFIN_DSUPBANKS 2 +#endif #else +#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH 0x8000 #define L1_DATA_B_LENGTH 0x8000 -#endif +#define BLKFIN_DCACHESIZE (0*1024) +#define BLKFIN_DSUPBANKS 0 +#endif /*CONFIG_BLKFIN_DCACHE*/ #endif diff -uprN linux-2.6-orig/include/asm-blackfin/mach-bf561/bf561.h linux-2.6/include/asm-blackfin/mach-bf561/bf561.h --- linux-2.6-orig/include/asm-blackfin/mach-bf561/bf561.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/mach-bf561/bf561.h 2007-03-05 13:54:10.000000000 +0800 @@ -74,14 +74,11 @@ * Blackfin Cache setup */ -#define BLKFIN_ICACHESIZE (16*1024) -#define BLKFIN_DCACHESIZE (32*1024) #define BLKFIN_ISUBBANKS 4 #define BLKFIN_IWAYS 4 #define BLKFIN_ILINES 32 -#define BLKFIN_DSUPBANKS 2 #define BLKFIN_DSUBBANKS 4 #define BLKFIN_DWAYS 2 #define BLKFIN_DLINES 64 @@ -351,7 +348,7 @@ #endif -#define L1_DMEMORY (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_DIRTY) +#define L1_DMEMORY (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_LOCK | CPLB_DIRTY) #define SDRAM_DNON_CHBL (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_DIRTY) #define SDRAM_EBIU (CPLB_SUPV_WR | CPLB_USER_WR | CPLB_USER_RD | CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_DIRTY) #define SDRAM_OOPS (CPLB_VALID | ANOMALY_05000158_WORKAROUND | CPLB_LOCK | CPLB_DIRTY) diff -uprN linux-2.6-orig/include/asm-blackfin/mach-bf561/mem_map.h linux-2.6/include/asm-blackfin/mach-bf561/mem_map.h --- linux-2.6-orig/include/asm-blackfin/mach-bf561/mem_map.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/mach-bf561/mem_map.h 2007-03-05 11:57:23.000000000 +0800 @@ -23,6 +23,12 @@ /* Level 1 Memory */ +#ifdef CONFIG_BLKFIN_CACHE +#define BLKFIN_ICACHESIZE (16*1024) +#else +#define BLKFIN_ICACHESIZE (0*1024) +#endif + /* Memory Map for ADSP-BF561 processors */ #ifdef CONFIG_BF561 @@ -33,18 +39,28 @@ #define L1_CODE_LENGTH 0x4000 #ifdef CONFIG_BLKFIN_DCACHE -#define DMEM_CNTR (ACACHE_BCACHE | ENDCPLB | PORT_PREF0) + +#ifdef CONFIG_BLKFIN_DCACHE_BANKA +#define DMEM_CNTR (ACACHE_BSRAM | ENDCPLB | PORT_PREF0) #define L1_DATA_A_LENGTH (0x8000 - 0x4000) +#define L1_DATA_B_LENGTH 0x8000 +#define BLKFIN_DCACHESIZE (16*1024) +#define BLKFIN_DSUPBANKS 1 #else -#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) -#define L1_DATA_A_LENGTH 0x8000 +#define DMEM_CNTR (ACACHE_BCACHE | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH (0x8000 - 0x4000) +#define L1_DATA_B_LENGTH (0x8000 - 0x4000) +#define BLKFIN_DCACHESIZE (32*1024) +#define BLKFIN_DSUPBANKS 2 #endif -#ifdef CONFIG_BLKFIN_DCACHE -#define L1_DATA_B_LENGTH (0x8000 - 0x4000) #else +#define DMEM_CNTR (ASRAM_BSRAM | ENDCPLB | PORT_PREF0) +#define L1_DATA_A_LENGTH 0x8000 #define L1_DATA_B_LENGTH 0x8000 -#endif +#define BLKFIN_DCACHESIZE (0*1024) +#define BLKFIN_DSUPBANKS 0 +#endif /*CONFIG_BLKFIN_DCACHE*/ #endif /* Level 2 Memory */ diff -uprN linux-2.6-orig/include/asm-blackfin/pgtable.h linux-2.6/include/asm-blackfin/pgtable.h --- linux-2.6-orig/include/asm-blackfin/pgtable.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/pgtable.h 2007-03-05 11:57:23.000000000 +0800 @@ -59,4 +59,12 @@ extern unsigned int kobjsize(const void #define VMALLOC_START 0 #define VMALLOC_END 0xffffffff +#define __HAVE_ARCH_ENTER_LAZY_CPU_MODE +#define arch_enter_lazy_cpu_mode() do {} while (0) +#define arch_leave_lazy_cpu_mode() do {} while (0) + +#define __HAVE_ARCH_ENTER_LAZY_MMU_MODE +#define arch_enter_lazy_mmu_mode() do {} while (0) +#define arch_leave_lazy_mmu_mode() do {} while (0) + #endif /* _BLACKFIN_PGTABLE_H */ diff -uprN linux-2.6-orig/include/asm-blackfin/uaccess.h linux-2.6/include/asm-blackfin/uaccess.h --- linux-2.6-orig/include/asm-blackfin/uaccess.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/uaccess.h 2007-03-05 11:57:23.000000000 +0800 @@ -41,48 +41,15 @@ extern int is_in_rom(unsigned long); * get_fs() == KERNEL_DS, checking is bypassed. */ -static inline int _access_ok(unsigned long addr, unsigned long size) -{ #ifdef CONFIG_NO_ACCESS_CHECK - return 1; +static inline int _access_ok(unsigned long addr, unsigned long size) { return 1; } #else - if (addr > (addr + size)) - return 0; - if (segment_eq(get_fs(),KERNEL_DS)) - return 1; -#ifdef CONFIG_MTD_UCLINUX - if (addr >= memory_start && (addr + size) <= memory_end) - return 1; - if (addr >= memory_mtd_end && (addr + size) <= physical_mem_end) - return 1; +#ifdef CONFIG_ACCESS_OK_L1 +extern int _access_ok(unsigned long addr, unsigned long size)__attribute__((l1_text)); #else - if (addr >= memory_start && (addr + size) <= physical_mem_end) - return 1; -#endif - if (addr >= (unsigned long)__init_begin && - addr + size <= (unsigned long)__init_end) - return 1; - if (addr >= L1_SCRATCH_START - && addr + size <= L1_SCRATCH_START + L1_SCRATCH_LENGTH) - return 1; -#if L1_CODE_LENGTH != 0 - if (addr >= L1_CODE_START + (_etext_l1 - _stext_l1) - && addr + size <= L1_CODE_START + L1_CODE_LENGTH) - return 1; -#endif -#if L1_DATA_A_LENGTH != 0 - if (addr >= L1_DATA_A_START + (_ebss_l1 - _sdata_l1) - && addr + size <= L1_DATA_A_START + L1_DATA_A_LENGTH) - return 1; +extern int _access_ok(unsigned long addr, unsigned long size); #endif -#if L1_DATA_B_LENGTH != 0 - if (addr >= L1_DATA_B_START - && addr + size <= L1_DATA_B_START + L1_DATA_B_LENGTH) - return 1; #endif - return 0; -#endif -} /* * The exception table consists of pairs of addresses: the first is the diff -uprN linux-2.6-orig/include/asm-blackfin/unistd.h linux-2.6/include/asm-blackfin/unistd.h --- linux-2.6-orig/include/asm-blackfin/unistd.h 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/include/asm-blackfin/unistd.h 2007-03-05 11:57:23.000000000 +0800 @@ -188,7 +188,7 @@ #define __NR_getcwd 183 #define __NR_capget 184 #define __NR_capset 185 - /* 186 __NR_sigaltstack obsolete */ +#define __NR_sigaltstack 186 #define __NR_sendfile 187 /* 188 __NR_getpmsg */ /* 189 __NR_putpmsg */ diff -uprN linux-2.6-orig/include/asm-blackfin/xor.h linux-2.6/include/asm-blackfin/xor.h --- linux-2.6-orig/include/asm-blackfin/xor.h 1970-01-01 08:00:00.000000000 +0800 +++ linux-2.6/include/asm-blackfin/xor.h 2007-03-05 11:57:23.000000000 +0800 @@ -0,0 +1 @@ +#include <asm-generic/xor.h> diff -uprN linux-2.6-orig/init/Kconfig linux-2.6/init/Kconfig --- linux-2.6-orig/init/Kconfig 2007-03-05 11:50:22.000000000 +0800 +++ linux-2.6/init/Kconfig 2007-03-05 11:57:23.000000000 +0800 @@ -336,7 +336,7 @@ menuconfig EMBEDDED config UID16 bool "Enable 16-bit UID system calls" if EMBEDDED - depends on ARM || CRIS || FRV || H8300 || X86_32 || M68K || (S390 && !64BIT) || SUPERH || SPARC32 || (SPARC64 && SPARC32_COMPAT) || UML || (X86_64 && IA32_EMULATION) || BFIN + depends on ARM || BFIN || CRIS || FRV || H8300 || X86_32 || M68K || (S390 && !64BIT) || SUPERH || SPARC32 || (SPARC64 && SPARC32_COMPAT) || UML || (X86_64 && IA32_EMULATION) default y help This enables the legacy 16-bit UID syscall wrappers. @@ -443,7 +443,6 @@ config FUTEX config EPOLL bool "Enable eventpoll support" if EMBEDDED default y - depends on MMU help Disabling this option will cause the kernel to be built without support for epoll family of system calls. _ Thanks -Bryan Wu ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-01 4:14 [PATCH -mm 1/5] Blackfin: blackfin architecture patch update Wu, Bryan 2007-03-03 20:38 ` Arnd Bergmann @ 2007-03-03 22:30 ` Arnd Bergmann 2007-03-03 22:50 ` bert hubert ` (3 more replies) 2007-03-05 9:23 ` Paul Mundt 2 siblings, 4 replies; 24+ messages in thread From: Arnd Bergmann @ 2007-03-03 22:30 UTC (permalink / raw) To: bryan.wu; +Cc: Andrew Morton, linux-kernel On Thursday 01 March 2007 05:14:40 Wu, Bryan wrote: > Here is the update version of blackfin-arch.patch in -mm tree. > simply add support to utrace and it was tested on blackfin STAMP board > as well as other following patches. Wow, this has come a long way since I looked at the patches last year, good work! I've gone through the complete patch again now, and these are the issues I've found in it. None of these are show-stoppers and I'd like to see it all go in during the next merge window. There should be enough time until then to address these points: > +EXPORT_SYMBOL(__ioremap); > +EXPORT_SYMBOL(strcmp); > +EXPORT_SYMBOL(strncmp); > +EXPORT_SYMBOL(dump_thread); > + > +EXPORT_SYMBOL(ip_fast_csum); > + > +EXPORT_SYMBOL(kernel_thread); > + > +EXPORT_SYMBOL(__up); > +EXPORT_SYMBOL(__down); > +EXPORT_SYMBOL(__down_trylock); > +EXPORT_SYMBOL(__down_interruptible); > + > +EXPORT_SYMBOL(is_in_rom); In general, please put EXPORT_SYMBOL lines below the definition of the symbol itself. This list of exports should only be used for symbols that come from assembly files. You should probably also think about whether some of them are better done as EXPORT_SYMBOL_GPL. > + pending = bfin_read_IPEND() & ~0x8000; > + other_ints = pending & (pending - 1); > + if (other_ints == 0) > + lower_to_irq14(); > + irq_exit(); > + The last line here has trailing whitespace. While this gets automatically removed by akpm's scripts, you're normally better off not adding it in the first place, because it may cause your follow-on patches not to apply, aside from being wrong to start with. > +void machine_halt(void) > +{ > + for (;;) > + /* nothing */ ; > +} > + > +void machine_power_off(void) > +{ > + for (;;) > + /* nothing */ ; > +} It might be nicer to make this for (;;) asm volatile ("idle"); Otherwise you end up burning CPU cycles after a halt without any particular need. > +#if defined(CONFIG_MTD_UCLINUX) > + /* generic memory mapped MTD driver */ > + memory_mtd_end = memory_end; > + > + mtd_phys = _ramstart; > + mtd_size = PAGE_ALIGN(*((unsigned long *)(mtd_phys + 8))); > + > +# if defined(CONFIG_EXT2_FS) || defined(CONFIG_EXT3_FS) > + if (*((unsigned short *)(mtd_phys + 0x438)) == EXT2_SUPER_MAGIC) > + mtd_size = > + PAGE_ALIGN(*((unsigned long *)(mtd_phys + 0x404)) << 10); > +# endif > + > +# if defined(CONFIG_CRAMFS) > + if (*((unsigned long *)(mtd_phys)) == CRAMFS_MAGIC) > + mtd_size = PAGE_ALIGN(*((unsigned long *)(mtd_phys + 0x4))); > +# endif > + > +# if defined(CONFIG_ROMFS_FS) > + if (((unsigned long *)mtd_phys)[0] == ROMSB_WORD0 > + && ((unsigned long *)mtd_phys)[1] == ROMSB_WORD1) > + mtd_size = > + PAGE_ALIGN(be32_to_cpu(((unsigned long *)mtd_phys)[2])); This detection seems to me like a strange thing to do in setup_arch(). It should be possible to do this much later, at a point where the system is much less fragile and e.g. printk works. It could even be moved into some place in the mtd code itself, since other architectures might want to do the same thing. > +#if defined(CONFIG_BF561) > +static struct cpu cpu[2]; > +#else > +static struct cpu cpu[1]; > +#endif > +static int __init topology_init(void) > +{ > +#if defined (CONFIG_BF561) > + register_cpu(&cpu[0], 0); > + register_cpu(&cpu[1], 1); > + return 0; > +#else > + return register_cpu(cpu, 0); > +#endif > +} I think you should try to avoid the special-case stuff here. You can have CONFIG_NR_CPUS in Kconfig set dependent on CONFIG_BF561 and change the code here (and similarly in other places) to static struct cpu cpu[NR_CPUS]; static int __init topology_init(void) { int i; for (i=0; i< NR_CPUS; i++) { register_cpu(&cpu[i], i); return 0; } > + for (i = ZERO_P; i <= L2_MEM; i++) { > + > + if (cplb_data[i].valid) { > + > + as_1m = cplb_data[i].start % SIZE_1M; > + > + /* We need to make sure all sections are properly 1M aligned > + * However between Kernel Memory and the Kernel mtd section, depending on the > + * rootfs size, there can be overlapping memory areas. > + */ > + > + if (as_1m) { > +#ifdef CONFIG_MTD_UCLINUX > + if (i == SDRAM_RAM_MTD) { > + if ((cplb_data[SDRAM_KERN].end + 1) > cplb_data[SDRAM_RAM_MTD].start) > + cplb_data[SDRAM_RAM_MTD].start = (cplb_data[i].start & (-2*SIZE_1M)) + SIZE_1M; I count 6 levels of indentation, which severely limits readability, especially when you have terms this complex in the last level. Please try to split up functions like this into smaller units. > +/* > + * ++roman (07/09/96): implemented signal stacks (specially for tosemu on > + * Atari :-) Current limitation: Only one sigstack can be active at one time. > + * If a second signal with SA_ONSTACK set arrives while working on a sigstack, > + * SA_ONSTACK is ignored. This behaviour avoids lots of trouble with nested > + * signal handlers! > + */ This comment is probably outdated. > +extern int setup_irq(unsigned int irq, struct irqaction *handler); this extern should be in a header. > +asmlinkage void trap_c(struct pt_regs *fp) > +{ > + int j, sig = 0; > + siginfo_t info; > + unsigned long trapnr = fp->seqstat & SEQSTAT_EXCAUSE; > + > +#ifdef CONFIG_KGDB > +# define CHK_DEBUGGER_TRAP() do { CHK_DEBUGGER(trapnr, sig, info.si_code, fp,); } while (0) > +# define CHK_DEBUGGER_TRAP_MAYBE() do { if (kgdb_connected) CHK_DEBUGGER_TRAP(); } while (0) > +#else > +# define CHK_DEBUGGER_TRAP() do { } while (0) > +# define CHK_DEBUGGER_TRAP_MAYBE() do { } while (0) > +#endif > + > + trace_buffer_save(j); > + > + /* trap_c() will be called for exceptions. During exceptions > + * processing, the pc value should be set with retx value. > + * With this change we can cleanup some code in signal.c- TODO > + */ > + fp->orig_pc = fp->retx; > + /* printk("exception: 0x%x, ipend=%x, reti=%x, retx=%x\n", > + trapnr, fp->ipend, fp->pc, fp->retx); */ > + > + /* send the appropriate signal to the user program */ > + switch (trapnr) { > + > + /* This table works in conjuction with the one in ./mach-common/entry.S > + * Some exceptions are handled there (in assembly, in exception space) > + * Some are handled here, (in C, in interrupt space) > + * Some, like CPLB, are handled in both, where the normal path is > + * handled in assembly/exception space, and the error path is handled > + * here > + */ > + > + /* 0x00 - Linux Syscall, getting here is an error */ > + /* 0x01 - userspace gdb breakpoint, handled here */ > + case VEC_EXCPT01: > + info.si_code = TRAP_ILLTRAP; > + sig = SIGTRAP; > + CHK_DEBUGGER_TRAP_MAYBE(); > + /* Check if this is a breakpoint in kernel space */ > + if (fp->ipend & 0xffc0) > + return; > + else > + break; > +#ifdef CONFIG_KGDB > + case VEC_EXCPT02 : /* gdb connection */ > + info.si_code = TRAP_ILLTRAP; > + sig = SIGTRAP; > + CHK_DEBUGGER_TRAP(); > + return; > +#else > + /* 0x02 - User Defined, Caught by default */ > +#endif > + /* 0x03 - Atomic test and set */ > + case VEC_EXCPT03: The table here probably gets converted by gcc into a lookup table, but I think it would be clearer to make that explicit and force the table here with something like: static void trap_breakpoint(struct pt_regs *fp, siginfo_t *info, unsigned int trapnr) { info->si_code = TRAP_ILLTRAP; CHK_DEBUGGER_TRAP_MAYBE(); if (fp->ipend & 0xffc0) return 0; return SIGTRAP; } #ifdef CONFIG_KGDB static void trap_kgdb(struct pt_regs *fp, siginfo_t *info, unsigned int trapnr) { info->si_code = TRAP_ILLTRAP; CHK_DEBUGGER_TRAP(); return SIGTRAP; } #endif static void (traps[])(struct pt_regs *fp, siginfo_t *info, unsigned int trapnr) = { [VEC_EXCPT01] &trap_breakpoint, /* 0x01 - userspace gdb breakpoint */ #ifdef CONFIG_KGDB [VEC_EXCPT02] &trap_kgdb, /* gdb connection */ #endif ... }; asmlinkage void trap_c(struct pt_regs *fp) { ... if (trapnr > 0x3f) goto error; if (!traps[trapnr]) goto out; ret = traps[trapno](fp, &info, trapnr); if (!ret) goto out; ... That should make sure you get the efficient trap handling almost and might even save some code space (you'd have to try that). > +asmlinkage int sys_bfin_spinlock(int *spinlock) > +{ > + int ret = 0; > + int tmp = 0; > + > + local_irq_disable(); > + ret = get_user(tmp, spinlock); > + if (ret == 0) { > + if (tmp) > + ret = 1; > + tmp = 1; > + put_user(tmp, spinlock); > + } > + local_irq_enable(); > + return ret; > +} I'm curious: In your dual-core bf561, don't you actually need to implement something that maintains atomicity across cores rather than just across processes? > +#include <net/checksum.h> > +#include <asm/checksum.h> > + > +#ifdef CONFIG_IP_CHECKSUM_L1 > +static unsigned short do_csum(const unsigned char *buff, int len)__attribute__((l1_text)); > +#endif > + > +static unsigned short do_csum(const unsigned char *buff, int len) > +{ > + register unsigned long sum = 0; > + int swappem = 0; > + > + if (1 & (unsigned long)buff) { > + sum = *buff << 8; > + buff++; > + len--; > + ++swappem; > + } For the non-L1 case, is there actually anything specific to blackfin in this code? I wonder if we should instead have a totally generic version of that in lib/csum.c for those architectures that don't want their own optimizations for this. > +static struct platform_device *stamp_devices[] __initdata = { > +#if defined(CONFIG_RTC_DRV_BFIN) || defined(CONFIG_RTC_DRV_BFIN_MODULE) > + &rtc_device, > +#endif > + > +#if defined(CONFIG_BFIN_CFPCMCIA) || defined(CONFIG_BFIN_CFPCMCIA_MODULE) > + &bfin_pcmcia_cf_device, > +#endif > + > +#if defined(CONFIG_USB_SL811_HCD) || defined(CONFIG_USB_SL811_HCD_MODULE) > + &sl811_hcd_device, > +#endif This array reminds me of the bad old days when all initialization functions had to be called from a common function. Normally I don't like to recommend using section magic for anything, but in this case I guess you'd be better off if you just do your own thing here, like: #define BFIN_PLATFORM_DEVICE(name) struct platform_device name \ __attribute((section("initdata.platformdev"))) then declare your platform devices in the respective driver, like static BFIN_PLATFORM_DEVICE(rtc_device) = { ... }; and loop through the new section like you do for the initcalls. > +void get_bf537_ether_addr(char *addr) > +{ > + /* currently the mac addr is saved in flash */ > + int flash_mac = 0x203f0000; > + *(u32 *)(&(addr[0])) = *(int *)flash_mac; > + flash_mac += 4; > + *(u16 *)(&(addr[4])) = (u16)*(int *)flash_mac; > +} > + > +EXPORT_SYMBOL(get_bf537_ether_addr); This looks wrong in a number of aspects. Are there other things stored in the flash so you can make this a more general flash abstraction? Is there more than one driver using this at all? If not, it should probably be part of the driver, with the exact address being a compile-time constant for that driver. It should probably also use bfin_readXX instead direct dereferences. > +/* The number of spurious interrupts */ > +volatile unsigned int num_spurious; Why volatile? It probably does not what you intended, maybe it should be atomic_t instead. > +/* BASE LEVEL interrupt handler routines */ > +asmlinkage void evt_emulation(void); > +asmlinkage void evt_exception(void); > +asmlinkage void trap(void); > +asmlinkage void evt_ivhw(void); > +asmlinkage void evt_timer(void); > +asmlinkage void evt_evt2(void); > +asmlinkage void evt_evt7(void); > +asmlinkage void evt_evt8(void); > +asmlinkage void evt_evt9(void); > +asmlinkage void evt_evt10(void); > +asmlinkage void evt_evt11(void); > +asmlinkage void evt_evt12(void); > +asmlinkage void evt_evt13(void); > +asmlinkage void evt_soft_int1(void); > +asmlinkage void evt_system_call(void); > +asmlinkage void init_exception_buff(void); move the extern definitions to a header file please > +/* BASE LEVEL interrupt handler routines */ > +asmlinkage void evt_emulation(void); > +asmlinkage void evt_exception(void); > +asmlinkage void trap(void); > +asmlinkage void evt_ivhw(void); > +asmlinkage void evt_timer(void); > +asmlinkage void evt_evt2(void); > +asmlinkage void evt_evt7(void); > +asmlinkage void evt_evt8(void); > +asmlinkage void evt_evt9(void); > +asmlinkage void evt_evt10(void); > +asmlinkage void evt_evt11(void); > +asmlinkage void evt_evt12(void); > +asmlinkage void evt_evt13(void); > +asmlinkage void evt_soft_int1(void); > +asmlinkage void evt_system_call(void); Then you don't have to write them twice ;-) > +static __inline__ void atomic_add(int i, atomic_t * v) > +{ > + long flags; > + > + local_irq_save(flags); > + v->counter += i; > + local_irq_restore(flags); > +} > + > +static __inline__ void atomic_sub(int i, atomic_t * v) > +{ > + long flags; > + > + local_irq_save(flags); > + v->counter -= i; > + local_irq_restore(flags); > + > +} The implementation of the atomic_t type looks very generic. Maybe it could be moved to asm-generic/atomic-irqdisable.h or similar to be shared by all architectures without atomic operations. > +struct spi_device_t { > + char *dev_name; > + > + unsigned short flag; > + unsigned short bdrate; > + > + unsigned short enable; > + unsigned short master; > + unsigned short out_opendrain; > + unsigned short polar; > + unsigned short phase; > + unsigned short byteorder; /* 0: MSB first; 1: LSB first; */ > + unsigned short size; /* 0: 8 bits; 1: 16 bits */ > + unsigned short emiso; > + unsigned short send_zero; > + unsigned short more_data; > + unsigned short slave_sel; > + unsigned short ti_mod; > + > + unsigned short dma; /* use dma mode or not */ > + unsigned short dma_config; /* only valid if dma enabled */ > + > + irqreturn_t(*irq_handler) (int irq, void *dev_id, > + struct pt_regs *regs); > + void *priv_data; > +}; How does this fit in with the generic SPI code? Does it duplicate stuff from there, or do you use it? > +static __inline__ void set_bit(int nr, volatile unsigned long *addr) > +{ > + int *a = (int *)addr; > + int mask; > + unsigned long flags; > + > + a += nr >> 5; > + mask = 1 << (nr & 0x1f); > + local_irq_save(flags); > + *a |= mask; > + local_irq_restore(flags); > +} > + > +static __inline__ void __set_bit(int nr, volatile unsigned long *addr) > +{ > + int *a = (int *)addr; > + int mask; > + > + a += nr >> 5; > + mask = 1 << (nr & 0x1f); > + *a |= mask; > +} same as for the atomic header, this could become generic. > +#if defined(ANOMALY_05000312) && defined(ANOMALY_05000244) > +#define SSYNC() do { int _tmp; \ > + __asm__ __volatile__ ("cli %0;\n\t"\ > + "nop;nop;\n\t"\ > + "ssync;\n\t"\ > + "sti %0;\n\t" \ > + :"=d"(_tmp):);\ > + } while (0) > +#elif defined(ANOMALY_05000312) && !defined(ANOMALY_05000244) > +#define SSYNC() do { int _tmp; \ > + __asm__ __volatile__ ("cli %0;\n\t"\ > + "ssync;\n\t"\ > + "sti %0;\n\t" \ > + :"=d"(_tmp):);\ > + } while (0) > +#elif !defined(ANOMALY_05000312) && defined(ANOMALY_05000244) > +#define SSYNC() do {__builtin_bfin_ssync();} while (0) > +#elif !defined(ANOMALY_05000312) && !defined(ANOMALY_05000244) > +#define SSYNC() do {__asm__ __volatile__ ("ssync;\n\t") } while (0) > +#endif > + > + > +#if defined(ANOMALY_05000312) && defined(ANOMALY_05000244) > +#define CSYNC() do { int _tmp; \ > + __asm__ __volatile__ ("cli %0;\n\t"\ > + "nop;nop;\n\t"\ > + "csync;\n\t"\ > + "sti %0;\n\t" \ > + :"=d"(_tmp):);\ > + } while (0) > +#elif defined(ANOMALY_05000312) && !defined(ANOMALY_05000244) > +#define CSYNC() do { int _tmp; \ > + __asm__ __volatile__ ("cli %0;\n\t"\ > + "csync;\n\t"\ > + "sti %0;\n\t" \ > + :"=d"(_tmp):);\ > + } while (0) > +#elif !defined(ANOMALY_05000312) && defined(ANOMALY_05000244) > +#define CSYNC() do {__builtin_bfin_csync();} while (0) > +#elif !defined(ANOMALY_05000312) && !defined(ANOMALY_05000244) > +#define CSYNC() do {__asm__ __volatile__ ("csync;\n\t") } while (0) > +#endif it would be nice if you could convert these to inline functions instead of macros. > +static inline struct task_struct *get_current(void) __attribute__ ((__const__)); > +static inline struct task_struct *get_current(void) > +{ > + return (current_thread_info()->task); > +} > + > +#define current (get_current()) You might want to maintain the current variable in the L1 memory, but probably that difference it not measurable. > +#ifdef BFIN_DMA_NDEBUG > +#define assert(expr) do {} while(0) > +#else > +#define assert(expr) \ > + if (!(expr)) { \ > + printk(KERN_INFO "Assertion failed! %s, %s, %s, line=%d \n", \ > + #expr, __FILE__,__FUNCTION__,__LINE__); \ > + } > +#endif Please don't define your own assert() functions, in the kernel we have BUG_ON and WARN_ON for this. > +#ifndef _BLACKFIN_DPMC_H_ > +#define _BLACKFIN_DPMC_H_ > + > +#define SLEEP_MODE 1 > +#define DEEP_SLEEP_MODE 2 > +#define ACTIVE_PLL_DISABLED 3 > +#define FULLON_MODE 4 > +#define ACTIVE_PLL_ENABLED 5 > +#define HIBERNATE_MODE 6 > + > +#define IOCTL_FULL_ON_MODE _IO('s', 0xA0) > +#define IOCTL_ACTIVE_MODE _IO('s', 0xA1) > +#define IOCTL_SLEEP_MODE _IO('s', 0xA2) > +#define IOCTL_DEEP_SLEEP_MODE _IO('s', 0xA3) > +#define IOCTL_HIBERNATE_MODE _IO('s', 0xA4) > +#define IOCTL_CHANGE_FREQUENCY _IOW('s', 0xA5, unsigned long) > +#define IOCTL_CHANGE_VOLTAGE _IOW('s', 0xA6, unsigned long) > +#define IOCTL_SET_CCLK _IOW('s', 0xA7, unsigned long) > +#define IOCTL_SET_SCLK _IOW('s', 0xA8, unsigned long) > +#define IOCTL_GET_PLLSTATUS _IOW('s', 0xA9, unsigned long) > +#define IOCTL_GET_CORECLOCK _IOW('s', 0xAA, unsigned long) > +#define IOCTL_GET_SYSTEMCLOCK _IOW('s', 0xAB, unsigned long) > +#define IOCTL_GET_VCO _IOW('s', 0xAC, unsigned long) > +#define IOCTL_DISABLE_WDOG_TIMER _IO('s', 0xAD) > +#define IOCTL_UNMASK_WDOG_WAKEUP_EVENT _IO('s',0xAE) > +#define IOCTL_PROGRAM_WDOG_TIMER _IOW('s',0xAF,unsigned long) > +#define IOCTL_CLEAR_WDOG_WAKEUP_EVENT _IO('s',0xB0) > +#define IOCTL_SLEEP_DEEPER_MODE _IO('s',0xB1) > + > +#define DPMC_MINOR 254 > + > +#define ON 0 > +#define OFF 1 > + > +unsigned long calc_volt(void); > +int calc_vlev(int vlt); > +unsigned long change_voltage(unsigned long volt); > +int calc_msel(int vco_hz); > +unsigned long change_frequency(unsigned long vco_mhz); > +int set_pll_div(unsigned short sel, unsigned char flag); In general, you should be careful not to mix in-kernel declarations with user-visible parts. In this case, the ioctl definitions need to be exported via 'make headers_install', while the function declarations clearly are not useful for user space and should therefore be inside of '#ifdef __KERNEL__'. You should then list this file as 'unifdef-y' in the Kbuild file. > + > +#ifdef __KERNEL__ > +#define SET_PERSONALITY(ex, ibcs2) set_personality((ibcs2)?PER_SVR4:PER_LINUX) > +#endif This looks bogus, I don't think you implement personalities. > +#define readb(addr) ({ unsigned __v; \ > + int _tmp; \ > + __asm__ __volatile__ ("cli %1;\n\t"\ > + "NOP;NOP;SSYNC;\n\t"\ > + "%0 = b [%2] (z);\n\t"\ > + "sti %1;\n\t" \ > + : "=d"(__v), "=d"(_tmp): "a"(addr)); (unsigned char)__v; }) > + > +#define readw(addr) ({ unsigned __v; \ > + int _tmp; \ > + __asm__ __volatile__ ("cli %1;\n\t"\ > + "NOP;NOP;SSYNC;\n\t"\ > + "%0 = w [%2] (z);\n\t"\ > + "sti %1;\n\t" \ > + : "=d"(__v), "=d"(_tmp): "a"(addr)); (unsigned short)__v; }) They should probably be inline functions. > +#define TCGETS 0x5401 > +#define TCSETS 0x5402 > +#define TCSETSW 0x5403 > +#define TCSETSF 0x5404 > +#define TCGETA 0x5405 > +#define TCSETA 0x5406 > +#define TCSETAW 0x5407 > +#define TCSETAF 0x5408 > +#define TCSBRK 0x5409 > +#define TCXONC 0x540A > +#define TCFLSH 0x540B > +#define TIOCEXCL 0x540C > +#define TIOCNXCL 0x540D > +#define TIOCSCTTY 0x540E > +#define TIOCGPGRP 0x540F > +#define TIOCSPGRP 0x5410 > +#define TIOCOUTQ 0x5411 > +#define TIOCSTI 0x5412 > +#define TIOCGWINSZ 0x5413 > +#define TIOCSWINSZ 0x5414 > +#define TIOCMGET 0x5415 > +#define TIOCMBIS 0x5416 > +#define TIOCMBIC 0x5417 > +#define TIOCMSET 0x5418 > +#define TIOCGSOFTCAR 0x5419 > +#define TIOCSSOFTCAR 0x541A > +#define FIONREAD 0x541B > +#define TIOCINQ FIONREAD > +#define TIOCLINUX 0x541C > +#define TIOCCONS 0x541D > +#define TIOCGSERIAL 0x541E > +#define TIOCSSERIAL 0x541F > +#define TIOCPKT 0x5420 > +#define FIONBIO 0x5421 > +#define TIOCNOTTY 0x5422 > +#define TIOCSETD 0x5423 > +#define TIOCGETD 0x5424 > +#define TCSBRKP 0x5425 /* Needed for POSIX tcsendbreak() */ > +#define TIOCTTYGSTRUCT 0x5426 /* For debugging only */ > +#define TIOCSBRK 0x5427 /* BSD compatibility */ > +#define TIOCCBRK 0x5428 /* BSD compatibility */ > +#define TIOCGSID 0x5429 /* Return the session ID of FD */ > +#define TIOCGPTN _IOR('T',0x30, unsigned int) /* Get Pty Number (of pty-mux device) */ > +#define TIOCSPTLCK _IOW('T',0x31, int) /* Lock/unlock Pty */ These look like another good candidate for an asm-generic version. > + > +extern void sys_free_irq(unsigned int irq, void *dev_id); This functions should probably be renamed. Anything that starts with 'sys_' is normally a system call. + > +/* Clock and System Control (0xFFC0 0400-0xFFC0 07FF) */ > +#define bfin_read_PLL_CTL() bfin_read16(PLL_CTL) > +#define bfin_write_PLL_CTL(val) bfin_write16(PLL_CTL,val) > +#define bfin_read_PLL_STAT() bfin_read16(PLL_STAT) > +#define bfin_write_PLL_STAT(val) bfin_write16(PLL_STAT,val) > +#define bfin_read_PLL_LOCKCNT() bfin_read16(PLL_LOCKCNT) > +#define bfin_write_PLL_LOCKCNT(val) bfin_write16(PLL_LOCKCNT,val) > +#define bfin_read_CHIPID() bfin_read32(CHIPID) > +#define bfin_read_SWRST() bfin_read16(SWRST) > +#define bfin_write_SWRST(val) bfin_write16(SWRST,val) I remember that we have discussed these macro abstractions before, but don't recall the result of the discussion. You have around 5000 such macros, and I still think it's not a helpful abstraction. It is much more common for device drivers to just use the read/write accessors directly. IIRC the objections that were raised (and my replies to them) were roughly: > the driver writer doesn't want to know the size of the variable, and > if it changes, they need to change every instance in the code, not > just the macro. A driver writer should better know the type of variable he is changing, e.g. because of the type he passes in and out. Also, hardware registers don't suddenly change in size. > These macros allow us to work around hardware bugs when accessing the > registers by simply redefining the accessor to do something more > complex. If there is a bug in the hardware, the workaround belongs into the driver. You can then still define a special inline function to access that particular register. Arnd <>< ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-03 22:30 ` Arnd Bergmann @ 2007-03-03 22:50 ` bert hubert 2007-03-03 23:05 ` Arnd Bergmann 2007-03-05 6:54 ` Aubrey Li ` (2 subsequent siblings) 3 siblings, 1 reply; 24+ messages in thread From: bert hubert @ 2007-03-03 22:50 UTC (permalink / raw) To: Arnd Bergmann; +Cc: bryan.wu, Andrew Morton, linux-kernel On Sat, Mar 03, 2007 at 11:30:49PM +0100, Arnd Bergmann wrote: > It might be nicer to make this > > for (;;) > asm volatile ("idle"); This looks remarkably like relax_cpu() -- http://www.PowerDNS.com Open source, database driven DNS Software http://netherlabs.nl Open and Closed source services ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-03 22:50 ` bert hubert @ 2007-03-03 23:05 ` Arnd Bergmann 0 siblings, 0 replies; 24+ messages in thread From: Arnd Bergmann @ 2007-03-03 23:05 UTC (permalink / raw) To: bert hubert; +Cc: bryan.wu, Andrew Morton, linux-kernel On Saturday 03 March 2007 23:50:02 bert hubert wrote: > > for (;;) > > asm volatile ("idle"); > > This looks remarkably like relax_cpu() Actually not: cpu_relax() is defined as barrier(), it can't call idle because that might make it sleep for a indefinite amount of time (until the next interrupt, but only if they are enabled). Some nice architectures provide a hardware mechanism to do cpu_relax, like going to low-power mode for a few microseconds, but this one doesn't seem to have it. Arnd <>< ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-03 22:30 ` Arnd Bergmann 2007-03-03 22:50 ` bert hubert @ 2007-03-05 6:54 ` Aubrey Li 2007-03-05 8:47 ` Arnd Bergmann 2007-03-05 7:34 ` Wu, Bryan 2007-03-06 2:09 ` Mike Frysinger 3 siblings, 1 reply; 24+ messages in thread From: Aubrey Li @ 2007-03-05 6:54 UTC (permalink / raw) To: Arnd Bergmann; +Cc: bryan.wu, Andrew Morton, linux-kernel On 3/4/07, Arnd Bergmann <arnd@arndb.de> wrote: > On Thursday 01 March 2007 05:14:40 Wu, Bryan wrote: > > Here is the update version of blackfin-arch.patch in -mm tree. > > simply add support to utrace and it was tested on blackfin STAMP board > > as well as other following patches. > > Wow, this has come a long way since I looked at the patches last > year, good work! > > I've gone through the complete patch again now, and these are the > issues I've found in it. None of these are show-stoppers and I'd > like to see it all go in during the next merge window. There should > be enough time until then to address these points: > > > +EXPORT_SYMBOL(__ioremap); > > +EXPORT_SYMBOL(strcmp); > > +EXPORT_SYMBOL(strncmp); > > +EXPORT_SYMBOL(dump_thread); > > + > > +EXPORT_SYMBOL(ip_fast_csum); > > + > > +EXPORT_SYMBOL(kernel_thread); > > + > > +EXPORT_SYMBOL(__up); > > +EXPORT_SYMBOL(__down); > > +EXPORT_SYMBOL(__down_trylock); > > +EXPORT_SYMBOL(__down_interruptible); > > + > > +EXPORT_SYMBOL(is_in_rom); > > In general, please put EXPORT_SYMBOL lines below the definition > of the symbol itself. This list of exports should only be used > for symbols that come from assembly files. What is the right way to export symbol coming from c files? > > You should probably also think about whether some of them are > better done as EXPORT_SYMBOL_GPL. It is probably not. We need code using those symbols, no matter it's non-GPLd or GPLd. > > > + pending = bfin_read_IPEND() & ~0x8000; > > + other_ints = pending & (pending - 1); > > + if (other_ints == 0) > > + lower_to_irq14(); > > + irq_exit(); > > + > > The last line here has trailing whitespace. While this gets automatically > removed by akpm's scripts, you're normally better off not adding it > in the first place, because it may cause your follow-on patches not > to apply, aside from being wrong to start with. > > > +void machine_halt(void) > > +{ > > + for (;;) > > + /* nothing */ ; > > +} > > + > > +void machine_power_off(void) > > +{ > > + for (;;) > > + /* nothing */ ; > > +} > > It might be nicer to make this > > for (;;) > asm volatile ("idle"); > > Otherwise you end up burning CPU cycles after a halt without > any particular need. Good suggestion, thanks. > > > +#if defined(CONFIG_MTD_UCLINUX) > > + /* generic memory mapped MTD driver */ > > + memory_mtd_end = memory_end; > > + > > + mtd_phys = _ramstart; > > + mtd_size = PAGE_ALIGN(*((unsigned long *)(mtd_phys + 8))); > > + > > +# if defined(CONFIG_EXT2_FS) || defined(CONFIG_EXT3_FS) > > + if (*((unsigned short *)(mtd_phys + 0x438)) == EXT2_SUPER_MAGIC) > > + mtd_size = > > + PAGE_ALIGN(*((unsigned long *)(mtd_phys + 0x404)) << 10); > > +# endif > > + > > +# if defined(CONFIG_CRAMFS) > > + if (*((unsigned long *)(mtd_phys)) == CRAMFS_MAGIC) > > + mtd_size = PAGE_ALIGN(*((unsigned long *)(mtd_phys + 0x4))); > > +# endif > > + > > +# if defined(CONFIG_ROMFS_FS) > > + if (((unsigned long *)mtd_phys)[0] == ROMSB_WORD0 > > + && ((unsigned long *)mtd_phys)[1] == ROMSB_WORD1) > > + mtd_size = > > + PAGE_ALIGN(be32_to_cpu(((unsigned long *)mtd_phys)[2])); > > This detection seems to me like a strange thing to do in setup_arch(). > It should be possible to do this much later, at a point where the system > is much less fragile and e.g. printk works. It could even be moved into > some place in the mtd code itself, since other architectures might want > to do the same thing. After download the rootfs image from host to the target ram, we need to move the image to the right place, so we need to know the size of the image at this time. > > > +#if defined(CONFIG_BF561) > > +static struct cpu cpu[2]; > > +#else > > +static struct cpu cpu[1]; > > +#endif > > +static int __init topology_init(void) > > +{ > > +#if defined (CONFIG_BF561) > > + register_cpu(&cpu[0], 0); > > + register_cpu(&cpu[1], 1); > > + return 0; > > +#else > > + return register_cpu(cpu, 0); > > +#endif > > +} > > I think you should try to avoid the special-case stuff here. You can > have CONFIG_NR_CPUS in Kconfig set dependent on CONFIG_BF561 and change > the code here (and similarly in other places) to > > static struct cpu cpu[NR_CPUS]; > static int __init topology_init(void) > { > int i; > for (i=0; i< NR_CPUS; i++) { > register_cpu(&cpu[i], i); > return 0; > } > Good suggestions, thanks. > > + for (i = ZERO_P; i <= L2_MEM; i++) { > > + > > + if (cplb_data[i].valid) { > > + > > + as_1m = cplb_data[i].start % SIZE_1M; > > + > > + /* We need to make sure all sections are properly 1M aligned > > + * However between Kernel Memory and the Kernel mtd section, depending > on the > > + * rootfs size, there can be overlapping memory areas. > > + */ > > + > > + if (as_1m) { > > +#ifdef CONFIG_MTD_UCLINUX > > + if (i == SDRAM_RAM_MTD) { > > + if ((cplb_data[SDRAM_KERN].end + 1) > cplb_data[SDRAM_RAM_MTD].start) > > + cplb_data[SDRAM_RAM_MTD].start = (cplb_data[i].start & (-2*SIZE_1M)) > + SIZE_1M; > > I count 6 levels of indentation, which severely limits readability, > especially when you have terms this complex in the last level. > Please try to split up functions like this into smaller units. > > > +/* > > + * ++roman (07/09/96): implemented signal stacks (specially for tosemu on > > + * Atari :-) Current limitation: Only one sigstack can be active at one > time. > > + * If a second signal with SA_ONSTACK set arrives while working on a > sigstack, > > + * SA_ONSTACK is ignored. This behaviour avoids lots of trouble with nested > > + * signal handlers! > > + */ > > This comment is probably outdated. > > > +extern int setup_irq(unsigned int irq, struct irqaction *handler); > > this extern should be in a header. > > > +asmlinkage void trap_c(struct pt_regs *fp) > > +{ > > + int j, sig = 0; > > + siginfo_t info; > > + unsigned long trapnr = fp->seqstat & SEQSTAT_EXCAUSE; > > + > > +#ifdef CONFIG_KGDB > > +# define CHK_DEBUGGER_TRAP() do { CHK_DEBUGGER(trapnr, sig, info.si_code, > fp,); } while (0) > > +# define CHK_DEBUGGER_TRAP_MAYBE() do { if (kgdb_connected) > CHK_DEBUGGER_TRAP(); } while (0) > > +#else > > +# define CHK_DEBUGGER_TRAP() do { } while (0) > > +# define CHK_DEBUGGER_TRAP_MAYBE() do { } while (0) > > +#endif > > + > > + trace_buffer_save(j); > > + > > + /* trap_c() will be called for exceptions. During exceptions > > + * processing, the pc value should be set with retx value. > > + * With this change we can cleanup some code in signal.c- TODO > > + */ > > + fp->orig_pc = fp->retx; > > + /* printk("exception: 0x%x, ipend=%x, reti=%x, retx=%x\n", > > + trapnr, fp->ipend, fp->pc, fp->retx); */ > > + > > + /* send the appropriate signal to the user program */ > > + switch (trapnr) { > > + > > + /* This table works in conjuction with the one in ./mach-common/entry.S > > + * Some exceptions are handled there (in assembly, in exception space) > > + * Some are handled here, (in C, in interrupt space) > > + * Some, like CPLB, are handled in both, where the normal path is > > + * handled in assembly/exception space, and the error path is handled > > + * here > > + */ > > + > > + /* 0x00 - Linux Syscall, getting here is an error */ > > + /* 0x01 - userspace gdb breakpoint, handled here */ > > + case VEC_EXCPT01: > > + info.si_code = TRAP_ILLTRAP; > > + sig = SIGTRAP; > > + CHK_DEBUGGER_TRAP_MAYBE(); > > + /* Check if this is a breakpoint in kernel space */ > > + if (fp->ipend & 0xffc0) > > + return; > > + else > > + break; > > +#ifdef CONFIG_KGDB > > + case VEC_EXCPT02 : /* gdb connection */ > > + info.si_code = TRAP_ILLTRAP; > > + sig = SIGTRAP; > > + CHK_DEBUGGER_TRAP(); > > + return; > > +#else > > + /* 0x02 - User Defined, Caught by default */ > > +#endif > > + /* 0x03 - Atomic test and set */ > > + case VEC_EXCPT03: > > > The table here probably gets converted by gcc into a lookup table, but > I think it would be clearer to make that explicit and force the table > here with something like: > > static void trap_breakpoint(struct pt_regs *fp, siginfo_t *info, > unsigned int trapnr) > { > info->si_code = TRAP_ILLTRAP; > CHK_DEBUGGER_TRAP_MAYBE(); > if (fp->ipend & 0xffc0) > return 0; > return SIGTRAP; > } > > #ifdef CONFIG_KGDB > static void trap_kgdb(struct pt_regs *fp, siginfo_t *info, > unsigned int trapnr) > { > info->si_code = TRAP_ILLTRAP; > CHK_DEBUGGER_TRAP(); > return SIGTRAP; > } > #endif > > static void (traps[])(struct pt_regs *fp, siginfo_t *info, > unsigned int trapnr) = { > [VEC_EXCPT01] &trap_breakpoint, /* 0x01 - userspace gdb breakpoint */ > #ifdef CONFIG_KGDB > [VEC_EXCPT02] &trap_kgdb, /* gdb connection */ > #endif > ... > }; > > asmlinkage void trap_c(struct pt_regs *fp) > { > ... > if (trapnr > 0x3f) > goto error; > if (!traps[trapnr]) > goto out; > ret = traps[trapno](fp, &info, trapnr); > if (!ret) > goto out; > ... > > That should make sure you get the efficient trap handling almost and might > even > save some code space (you'd have to try that). > > > +asmlinkage int sys_bfin_spinlock(int *spinlock) > > +{ > > + int ret = 0; > > + int tmp = 0; > > + > > + local_irq_disable(); > > + ret = get_user(tmp, spinlock); > > + if (ret == 0) { > > + if (tmp) > > + ret = 1; > > + tmp = 1; > > + put_user(tmp, spinlock); > > + } > > + local_irq_enable(); > > + return ret; > > +} > > I'm curious: In your dual-core bf561, don't you actually need to implement > something that maintains atomicity across cores rather than just across > processes? Yes, bf561 is a dual-core processor, but we are using only one core of bf561 now. IMHO, BF561 architecture was not designed for SMP or NUMA. > > > +#include <net/checksum.h> > > +#include <asm/checksum.h> > > + > > +#ifdef CONFIG_IP_CHECKSUM_L1 > > +static unsigned short do_csum(const unsigned char *buff, int > len)__attribute__((l1_text)); > > +#endif > > + > > +static unsigned short do_csum(const unsigned char *buff, int len) > > +{ > > + register unsigned long sum = 0; > > + int swappem = 0; > > + > > + if (1 & (unsigned long)buff) { > > + sum = *buff << 8; > > + buff++; > > + len--; > > + ++swappem; > > + } > > For the non-L1 case, is there actually anything specific to blackfin in > this code? I wonder if we should instead have a totally generic version > of that in lib/csum.c for those architectures that don't want their > own optimizations for this. > > > +static struct platform_device *stamp_devices[] __initdata = { > > +#if defined(CONFIG_RTC_DRV_BFIN) || defined(CONFIG_RTC_DRV_BFIN_MODULE) > > + &rtc_device, > > +#endif > > + > > +#if defined(CONFIG_BFIN_CFPCMCIA) || defined(CONFIG_BFIN_CFPCMCIA_MODULE) > > + &bfin_pcmcia_cf_device, > > +#endif > > + > > +#if defined(CONFIG_USB_SL811_HCD) || defined(CONFIG_USB_SL811_HCD_MODULE) > > + &sl811_hcd_device, > > +#endif > > This array reminds me of the bad old days when all initialization functions > had to be called from a common function. Normally I don't like to recommend > using section magic for anything, but in this case I guess you'd be > better off if you just do your own thing here, like: > > #define BFIN_PLATFORM_DEVICE(name) struct platform_device name \ > __attribute((section("initdata.platformdev"))) > > then declare your platform devices in the respective driver, like > > static BFIN_PLATFORM_DEVICE(rtc_device) = { > ... > }; > > and loop through the new section like you do for the initcalls. > > > +void get_bf537_ether_addr(char *addr) > > +{ > > + /* currently the mac addr is saved in flash */ > > + int flash_mac = 0x203f0000; > > + *(u32 *)(&(addr[0])) = *(int *)flash_mac; > > + flash_mac += 4; > > + *(u16 *)(&(addr[4])) = (u16)*(int *)flash_mac; > > +} > > + > > +EXPORT_SYMBOL(get_bf537_ether_addr); > > This looks wrong in a number of aspects. Are there other things > stored in the flash so you can make this a more general flash > abstraction? Is there more than one driver using this at all? > If not, it should probably be part of the driver, with the > exact address being a compile-time constant for that driver. > > It should probably also use bfin_readXX instead direct dereferences. Good suggestions, thanks. > > > +/* The number of spurious interrupts */ > > +volatile unsigned int num_spurious; > > Why volatile? It probably does not what you intended, maybe it should > be atomic_t instead. > > > +/* BASE LEVEL interrupt handler routines */ > > +asmlinkage void evt_emulation(void); > > +asmlinkage void evt_exception(void); > > +asmlinkage void trap(void); > > +asmlinkage void evt_ivhw(void); > > +asmlinkage void evt_timer(void); > > +asmlinkage void evt_evt2(void); > > +asmlinkage void evt_evt7(void); > > +asmlinkage void evt_evt8(void); > > +asmlinkage void evt_evt9(void); > > +asmlinkage void evt_evt10(void); > > +asmlinkage void evt_evt11(void); > > +asmlinkage void evt_evt12(void); > > +asmlinkage void evt_evt13(void); > > +asmlinkage void evt_soft_int1(void); > > +asmlinkage void evt_system_call(void); > > +asmlinkage void init_exception_buff(void); > > move the extern definitions to a header file please > > > +/* BASE LEVEL interrupt handler routines */ > > +asmlinkage void evt_emulation(void); > > +asmlinkage void evt_exception(void); > > +asmlinkage void trap(void); > > +asmlinkage void evt_ivhw(void); > > +asmlinkage void evt_timer(void); > > +asmlinkage void evt_evt2(void); > > +asmlinkage void evt_evt7(void); > > +asmlinkage void evt_evt8(void); > > +asmlinkage void evt_evt9(void); > > +asmlinkage void evt_evt10(void); > > +asmlinkage void evt_evt11(void); > > +asmlinkage void evt_evt12(void); > > +asmlinkage void evt_evt13(void); > > +asmlinkage void evt_soft_int1(void); > > +asmlinkage void evt_system_call(void); > > Then you don't have to write them twice ;-) > > > +static __inline__ void atomic_add(int i, atomic_t * v) > > +{ > > + long flags; > > + > > + local_irq_save(flags); > > + v->counter += i; > > + local_irq_restore(flags); > > +} > > + > > +static __inline__ void atomic_sub(int i, atomic_t * v) > > +{ > > + long flags; > > + > > + local_irq_save(flags); > > + v->counter -= i; > > + local_irq_restore(flags); > > + > > +} > > The implementation of the atomic_t type looks very generic. Maybe it could > be moved to asm-generic/atomic-irqdisable.h or similar to be shared by > all architectures without atomic operations. > > > +struct spi_device_t { > > + char *dev_name; > > + > > + unsigned short flag; > > + unsigned short bdrate; > > + > > + unsigned short enable; > > + unsigned short master; > > + unsigned short out_opendrain; > > + unsigned short polar; > > + unsigned short phase; > > + unsigned short byteorder; /* 0: MSB first; 1: LSB first; */ > > + unsigned short size; /* 0: 8 bits; 1: 16 bits */ > > + unsigned short emiso; > > + unsigned short send_zero; > > + unsigned short more_data; > > + unsigned short slave_sel; > > + unsigned short ti_mod; > > + > > + unsigned short dma; /* use dma mode or not */ > > + unsigned short dma_config; /* only valid if dma enabled */ > > + > > + irqreturn_t(*irq_handler) (int irq, void *dev_id, > > + struct pt_regs *regs); > > + void *priv_data; > > +}; > > How does this fit in with the generic SPI code? Does it duplicate stuff > from there, or do you use it? We use our own. We have dma which can be used for SPI operations. > > > +static __inline__ void set_bit(int nr, volatile unsigned long *addr) > > +{ > > + int *a = (int *)addr; > > + int mask; > > + unsigned long flags; > > + > > + a += nr >> 5; > > + mask = 1 << (nr & 0x1f); > > + local_irq_save(flags); > > + *a |= mask; > > + local_irq_restore(flags); > > +} > > + > > +static __inline__ void __set_bit(int nr, volatile unsigned long *addr) > > +{ > > + int *a = (int *)addr; > > + int mask; > > + > > + a += nr >> 5; > > + mask = 1 << (nr & 0x1f); > > + *a |= mask; > > +} > > same as for the atomic header, this could become generic. > > > +#if defined(ANOMALY_05000312) && defined(ANOMALY_05000244) > > +#define SSYNC() do { int _tmp; \ > > + __asm__ __volatile__ ("cli %0;\n\t"\ > > + "nop;nop;\n\t"\ > > + "ssync;\n\t"\ > > + "sti %0;\n\t" \ > > + :"=d"(_tmp):);\ > > + } while (0) > > +#elif defined(ANOMALY_05000312) && !defined(ANOMALY_05000244) > > +#define SSYNC() do { int _tmp; \ > > + __asm__ __volatile__ ("cli %0;\n\t"\ > > + "ssync;\n\t"\ > > + "sti %0;\n\t" \ > > + :"=d"(_tmp):);\ > > + } while (0) > > +#elif !defined(ANOMALY_05000312) && defined(ANOMALY_05000244) > > +#define SSYNC() do {__builtin_bfin_ssync();} while (0) > > +#elif !defined(ANOMALY_05000312) && !defined(ANOMALY_05000244) > > +#define SSYNC() do {__asm__ __volatile__ ("ssync;\n\t") } while (0) > > +#endif > > + > > + > > +#if defined(ANOMALY_05000312) && defined(ANOMALY_05000244) > > +#define CSYNC() do { int _tmp; \ > > + __asm__ __volatile__ ("cli %0;\n\t"\ > > + "nop;nop;\n\t"\ > > + "csync;\n\t"\ > > + "sti %0;\n\t" \ > > + :"=d"(_tmp):);\ > > + } while (0) > > +#elif defined(ANOMALY_05000312) && !defined(ANOMALY_05000244) > > +#define CSYNC() do { int _tmp; \ > > + __asm__ __volatile__ ("cli %0;\n\t"\ > > + "csync;\n\t"\ > > + "sti %0;\n\t" \ > > + :"=d"(_tmp):);\ > > + } while (0) > > +#elif !defined(ANOMALY_05000312) && defined(ANOMALY_05000244) > > +#define CSYNC() do {__builtin_bfin_csync();} while (0) > > +#elif !defined(ANOMALY_05000312) && !defined(ANOMALY_05000244) > > +#define CSYNC() do {__asm__ __volatile__ ("csync;\n\t") } while (0) > > +#endif > > it would be nice if you could convert these to inline functions > instead of macros. > > > +static inline struct task_struct *get_current(void) __attribute__ > ((__const__)); > > +static inline struct task_struct *get_current(void) > > +{ > > + return (current_thread_info()->task); > > +} > > + > > +#define current (get_current()) > > You might want to maintain the current variable in the L1 memory, but > probably that difference it not measurable. > > > +#ifdef BFIN_DMA_NDEBUG > > +#define assert(expr) do {} while(0) > > +#else > > +#define assert(expr) \ > > + if (!(expr)) { \ > > + printk(KERN_INFO "Assertion failed! %s, %s, %s, line=%d \n", \ > > + #expr, __FILE__,__FUNCTION__,__LINE__); \ > > + } > > +#endif > > Please don't define your own assert() functions, in the kernel we have > BUG_ON and WARN_ON for this. > > > +#ifndef _BLACKFIN_DPMC_H_ > > +#define _BLACKFIN_DPMC_H_ > > + > > +#define SLEEP_MODE 1 > > +#define DEEP_SLEEP_MODE 2 > > +#define ACTIVE_PLL_DISABLED 3 > > +#define FULLON_MODE 4 > > +#define ACTIVE_PLL_ENABLED 5 > > +#define HIBERNATE_MODE 6 > > + > > +#define IOCTL_FULL_ON_MODE _IO('s', 0xA0) > > +#define IOCTL_ACTIVE_MODE _IO('s', 0xA1) > > +#define IOCTL_SLEEP_MODE _IO('s', 0xA2) > > +#define IOCTL_DEEP_SLEEP_MODE _IO('s', 0xA3) > > +#define IOCTL_HIBERNATE_MODE _IO('s', 0xA4) > > +#define IOCTL_CHANGE_FREQUENCY _IOW('s', 0xA5, unsigned long) > > +#define IOCTL_CHANGE_VOLTAGE _IOW('s', 0xA6, unsigned long) > > +#define IOCTL_SET_CCLK _IOW('s', 0xA7, unsigned long) > > +#define IOCTL_SET_SCLK _IOW('s', 0xA8, unsigned long) > > +#define IOCTL_GET_PLLSTATUS _IOW('s', 0xA9, unsigned long) > > +#define IOCTL_GET_CORECLOCK _IOW('s', 0xAA, unsigned long) > > +#define IOCTL_GET_SYSTEMCLOCK _IOW('s', 0xAB, unsigned long) > > +#define IOCTL_GET_VCO _IOW('s', 0xAC, unsigned long) > > +#define IOCTL_DISABLE_WDOG_TIMER _IO('s', 0xAD) > > +#define IOCTL_UNMASK_WDOG_WAKEUP_EVENT _IO('s',0xAE) > > +#define IOCTL_PROGRAM_WDOG_TIMER _IOW('s',0xAF,unsigned long) > > +#define IOCTL_CLEAR_WDOG_WAKEUP_EVENT _IO('s',0xB0) > > +#define IOCTL_SLEEP_DEEPER_MODE _IO('s',0xB1) > > + > > +#define DPMC_MINOR 254 > > + > > +#define ON 0 > > +#define OFF 1 > > + > > +unsigned long calc_volt(void); > > +int calc_vlev(int vlt); > > +unsigned long change_voltage(unsigned long volt); > > +int calc_msel(int vco_hz); > > +unsigned long change_frequency(unsigned long vco_mhz); > > +int set_pll_div(unsigned short sel, unsigned char flag); > > In general, you should be careful not to mix in-kernel declarations with > user-visible parts. In this case, the ioctl definitions need to be exported > via 'make headers_install', while the function declarations clearly are > not useful for user space and should therefore be inside of > '#ifdef __KERNEL__'. You should then list this file as 'unifdef-y' in the > Kbuild file. > > > + > > +#ifdef __KERNEL__ > > +#define SET_PERSONALITY(ex, ibcs2) > set_personality((ibcs2)?PER_SVR4:PER_LINUX) > > +#endif > > This looks bogus, I don't think you implement personalities. > > > +#define readb(addr) ({ unsigned __v; \ > > + int _tmp; \ > > + __asm__ __volatile__ ("cli %1;\n\t"\ > > + "NOP;NOP;SSYNC;\n\t"\ > > + "%0 = b [%2] (z);\n\t"\ > > + "sti %1;\n\t" \ > > + : "=d"(__v), "=d"(_tmp): "a"(addr)); (unsigned char)__v; }) > > + > > +#define readw(addr) ({ unsigned __v; \ > > + int _tmp; \ > > + __asm__ __volatile__ ("cli %1;\n\t"\ > > + "NOP;NOP;SSYNC;\n\t"\ > > + "%0 = w [%2] (z);\n\t"\ > > + "sti %1;\n\t" \ > > + : "=d"(__v), "=d"(_tmp): "a"(addr)); (unsigned short)__v; }) > > They should probably be inline functions. > > > +#define TCGETS 0x5401 > > +#define TCSETS 0x5402 > > +#define TCSETSW 0x5403 > > +#define TCSETSF 0x5404 > > +#define TCGETA 0x5405 > > +#define TCSETA 0x5406 > > +#define TCSETAW 0x5407 > > +#define TCSETAF 0x5408 > > +#define TCSBRK 0x5409 > > +#define TCXONC 0x540A > > +#define TCFLSH 0x540B > > +#define TIOCEXCL 0x540C > > +#define TIOCNXCL 0x540D > > +#define TIOCSCTTY 0x540E > > +#define TIOCGPGRP 0x540F > > +#define TIOCSPGRP 0x5410 > > +#define TIOCOUTQ 0x5411 > > +#define TIOCSTI 0x5412 > > +#define TIOCGWINSZ 0x5413 > > +#define TIOCSWINSZ 0x5414 > > +#define TIOCMGET 0x5415 > > +#define TIOCMBIS 0x5416 > > +#define TIOCMBIC 0x5417 > > +#define TIOCMSET 0x5418 > > +#define TIOCGSOFTCAR 0x5419 > > +#define TIOCSSOFTCAR 0x541A > > +#define FIONREAD 0x541B > > +#define TIOCINQ FIONREAD > > +#define TIOCLINUX 0x541C > > +#define TIOCCONS 0x541D > > +#define TIOCGSERIAL 0x541E > > +#define TIOCSSERIAL 0x541F > > +#define TIOCPKT 0x5420 > > +#define FIONBIO 0x5421 > > +#define TIOCNOTTY 0x5422 > > +#define TIOCSETD 0x5423 > > +#define TIOCGETD 0x5424 > > +#define TCSBRKP 0x5425 /* Needed for POSIX tcsendbreak() */ > > +#define TIOCTTYGSTRUCT 0x5426 /* For debugging only */ > > +#define TIOCSBRK 0x5427 /* BSD compatibility */ > > +#define TIOCCBRK 0x5428 /* BSD compatibility */ > > +#define TIOCGSID 0x5429 /* Return the session ID of FD */ > > +#define TIOCGPTN _IOR('T',0x30, unsigned int) /* Get Pty Number (of pty-mux > device) */ > > +#define TIOCSPTLCK _IOW('T',0x31, int) /* Lock/unlock Pty */ > > These look like another good candidate for an asm-generic version. > > > + > > +extern void sys_free_irq(unsigned int irq, void *dev_id); > > This functions should probably be renamed. Anything that starts with > 'sys_' is normally a system call. > > + > > +/* Clock and System Control (0xFFC0 0400-0xFFC0 07FF) */ > > +#define bfin_read_PLL_CTL() bfin_read16(PLL_CTL) > > +#define bfin_write_PLL_CTL(val) bfin_write16(PLL_CTL,val) > > +#define bfin_read_PLL_STAT() bfin_read16(PLL_STAT) > > +#define bfin_write_PLL_STAT(val) bfin_write16(PLL_STAT,val) > > +#define bfin_read_PLL_LOCKCNT() bfin_read16(PLL_LOCKCNT) > > +#define bfin_write_PLL_LOCKCNT(val) bfin_write16(PLL_LOCKCNT,val) > > +#define bfin_read_CHIPID() bfin_read32(CHIPID) > > +#define bfin_read_SWRST() bfin_read16(SWRST) > > +#define bfin_write_SWRST(val) bfin_write16(SWRST,val) > > I remember that we have discussed these macro abstractions before, but don't > recall the result of the discussion. You have around 5000 such macros, > and I still think it's not a helpful abstraction. It is much more common > for device drivers to just use the read/write accessors directly. IIRC > the objections that were raised (and my replies to them) were roughly: > > > the driver writer doesn't want to know the size of the variable, and > > if it changes, they need to change every instance in the code, not > > just the macro. > > A driver writer should better know the type of variable he is changing, > e.g. because of the type he passes in and out. Also, hardware registers > don't suddenly change in size. > > > These macros allow us to work around hardware bugs when accessing the > > registers by simply redefining the accessor to do something more > > complex. > > If there is a bug in the hardware, the workaround belongs into the > driver. You can then still define a special inline function to > access that particular register. -Aubrey ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 6:54 ` Aubrey Li @ 2007-03-05 8:47 ` Arnd Bergmann 2007-03-05 9:19 ` Wu, Bryan 0 siblings, 1 reply; 24+ messages in thread From: Arnd Bergmann @ 2007-03-05 8:47 UTC (permalink / raw) To: Aubrey Li; +Cc: bryan.wu, Andrew Morton, linux-kernel On Monday 05 March 2007, Aubrey Li wrote: > On 3/4/07, Arnd Bergmann <arnd@arndb.de> wrote: > > In general, please put EXPORT_SYMBOL lines below the definition > > of the symbol itself. This list of exports should only be used > > for symbols that come from assembly files. > > What is the right way to export symbol coming from c files? As I said, below the symbol definition, like int global_var; EXPORT_SYMBOL(global_var); int global_function(void) { return 3; } EXPORT_SYMBOL(global_function); > > This detection seems to me like a strange thing to do in setup_arch(). > > It should be possible to do this much later, at a point where the system > > is much less fragile and e.g. printk works. It could even be moved into > > some place in the mtd code itself, since other architectures might want > > to do the same thing. > > After download the rootfs image from host to the target ram, we need > to move the image to the right place, so we need to know the size of > the image at this time. Well, it doesn't have to be in the modular part of the kernel, but some place later than setup_arch() would be a step in the right direction. If you need it before the file systems, an arch_initcall() might be the right place. > > I'm curious: In your dual-core bf561, don't you actually need to implement > > something that maintains atomicity across cores rather than just across > > processes? > > Yes, bf561 is a dual-core processor, but we are using only one core of > bf561 now. > IMHO, BF561 architecture was not designed for SMP or NUMA. Interesting, so what is the intended use of the other core? Does the hardware have any way of supporting concurrency between the cores, other than sending interrupts between them? > > How does this fit in with the generic SPI code? Does it duplicate stuff > > from there, or do you use it? > > We use our own. We have dma which can be used for SPI operations. I just looked again at your code. My question was more directed at whether you use your own SPI abstraction layer instead of drivers/spi, which you fortunately don't. The piece I was missing however is the spi_bfin5xx.c driver, which was not part of this patch, though you seem to rely on it. Is that already part of the -mm kernel? Arnd <>< ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 8:47 ` Arnd Bergmann @ 2007-03-05 9:19 ` Wu, Bryan 2007-03-05 16:43 ` Arnd Bergmann 0 siblings, 1 reply; 24+ messages in thread From: Wu, Bryan @ 2007-03-05 9:19 UTC (permalink / raw) To: Arnd Bergmann; +Cc: Aubrey Li, bryan.wu, Andrew Morton, linux-kernel On Mon, 2007-03-05 at 09:47 +0100, Arnd Bergmann wrote: > On Monday 05 March 2007, Aubrey Li wrote: > > On 3/4/07, Arnd Bergmann <arnd@arndb.de> wrote: > > > > In general, please put EXPORT_SYMBOL lines below the definition > > > of the symbol itself. This list of exports should only be used > > > for symbols that come from assembly files. > > > > What is the right way to export symbol coming from c files? > > As I said, below the symbol definition, like > > int global_var; > EXPORT_SYMBOL(global_var); > > int global_function(void) > { > return 3; > } > EXPORT_SYMBOL(global_function); > Got it, the rule will be followed. > > > I'm curious: In your dual-core bf561, don't you actually need to implement > > > something that maintains atomicity across cores rather than just across > > > processes? > > > > Yes, bf561 is a dual-core processor, but we are using only one core of > > bf561 now. > > IMHO, BF561 architecture was not designed for SMP or NUMA. > > Interesting, so what is the intended use of the other core? Does the > hardware have any way of supporting concurrency between the cores, > other than sending interrupts between them? > AFAIK, the dual-core in BF561 have their own L1 memory which can be configured as cache respectively. So there is no hardware to help cache coherence management, it is very hard to implement SMP, right? Maybe NUMA is a solution, but it is not a wonderful solution. In some application product, BF561 core A is running Linux kernel +Applications while BF561 core B is just for some complicated video/audio codec algorithm. Any Linux multicore solution in BF561 situation is highly welcome. > > > How does this fit in with the generic SPI code? Does it duplicate stuff > > > from there, or do you use it? > > > > We use our own. We have dma which can be used for SPI operations. > > I just looked again at your code. My question was more directed at whether > you use your own SPI abstraction layer instead of drivers/spi, which > you fortunately don't. The piece I was missing however is the spi_bfin5xx.c > driver, which was not part of this patch, though you seem to rely on it. > Is that already part of the -mm kernel? > > Arnd <>< I will send out the SPI patch ASAP. and your signature is very interesting, it looks like a fish -:) Another question: when is the merge point from -mm to linus mainline, is it the same as the merge window after 2.6.21 released? Thanks again -Bryan Wu ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 9:19 ` Wu, Bryan @ 2007-03-05 16:43 ` Arnd Bergmann 0 siblings, 0 replies; 24+ messages in thread From: Arnd Bergmann @ 2007-03-05 16:43 UTC (permalink / raw) To: bryan.wu; +Cc: Aubrey Li, Andrew Morton, linux-kernel On Monday 05 March 2007, Wu, Bryan wrote: > Maybe NUMA is a solution, but it is not a wonderful solution. NUMA doesn't help you. Linux only runs on cache-coherent NUMA, which this isn't. > In some application product, BF561 core A is running Linux kernel > +Applications while BF561 core B is just for some complicated > video/audio codec algorithm. > > Any Linux multicore solution in BF561 situation is highly welcome. You definitely can't use the cache mode in this case, but one idea that should make atomic instructions work is to always do these on one of the two cores, and use cross-core interrupts to trigger an update. It's probably pretty inefficient and you also need to do something about atomic updates (spinlock_t and atomic_t) when interrupts are disabled. > Another question: when is the merge point from -mm to linus mainline, is > it the same as the merge window after 2.6.21 released? It's the same. Arnd <>< ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-03 22:30 ` Arnd Bergmann 2007-03-03 22:50 ` bert hubert 2007-03-05 6:54 ` Aubrey Li @ 2007-03-05 7:34 ` Wu, Bryan 2007-03-05 8:10 ` Arnd Bergmann 2007-03-06 2:09 ` Mike Frysinger 3 siblings, 1 reply; 24+ messages in thread From: Wu, Bryan @ 2007-03-05 7:34 UTC (permalink / raw) To: Arnd Bergmann; +Cc: bryan.wu, Andrew Morton, linux-kernel On Sat, 2007-03-03 at 17:30 -0500, Arnd Bergmann wrote: > On Thursday 01 March 2007 05:14:40 Wu, Bryan wrote: > > Here is the update version of blackfin-arch.patch in -mm tree. > > simply add support to utrace and it was tested on blackfin STAMP > board > > as well as other following patches. > > Wow, this has come a long way since I looked at the patches last > year, good work! Yeah, old friend comes back. Thanks a lot for your review. We fixed lots of things according to your review since last year. > > I've gone through the complete patch again now, and these are the > issues I've found in it. None of these are show-stoppers and I'd > like to see it all go in during the next merge window. There should > be enough time until then to address these points: Lots of valuable comments got from you, we will get things done soon. So could please give us some information about the merge window schedule, we may try to catch this. > + > > +/* Clock and System Control (0xFFC0 0400-0xFFC0 07FF) */ > > +#define bfin_read_PLL_CTL() bfin_read16(PLL_CTL) > > +#define bfin_write_PLL_CTL(val) > bfin_write16(PLL_CTL,val) > > +#define bfin_read_PLL_STAT() bfin_read16(PLL_STAT) > > +#define bfin_write_PLL_STAT(val) > bfin_write16(PLL_STAT,val) > > +#define bfin_read_PLL_LOCKCNT() > bfin_read16(PLL_LOCKCNT) > > +#define bfin_write_PLL_LOCKCNT(val) > bfin_write16(PLL_LOCKCNT,val) > > +#define bfin_read_CHIPID() bfin_read32(CHIPID) > > +#define bfin_read_SWRST() bfin_read16(SWRST) > > +#define bfin_write_SWRST(val) > bfin_write16(SWRST,val) > > I remember that we have discussed these macro abstractions before, but > don't > recall the result of the discussion. You have around 5000 such macros, > and I still think it's not a helpful abstraction. It is much more > common > for device drivers to just use the read/write accessors directly. IIRC > the objections that were raised (and my replies to them) were roughly: > > > the driver writer doesn't want to know the size of the variable, and > > if it changes, they need to change every instance in the code, not > > just the macro. > > A driver writer should better know the type of variable he is > changing, > e.g. because of the type he passes in and out. Also, hardware > registers > don't suddenly change in size. > > > These macros allow us to work around hardware bugs when accessing > the > > registers by simply redefining the accessor to do something more > > complex. > > If there is a bug in the hardware, the workaround belongs into the > driver. You can then still define a special inline function to > access that particular register. > Oh, if we should fix this issue, there are lots of work to do because tons of drivers rely on this. Maybe after some team internal discussion, we will give a solution to this. Thanks again Best Regards, -Bryan Wu ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 7:34 ` Wu, Bryan @ 2007-03-05 8:10 ` Arnd Bergmann 0 siblings, 0 replies; 24+ messages in thread From: Arnd Bergmann @ 2007-03-05 8:10 UTC (permalink / raw) To: bryan.wu; +Cc: Andrew Morton, linux-kernel On Monday 05 March 2007, Wu, Bryan wrote: > So could please give us some information about the merge window > schedule, we may try to catch this. The merge window opens after 2.6.21 gets released and is open for two weeks aftre that. The idea is however that you have everything ready at the start of the merge window. > Oh, if we should fix this issue, there are lots of work to do because > tons of drivers rely on this. Maybe after some team internal discussion, > we will give a solution to this. You can probably use a short perl script (or similar) to automate the conversion. Arnd <>< ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-03 22:30 ` Arnd Bergmann ` (2 preceding siblings ...) 2007-03-05 7:34 ` Wu, Bryan @ 2007-03-06 2:09 ` Mike Frysinger 3 siblings, 0 replies; 24+ messages in thread From: Mike Frysinger @ 2007-03-06 2:09 UTC (permalink / raw) To: Arnd Bergmann; +Cc: bryan.wu, Andrew Morton, linux-kernel On 3/3/07, Arnd Bergmann <arnd@arndb.de> wrote: > > +#if defined(CONFIG_MTD_UCLINUX) > > + /* generic memory mapped MTD driver */ > > + memory_mtd_end = memory_end; > > + > > + mtd_phys = _ramstart; > > + mtd_size = PAGE_ALIGN(*((unsigned long *)(mtd_phys + 8))); > > + > > +# if defined(CONFIG_EXT2_FS) || defined(CONFIG_EXT3_FS) > > + if (*((unsigned short *)(mtd_phys + 0x438)) == EXT2_SUPER_MAGIC) > > + mtd_size = > > + PAGE_ALIGN(*((unsigned long *)(mtd_phys + 0x404)) << 10); > > +# endif > > + > > +# if defined(CONFIG_CRAMFS) > > + if (*((unsigned long *)(mtd_phys)) == CRAMFS_MAGIC) > > + mtd_size = PAGE_ALIGN(*((unsigned long *)(mtd_phys + 0x4))); > > +# endif > > + > > +# if defined(CONFIG_ROMFS_FS) > > + if (((unsigned long *)mtd_phys)[0] == ROMSB_WORD0 > > + && ((unsigned long *)mtd_phys)[1] == ROMSB_WORD1) > > + mtd_size = > > + PAGE_ALIGN(be32_to_cpu(((unsigned long *)mtd_phys)[2])); > > This detection seems to me like a strange thing to do in setup_arch(). > It should be possible to do this much later, at a point where the system > is much less fragile and e.g. printk works. It could even be moved into > some place in the mtd code itself, since other architectures might want > to do the same thing. the MTD code already does this ... the difference is that the blackfin port changes the memory layout dynamically by relocating the rootfs from the end of its data section to the end of memory ... that's why it's done so early in the boot process, it shrinks the memory pool available to the kernel allocator > > +#define TCGETS 0x5401 > > +#define TCSETS 0x5402 > > +#define TCSETSW 0x5403 > > +#define TCSETSF 0x5404 > > +#define TCGETA 0x5405 > > +#define TCSETA 0x5406 > > +#define TCSETAW 0x5407 > > +#define TCSETAF 0x5408 > > +#define TCSBRK 0x5409 > > +#define TCXONC 0x540A > > +#define TCFLSH 0x540B > > +#define TIOCEXCL 0x540C > > +#define TIOCNXCL 0x540D > > +#define TIOCSCTTY 0x540E > > +#define TIOCGPGRP 0x540F > > +#define TIOCSPGRP 0x5410 > > +#define TIOCOUTQ 0x5411 > > +#define TIOCSTI 0x5412 > > +#define TIOCGWINSZ 0x5413 > > +#define TIOCSWINSZ 0x5414 > > +#define TIOCMGET 0x5415 > > +#define TIOCMBIS 0x5416 > > +#define TIOCMBIC 0x5417 > > +#define TIOCMSET 0x5418 > > +#define TIOCGSOFTCAR 0x5419 > > +#define TIOCSSOFTCAR 0x541A > > +#define FIONREAD 0x541B > > +#define TIOCINQ FIONREAD > > +#define TIOCLINUX 0x541C > > +#define TIOCCONS 0x541D > > +#define TIOCGSERIAL 0x541E > > +#define TIOCSSERIAL 0x541F > > +#define TIOCPKT 0x5420 > > +#define FIONBIO 0x5421 > > +#define TIOCNOTTY 0x5422 > > +#define TIOCSETD 0x5423 > > +#define TIOCGETD 0x5424 > > +#define TCSBRKP 0x5425 /* Needed for POSIX tcsendbreak() */ > > +#define TIOCTTYGSTRUCT 0x5426 /* For debugging only */ > > +#define TIOCSBRK 0x5427 /* BSD compatibility */ > > +#define TIOCCBRK 0x5428 /* BSD compatibility */ > > +#define TIOCGSID 0x5429 /* Return the session ID of FD */ > > +#define TIOCGPTN _IOR('T',0x30, unsigned int) /* Get Pty Number (of pty-mux > device) */ > > +#define TIOCSPTLCK _IOW('T',0x31, int) /* Lock/unlock Pty */ > > These look like another good candidate for an asm-generic version. yeah, most everyone has the same ioctl values for terminal cruft > > These macros allow us to work around hardware bugs when accessing the > > registers by simply redefining the accessor to do something more > > complex. > > If there is a bug in the hardware, the workaround belongs into the > driver. You can then still define a special inline function to > access that particular register. not when it's an anomaly in a register that is not driver specific (like the VR_CTL register), or it's a peripheral that has no generic driver [yet] (like the SPORT or PPI) -mike ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-01 4:14 [PATCH -mm 1/5] Blackfin: blackfin architecture patch update Wu, Bryan 2007-03-03 20:38 ` Arnd Bergmann 2007-03-03 22:30 ` Arnd Bergmann @ 2007-03-05 9:23 ` Paul Mundt 2007-03-05 12:32 ` Bernd Schmidt ` (2 more replies) 2 siblings, 3 replies; 24+ messages in thread From: Paul Mundt @ 2007-03-05 9:23 UTC (permalink / raw) To: Wu, Bryan; +Cc: Andrew Morton, linux-kernel On Thu, Mar 01, 2007 at 12:14:40PM +0800, Wu, Bryan wrote: > Here is the update version of blackfin-arch.patch in -mm tree. > simply add support to utrace and it was tested on blackfin STAMP board > as well as other following patches. > > The whole patch is located at URL: > https://blackfin.uclinux.org/gf/download/frsrelease/39/2583/blackfin-arch.patch It would be nice if this could be split and posted incrementally for review. It's a bit of a pain reading through the entire thing in one shot. Anyways.. > Index: linux-2.6/arch/blackfin/Kconfig > =================================================================== > --- /dev/null 1970-01-01 00:00:00.000000000 +0000 > +++ linux-2.6/arch/blackfin/Kconfig 2007-03-01 10:30:27.000000000 +0800 [snip] > +config ZONE_DMA > + bool > + default y > + You don't need this, just get rid of it and shove everything in ZONE_NORMAL. > +config BLACKFIN > + bool > + default y > + > +config BFIN > + bool > + default y > + Why are there two of these? > +config GENERIC_IRQ_PROBE > + bool > + default y > + Whitespace damage. > +config UCLINUX > + bool > + default y > + Dead symbol. > +comment "Memory Optimizations" > + > +config I_ENTRY_L1 > + bool "Locate interrupt entry code in L1 Memory" > + default y > + help > + If enabled interrupt entry code (STORE/RESTORE CONTEXT) is linked > + into L1 instruction memory.(less latency) > + Wow, this is really crying out for a special linker section with slightly more intelligent relocation logic. You should flag the performance critical parts to be located in L1 memory directly with a section attribute, rather than making everything selectable. If you overflow you can simply spill in to main memory. > +choice > + prompt "Uncached SDRAM region" > + default DMA_UNCACHED_1M > + depends BFIN_DMA_5XX > +config DMA_UNCACHED_2M > + bool "Enable 2M DMA Zone" > +config DMA_UNCACHED_1M > + bool "Enable 1M DMA Zone" > +config DMA_UNCACHED_NONE > + bool "Disable DMA Zone" > +endchoice > + Contrary to the comment, this is not a zone. > +config DEBUG_HUNT_FOR_ZERO > + bool "Catch NULL pointer reads/writes" > + default y > + help > + Say Y here to catch reads/writes to anywhere in the memory range > + from 0x0000 - 0x0FFF (the first 4k) of memory. This is useful in > + catching common programming errors such as NULL pointer dereferences. > + > + Misbehaving applications will be killed (generate a SEGV) while the > + kernel will trigger a panic. > + > + Enabling this option will take up an extra entry in CPLB table. > + Otherwise, there is no extra overhead. > + Is this sane to have conditional? > +config BOOTPARAM > + bool "Compiled-in Kernel Boot Parameter" > + > +config BOOTPARAM_STRING > + string "Kernel Boot Parameter" > + default "console=ttyS0,57600" > + depends on BOOTPARAM > + Any reason not to use CMDLINE_BOOL/CMDLINE like every other platform? > +config NO_KERNEL_MSG > + bool "Suppress Kernel BUG Messages" > + help > + Do not output any debug BUG messages within the kernel. > + This is useless. For starters, CONFIG_BUG already does this. Secondly, this isn't used anywhere. > +int __init blackfin_dma_init(void) > +{ > + int i; > + > + printk(KERN_INFO "Blackfin DMA Controller\n"); > + > + for (i = 0; i < MAX_BLACKFIN_DMA_CHANNEL; i++) { > + dma_ch[i].chan_status = DMA_CHANNEL_FREE; > + dma_ch[i].regs = base_addr[i]; > + init_MUTEX(&(dma_ch[i].dmalock)); The dmalock is only ever used as a mutex, use that instead. > +void dma_alloc_init(unsigned long start, unsigned long end) > +{ > + spin_lock_init(&dma_page_lock); > + dma_initialized = 0; > + > + dma_page = (unsigned int *)__get_free_page(GFP_KERNEL); > + memset(dma_page, 0, PAGE_SIZE); > + dma_base = PAGE_ALIGN(start); > + dma_size = PAGE_ALIGN(end) - PAGE_ALIGN(start); > + dma_pages = dma_size >> PAGE_SHIFT; > + memset((void *)dma_base, 0, DMA_UNCACHED_REGION); > + dma_initialized = 1; > + > + printk(KERN_INFO "%s: dma_page @ 0x%p - %d pages at 0x%08lx\n", __FUNCTION__, > + dma_page, dma_pages, dma_base); > +} That's an "interesting" way of doing a bitmap. Please use a proper bitmap for dma_page, there's infrastructure for all of this already without having to come up with new schemes. dma_declare_coherent() is a good example. > +void *dma_alloc_coherent(struct device *dev, size_t size, > + dma_addr_t * dma_handle, gfp_t gfp) > +{ > + void *ret; > + > + ret = (void *)__alloc_dma_pages(get_pages(size)); > + > + if (ret) { > + memset(ret, 0, size); > + *dma_handle = virt_to_phys(ret); > + } > + No dcache write-back? > +dma_addr_t > +dma_map_single(struct device *dev, void *ptr, size_t size, > + enum dma_data_direction direction) > +{ > + BUG_ON(direction == DMA_NONE); > + > + blackfin_dcache_invalidate_range((unsigned long)ptr, > + (unsigned long)ptr + size); > + > + return (dma_addr_t) ptr; > +} > + > +int > +dma_map_sg(struct device *dev, struct scatterlist *sg, int nents, > + enum dma_data_direction direction) > +{ > + int i; > + > + BUG_ON(direction == DMA_NONE); > + > + for (i = 0; i < nents; i++) > + invalidate_dcache_range(sg_dma_address(&sg[i]), > + sg_dma_address(&sg[i]) + > + sg_dma_len(&sg[i])); > + Why are you using the different flushing routines here? > +#ifdef DEBUG_SERIAL_EARLY_INIT > + bfin_console_init(); /* early console registration */ > + /* this give a chance to get printk() working before crash. */ > +#endif > + Why not expose this as a sensible early_printk implementation? > +#ifdef CONFIG_CONSOLE > +#ifdef CONFIG_FRAMEBUFFER > + conswitchp = &fb_con; > +#else > + conswitchp = 0; > +#endif > +#endif > + Can't you do this in the defconfig? > + /* check the size of the l1 area */ > + l1_length = _etext_l1 - _stext_l1; > + if (l1_length > L1_CODE_LENGTH) > + panic("L1 memory overflow\n"); > + > + l1_length = _ebss_l1 - _sdata_l1; > + if (l1_length > L1_DATA_A_LENGTH) > + panic("L1 memory overflow\n"); > + That's not very nice. You can figure this out already at link time. > +void do_gettimeofday(struct timeval *tv) > +int do_settimeofday(struct timespec *tv) These can use CONFIG_GENERIC_TIME. > +static struct platform_device *cm_bf533_devices[] __initdata = { > +#if defined(CONFIG_SERIAL_BFIN) || defined(CONFIG_SERIAL_BFIN_MODULE) > + &bfin_uart_device, > +#endif > + Why? You'll already free this up if nothing claims it. One of the benefits of having the driver model is that we're able to get rid of this sort of ifdef abortion. > + /* > + * initialize the bad page table and bad page to point > + * to a couple of allocated pages > + */ > + empty_bad_page_table = (unsigned long)alloc_bootmem_pages(PAGE_SIZE); > + empty_bad_page = (unsigned long)alloc_bootmem_pages(PAGE_SIZE); > + empty_zero_page = (unsigned long)alloc_bootmem_pages(PAGE_SIZE); > + memset((void *)empty_zero_page, 0, PAGE_SIZE); > + dcache handling? > + tmp = (unsigned long)l1sram_alloc(sizeof(struct l1_scratch_task_info)); > + if (tmp != (unsigned long)L1_SCRATCH_TASK_INFO) { > + printk(KERN_EMERG "mem_init(): Did not get the right address from l1sram_alloc: %08lx != %08lx\n", > + tmp, (unsigned long)L1_SCRATCH_TASK_INFO); > + panic("No L1, time to give up\n"); > + } This platform seems to really want to panic() at the first sign of trouble. This is not a good policy, especially for something that's only a micro-optimization. > Index: linux-2.6/arch/blackfin/mm/kmap.c All of this can be inlined in io.h, there's nothing noteworthy here. > +static struct semaphore pfmon_sem; > + Use a mutex. > +int __init oprofile_arch_init(struct oprofile_operations *ops) > +{ > +#ifdef CONFIG_HARDWARE_PM [snip] > +#else > + return -1; > +#endif > +} > + Uh.. fix your dependencies. > +unsigned curr_pfctl, curr_count[2]; > + Globals? > Index: linux-2.6/include/asm-blackfin/bug.h > =================================================================== > --- /dev/null 1970-01-01 00:00:00.000000000 +0000 > +++ linux-2.6/include/asm-blackfin/bug.h 2007-03-01 10:30:27.000000000 +0800 > @@ -0,0 +1,15 @@ > +#ifndef _BLACKFIN_BUG_H > +#define _BLACKFIN_BUG_H > + > +#ifdef CONFIG_BUG > +#define HAVE_ARCH_BUG > +#define BUG() do { \ > + dump_stack(); \ > + printk(KERN_WARNING "\nkernel BUG at %s:%d!\n",\ > + __FILE__, __LINE__); \ > + panic("BUG!"); \ > +} while (0) > +#endif > + > +#include <asm-generic/bug.h> > +#endif What do you need HAVE_ARCH_BUG for? You're not doing anything with it.. > +#ifndef __ARCH_BLACKFIN_CACHE_H > +#define __ARCH_BLACKFIN_CACHE_H > + > +/* bytes per L1 cache line */ > +#define L1_CACHE_SHIFT 5 /* BlackFin loads 32 bytes for cache */ > +#define L1_CACHE_BYTES (1 << L1_CACHE_SHIFT) > + > +/* For speed we do need to align these ...MaTed---*/ > +/* But include/linux/cache.h does this for us if we DO not define ...MaTed---*/ > +#define __cacheline_aligned /***** maybe no need this Tony *****/ > +#define ____cacheline_aligned > + What the hell? > +static inline void flush_icache_range(unsigned start, unsigned end) > +{ > +#if defined(CONFIG_BLKFIN_DCACHE) && defined(CONFIG_BLKFIN_CACHE) > + > +# if defined(CONFIG_BLKFIN_WT) > + blackfin_icache_flush_range((start), (end)); > +# else > + blackfin_icache_dcache_flush_range((start), (end)); > +# endif > + > +#else > + > +# if defined(CONFIG_BLKFIN_CACHE) > + blackfin_icache_flush_range((start), (end)); > +# endif > +# if defined(CONFIG_BLKFIN_DCACHE) > + blackfin_dcache_flush_range((start), (end)); > +# endif > + > +#endif > +} > + This would probably be cleaner out-of-line, you can hide most of this ugliness in your Makefile instead. > +#pragma pack(2) > +struct dmasg_t { > + unsigned long next_desc_addr; > + unsigned long start_addr; > + unsigned short cfg; > + unsigned short x_count; > + short x_modify; > + unsigned short y_count; > + short y_modify; > +}; > +#pragma pack() > + Do you really need to use pragma? > +struct dma_register_t { Why is there a _t here? > +#define STR(X) STR1(X) > +#define STR1(X) #X You can use __stringify() for this if you feel you must. > +static inline unsigned long bfin_get_addr_from_rp(unsigned long *ptr, > + unsigned long relval, > + unsigned long flags, > + unsigned long *persistent) > +{ Some of these look frighteningly large to be inlined.. > +#define dma_cache_inv(_start,_size) do { blkfin_inv_cache_all();} while (0) > +#define dma_cache_wback(_start,_size) do { } while (0) > +#define dma_cache_wback_inv(_start,_size) do { blkfin_inv_cache_all();} while (0) What's the point of having selective flushing if you're 1) going to blow it all away, and 2) not use these in the dma-mapping cases? > +/* Clock and System Control (0xFFC0 0400-0xFFC0 07FF) */ > +#define bfin_read_PLL_CTL() bfin_read16(PLL_CTL) > +#define bfin_write_PLL_CTL(val) bfin_write16(PLL_CTL,val) > +#define bfin_read_PLL_STAT() bfin_read16(PLL_STAT) > +#define bfin_write_PLL_STAT(val) bfin_write16(PLL_STAT,val) > +#define bfin_read_PLL_LOCKCNT() bfin_read16(PLL_LOCKCNT) > +#define bfin_write_PLL_LOCKCNT(val) bfin_write16(PLL_LOCKCNT,val) What sort of magical abstraction is this? Is there some reason you can't just read and write the registers directly rather than having a wrapper for _every_ possible register? There are literally _thousands_ of lines of this stuff, and I imagine that careful auditing would reveal that not even 5% of them are actually used by the port. Presumably you have a processor manual, use that when you need it, rather than inlining this crap in the kernel. ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 9:23 ` Paul Mundt @ 2007-03-05 12:32 ` Bernd Schmidt 2007-03-05 12:39 ` Paul Mundt 2007-03-06 2:04 ` Mike Frysinger 2007-03-21 15:44 ` Mike Frysinger 2 siblings, 1 reply; 24+ messages in thread From: Bernd Schmidt @ 2007-03-05 12:32 UTC (permalink / raw) To: Paul Mundt, Wu, Bryan, Andrew Morton, linux-kernel Paul Mundt wrote: >> +comment "Memory Optimizations" >> + >> +config I_ENTRY_L1 >> + bool "Locate interrupt entry code in L1 Memory" >> + default y >> + help >> + If enabled interrupt entry code (STORE/RESTORE CONTEXT) is linked >> + into L1 instruction memory.(less latency) >> + > Wow, this is really crying out for a special linker section with slightly > more intelligent relocation logic. You should flag the performance > critical parts to be located in L1 memory directly with a section > attribute, rather than making everything selectable. If you overflow you > can simply spill in to main memory. This is done intentionally, because it's also possible for user code to be loaded into L1 memory. We want to give users the option to avoid filling it all up with kernel code. Bernd -- This footer brought to you by insane German lawmakers. Analog Devices GmbH Wilhelm-Wagenfeld-Str. 6 80807 Muenchen Registergericht Muenchen HRB 40368 Geschaeftsfuehrer Thomas Wessel, Vincent Roche, Joseph E. McDonough ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 12:32 ` Bernd Schmidt @ 2007-03-05 12:39 ` Paul Mundt 2007-03-05 13:26 ` Robin Getz 0 siblings, 1 reply; 24+ messages in thread From: Paul Mundt @ 2007-03-05 12:39 UTC (permalink / raw) To: Bernd Schmidt; +Cc: Wu, Bryan, Andrew Morton, linux-kernel On Mon, Mar 05, 2007 at 01:32:07PM +0100, Bernd Schmidt wrote: > Paul Mundt wrote: > >>+comment "Memory Optimizations" > >>+ > >>+config I_ENTRY_L1 > >>+ bool "Locate interrupt entry code in L1 Memory" > >>+ default y > >>+ help > >>+ If enabled interrupt entry code (STORE/RESTORE CONTEXT) is linked > >>+ into L1 instruction memory.(less latency) > >>+ > >Wow, this is really crying out for a special linker section with slightly > >more intelligent relocation logic. You should flag the performance > >critical parts to be located in L1 memory directly with a section > >attribute, rather than making everything selectable. If you overflow you > >can simply spill in to main memory. > > This is done intentionally, because it's also possible for user code to > be loaded into L1 memory. We want to give users the option to avoid > filling it all up with kernel code. > So then why not make the userspace component of it optional and allow a size cap for kernel usage that's configurable if it's enabled? This degree of abstraction is almost worse than no abstraction. ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 12:39 ` Paul Mundt @ 2007-03-05 13:26 ` Robin Getz 2007-03-05 14:00 ` Paul Mundt 0 siblings, 1 reply; 24+ messages in thread From: Robin Getz @ 2007-03-05 13:26 UTC (permalink / raw) To: Paul Mundt; +Cc: Bernd Schmidt, Wu, Bryan, Andrew Morton, linux-kernel On Mon 5 Mar 2007 07:39, Paul Mundt pondered: > On Mon, Mar 05, 2007 at 01:32:07PM +0100, Bernd Schmidt wrote: > > Paul Mundt wrote: > > >>+comment "Memory Optimizations" > > >>+ > > >>+config I_ENTRY_L1 > > >>+ bool "Locate interrupt entry code in L1 Memory" > > >>+ default y > > >>+ help > > >>+ If enabled interrupt entry code (STORE/RESTORE CONTEXT) is > > >> linked + into L1 instruction memory.(less latency) > > >>+ > > > > > >Wow, this is really crying out for a special linker section with > > > slightly more intelligent relocation logic. You should flag the > > > performance critical parts to be located in L1 memory directly with a > > > section attribute, rather than making everything selectable. If you > > > overflow you can simply spill in to main memory. > > > > This is done intentionally, because it's also possible for user code to > > be loaded into L1 memory. We want to give users the option to avoid > > filling it all up with kernel code. > > So then why not make the userspace component of it optional and allow a > size cap for kernel usage that's configurable if it's enabled? This degree > of abstraction is almost worse than no abstraction. I don't understand why you think lots of options are a bad thing?? For most embedded targets, people want/need easy knobs to turn to try and optimise the system level performance. I would guess that SH users want to do the same thing? That is what this does - it is just a easy to use knob. -Robin ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 13:26 ` Robin Getz @ 2007-03-05 14:00 ` Paul Mundt 2007-03-05 16:29 ` Robin Getz 0 siblings, 1 reply; 24+ messages in thread From: Paul Mundt @ 2007-03-05 14:00 UTC (permalink / raw) To: Robin Getz; +Cc: Bernd Schmidt, Wu, Bryan, Andrew Morton, linux-kernel On Mon, Mar 05, 2007 at 08:26:56AM -0500, Robin Getz wrote: > On Mon 5 Mar 2007 07:39, Paul Mundt pondered: > > On Mon, Mar 05, 2007 at 01:32:07PM +0100, Bernd Schmidt wrote: > > > This is done intentionally, because it's also possible for user code to > > > be loaded into L1 memory.?? We want to give users the option to avoid > > > filling it all up with kernel code. > > > > So then why not make the userspace component of it optional and allow a > > size cap for kernel usage that's configurable if it's enabled? This degree > > of abstraction is almost worse than no abstraction. > > I don't understand why you think lots of options are a bad thing?? > > For most embedded targets, people want/need easy knobs to turn to try and > optimise the system level performance. I would guess that SH users want to do > the same thing? > > That is what this does - it is just a easy to use knob. > This is hardly a knob, you're adding one config option per function to relocate in to the L1 memory, leaving it up to the user to decide what's best positioned there from the kernel point of view and what's left with userspace to play with. This is simply _not_ how you want to do this sort of interface, rather than making any sort of usability decisions, you've pushed it all on the user under the label of flexibility. What happens now if you suddenly start having other blocks of SRAM in future parts that are either shared across CPUs or just more buffer space for a single CPU? Do you start to repeat the config options for that space, too? There are things that will be a win to have located in on-chip SRAM, and others that will be less important. If you're concerned about this, you should simply pinpoint the hot paths that benefit the most from being relocated and weight that against a build-time configuration of how much room the user wants to play with. This way you can figure out all of your limits directly at link time, as you're arguably looking at an effectively static configuration anyways. If you really want to break it down on priority, use something similar to initcall levels. Start with the most critical bits, and stash as much as possible in whatever on-chip memory you have available (while heeding the user constraints) and then spill the rest to system memory. Throwing this all at the user simply shows that the functions being relocated haven't been profiled adequately with real workloads. You can't seriously expect your users to define what's the most timing critical and hope to get useful result. These are simply not things the user should ever _care_ about. If a user wants to use on-chip memory, presumably they have a target application in mind, and they know how much space they need. Beyond that, they expect the kernel to do the best it can with the space that's left over for it to play with. If a user has to sit around profiling their workload to figure out what config options to set to chew through the rest of the L1 memory, you've completely lost at intuitive design. This is like taking the KDE approach to UI design and applying it to the kernel, exposing every possible setting as a user-settable option and avoiding setting any sort of sane default in the hope that user knows best. This simply doesn't work. ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 14:00 ` Paul Mundt @ 2007-03-05 16:29 ` Robin Getz 2007-03-05 17:32 ` Paul Mundt 0 siblings, 1 reply; 24+ messages in thread From: Robin Getz @ 2007-03-05 16:29 UTC (permalink / raw) To: Paul Mundt; +Cc: Bernd Schmidt, Wu, Bryan, Andrew Morton, linux-kernel On Mon 5 Mar 2007 09:00, Paul Mundt pondered: > On Mon, Mar 05, 2007 at 08:26:56AM -0500, Robin Getz wrote: > > On Mon 5 Mar 2007 07:39, Paul Mundt pondered: > > > On Mon, Mar 05, 2007 at 01:32:07PM +0100, Bernd Schmidt wrote: > > > > That is what this does - it is just a easy to use knob. > > This is hardly a knob, you're adding one config option per function to > relocate in to the L1 memory, leaving it up to the user to decide what's > best positioned there from the kernel point of view and what's left with > userspace to play with. This is simply _not_ how you want to do this > sort > of interface, rather than making any sort of usability decisions, you've > pushed it all on the user under the label of flexibility. > > What happens now if you suddenly start having other blocks of SRAM in > future parts that are either shared across CPUs or just more buffer > space > for a single CPU? Do you start to repeat the config options for that > space, too? > > There are things that will be a win to have located in on-chip SRAM, and > others that will be less important. If you're concerned about this, you > should simply pinpoint the hot paths that benefit the most from being > relocated and weight that against a build-time configuration of how much > room the user wants to play with. This way you can figure out all of > your > limits directly at link time, as you're arguably looking at an > effectively static configuration anyways. > > If you really want to break it down on priority, use something similar > to > initcall levels. Start with the most critical bits, and stash as much as > possible in whatever on-chip memory you have available (while heeding > the > user constraints) and then spill the rest to system memory. > > Throwing this all at the user simply shows that the functions being > relocated haven't been profiled adequately with real workloads. Actually - that is not true at all - we have profiled these extensively across many workloads on various real world applications from fax machine to bar code scanner, to software radio, to internet radio - and that is the point - it all depends on what the user application/drivers that are loaded are doing. I can't decided what is best for the user, because I don't know what they are doing. > You can't seriously expect your users to define what's the most timing > critical and hope to get useful result. Yeah, we do - and the users who want to get the best performance - spend the effort in doing so - the others live with the defaults. > These are simply not things the user should ever _care_ about. Maybe users on Blackfin are more sophisticated than ones using SH? The reason we did this - is because we saw many end users doing the same thing, and wanted to give people a standard way to do it - that we could test/stand behind and give people support. > If a user > wants to use on-chip memory, presumably they have a target application > in mind, and they know how much space they need. Beyond that, they expect > the kernel to do the best it can with the space that's left over for it > to play with. And that is the problem - the kernel doesn't know how much on chip memory an application which will be loaded in the future might need. The user needs to be able to control the amount of on-chip memory that the kernel uses - hence the knobs. > If a user has to sit around profiling their workload to > figure out what config options to set to chew through the rest of the L1 > memory, you've completely lost at intuitive design. Embedded has never been intuitive. Normally, the majority of end users leave things set by the default, meaning their application doesn't use the on-chip memory, and the kernel is allowed to. If they want to start using on-chip memory, they are forced to do a re-compile, with the kernel functions using off-chip memory, and see how much of on-chip memory their application uses. They calculate the difference, and do a little system level profiling, and put the most used kernel functions back. Although this isn't intuitive., it is pretty straight forward and mechanical. People do it all the time. > This is like taking the KDE approach to UI design and applying it to the > kernel, exposing every possible setting as a user-settable option and > avoiding setting any sort of sane default in the hope that user knows > best. This simply doesn't work. There is a certain level of user sophistication expected when designing a deeply embedded application. I don't think this is a good comparison. Besides - I like (and use) all the knobs in KDE. You are asking me to try to decide what is best for all the potential end users, with out knowing what the end user is doing. What ends up happening is that people make modifications (normally poor ones) and then still ask for help when their modifications break. I would rather give them a supported (but arguablely confusing) method of customisation. (This is why it was added - end users were asking for it, and when we didn't have it, they would try to make the modifications themselves, break it, and report a bug. After telling 8 different people the same thing, when do you decide this is a feature you need to support?) -Robin -Robin ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 16:29 ` Robin Getz @ 2007-03-05 17:32 ` Paul Mundt 2007-03-05 22:06 ` Robin Getz 0 siblings, 1 reply; 24+ messages in thread From: Paul Mundt @ 2007-03-05 17:32 UTC (permalink / raw) To: Robin Getz; +Cc: Bernd Schmidt, Wu, Bryan, Andrew Morton, linux-kernel On Mon, Mar 05, 2007 at 11:29:19AM -0500, Robin Getz wrote: > On Mon 5 Mar 2007 09:00, Paul Mundt pondered: > > Throwing this all at the user simply shows that the functions being > > relocated haven't been profiled adequately with real workloads. > > Actually - that is not true at all - we have profiled these extensively > across many workloads on various real world applications from fax > machine to bar code scanner, to software radio, to internet radio - and > that is the point - it all depends on what the user application/drivers > that are loaded are doing. I can't decided what is best for the user, > because I don't know what they are doing. > And there are just as many cases that are common regardless of the workload that are simply good decisions. Exception vectors are a good example, either you get a boost from having them in L1 memory or you don't, the workload really doesn't matter. Workloads will of course have variations, but it's optimizing the hot paths that end up being quite common that end up being the most useful. These are not things that should be required from the end user. > > If a user wants to use on-chip memory, presumably they have a target > > application in mind, and they know how much space they need. Beyond > > that, they expect the kernel to do the best it can with the space > > that's left over for it to play with. > > And that is the problem - the kernel doesn't know how much on chip memory an > application which will be loaded in the future might need. The user needs to > be able to control the amount of on-chip memory that the kernel uses - hence > the knobs. > That's crap, for the exact reason that these are not run-time configurable knobs, they are static build-time components. If a user has to scale back the kernel users until their application fits, then they already know how much space they're going to need. Once you know that, reserving a chunk for userspace and limiting the kernel's usage is quite simple. > > If a user has to sit around profiling their workload to > > figure out what config options to set to chew through the rest of the L1 > > memory, you've completely lost at intuitive design. > > Embedded has never been intuitive. > Embedded has also never been about throwing everything at the user that's possible. Flexibility is good. Intelligent design is good. Blindly tossing everything the user's way and hoping for the best is not. > Normally, the majority of end users leave things set by the default, meaning > their application doesn't use the on-chip memory, and the kernel is allowed > to. If they want to start using on-chip memory, they are forced to do a > re-compile, with the kernel functions using off-chip memory, and see how much > of on-chip memory their application uses. They calculate the difference, and > do a little system level profiling, and put the most used kernel functions > back. > Yes, precisely. This is all something that should be done by the kernel _for_ the user. They can of course add their own instrumentation to the hot paths that their workload hits that aren't generally handled, but this should not be a required step. If I'm working on an embedded application, I expect my kernel to do some sensible things with the hardware available. I don't want to have to manually configure ever possible thing in order to get a system that behaves in a reasonable manner, and I certainly don't want to have to do this multiple times with a profiler for paths that are inevitably going to be quite common anyways. > You are asking me to try to decide what is best for all the potential end > users, with out knowing what the end user is doing. What ends up happening is > that people make modifications (normally poor ones) and then still ask for > help when their modifications break. > No, I'm asking you to set some sensible defaults, provide an interface that works for the common cases, and allow the special cases to remain simply that. > I would rather give them a supported (but arguablely confusing) method of > customisation. (This is why it was added - end users were asking for it, and > when we didn't have it, they would try to make the modifications themselves, > break it, and report a bug. After telling 8 different people the same thing, > when do you decide this is a feature you need to support?) > There's also a substantial differences between what customers are telling you they want, what they're trying to accomplish, and what they actually need. If you want to add something simple, setup an attribute tag for the section and tell the customers to flag the cases they're most concerned with. Obviously you expect them to be sitting there with a profiler open anyways, so this should not be a problem. Also note that we are not required to accept bad code in the kernel because you happen to have customer requirements for it. The point is, doing this as a config option design is just broken. It doesn't scale for new code, and it doesn't scale as soon as you start throwing in more blocks of SRAM with varying locality, latencies, etc. Littering around ifdefs for this stuff all over the code is equally ugly. I'll reiterate my original suggestion, allow the user to cap the amount of memory that's used by the kernel, then add something akin to initcall levels that you can use for prioritizing items to be relocated in on-chip memory, while spilling the rest to RAM. With this sort of scheme you can also trivially spill in to additional SRAM pools and whatever else you have handy as you go along. You can of course argue that as an embedded platform the user must make all of these decisions on their own, but there can be a compromise if the kernel's at least half-way intelligent about matters. Embedded has never been about hand-waving and expecting the user to solve all of the difficult problems just because they can. While it may not count for much, until something is done about this and the other issues that were raised earlier in the thread, I'd have to NAK this patch. If not wanting to manually select everything that goes in to L1 memory makes me too unsophisticated by blackfin standards, so be it ;-) ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 17:32 ` Paul Mundt @ 2007-03-05 22:06 ` Robin Getz 0 siblings, 0 replies; 24+ messages in thread From: Robin Getz @ 2007-03-05 22:06 UTC (permalink / raw) To: Paul Mundt; +Cc: Bernd Schmidt, Wu, Bryan, Andrew Morton, linux-kernel On Mon 5 Mar 2007 12:32, Paul Mundt pondered: > On Mon, Mar 05, 2007 at 11:29:19AM -0500, Robin Getz wrote: > > On Mon 5 Mar 2007 09:00, Paul Mundt pondered: > > > Throwing this all at the user simply shows that the functions being > > > relocated haven't been profiled adequately with real workloads. > > > > Actually - that is not true at all - we have profiled these > > extensively > > across many workloads on various real world applications from fax > > machine to bar code scanner, to software radio, to internet radio - > > and that is the point - it all depends on what the user > > application/drivers > > that are loaded are doing. I can't decided what is best for the user, > > because I don't know what they are doing. > > And there are just as many cases that are common regardless of the > workload that are simply good decisions. Exception vectors are a good > example, either you get a boost from having them in L1 memory or you > don't, the workload really doesn't matter. Then you have not profiled the same applications I have. I see different results in on this platform. > Workloads will of course have variations, but it's optimizing the hot > paths that end up being quite common that end up being the most useful. > These are not things that should be required from the end user. The end user is the only one who knows what is work load is. > > > If a user wants to use on-chip memory, presumably they have a target > > > application in mind, and they know how much space they need. Beyond > > > that, they expect the kernel to do the best it can with the space > > > that's left over for it to play with. > > > > And that is the problem - the kernel doesn't know how much on chip > > memory an application which will be loaded in the future might need. > > The user needs to be able to control the amount of on-chip memory > > that the kernel uses - hence the knobs. > > That's crap, for the exact reason that these are not run-time > configurable knobs, they are static build-time components. If a user has > to scale back the kernel users until their application fits, then they > already know how much space they're going to need. Once you know that, > reserving a chunk for userspace and limiting the kernel's usage is quite > simple. OK - you want me to remove 14 selections (yeah - I agree that is alot), and replace them with one selection, and a header file, that the user has to maintain separately from the .config, which everyone does already? IMHO - that is harder for an end user to maintain than our existing proposal. > > > If a user has to sit around profiling their workload to > > > figure out what config options to set to chew through the rest of > > > the L1 memory, you've completely lost at intuitive design. > > > > Embedded has never been intuitive. > > Embedded has also never been about throwing everything at the user > that's possible. Flexibility is good. Intelligent design is good. > Blindly tossing everything the user's way and hoping for the best > is not. default settings? Hope is not a strategy that we use. Careful design is (most of the time). > > Normally, the majority of end users leave things set by the default, > > meaning their application doesn't use the on-chip memory, and the > > kernel is allowed to. If they want to start using on-chip memory, > > they are forced to do a re-compile, with the kernel functions > > using off-chip memory, and see how much of on-chip memory > > their application uses. They calculate the difference, and > > do a little system level profiling, and put the most used kernel > > functions back. > > Yes, precisely. This is all something that should be done by the kernel > _for_ the user. They can of course add their own instrumentation to the > hot paths that their workload hits that aren't generally handled, but > this should not be a required step. > > If I'm working on an embedded application, I expect my kernel to do some > sensible things with the hardware available. I don't want to have to > manually configure ever possible thing in order to get a system that > behaves in a reasonable manner, and I certainly don't want to have to do > this multiple times with a profiler for paths that are inevitably going > to be quite common anyways. People do all kinds of things to make their application work better on a processor that cost less than lunch for one person. The applications always work without system level tuning - but they can work up to 2x faster with a little performance tweaking. To most people, spending the day or two - doing the tweaking to get max performance that the chip is capable of is worth the effort. > > You are asking me to try to decide what is best for all the potential > > end users, with out knowing what the end user is doing. What ends up > > happening is that people make modifications (normally poor ones) and > > then still ask for help when their modifications break. > > No, I'm asking you to set some sensible defaults, provide an interface > that works for the common cases, and allow the special cases to remain > simply that. In my opinion - that is what we have. > > I would rather give them a supported (but arguablely confusing) method > > of customisation. (This is why it was added - end users were asking for > > it, and when we didn't have it, they would try to make the > > modifications themselves, break it, and report a bug. After telling > > 8 different people the same thing, >> > There's also a substantial differences between what customers are > telling > you they want, what they're trying to accomplish, and what they actually > need. If you want to add something simple, setup an attribute tag for > the > section and tell the customers to flag the cases they're most concerned > with. Obviously you expect them to be sitting there with a profiler open > anyways, so this should not be a problem. This is what people do today (without our proposal). This is what was causing all the problems. This is what we are trying to avoid. > Also note that we are not required to accept bad code in the kernel > because you happen to have customer requirements for it. I am not asking you to accept bad code. I'm trying to explain why we came to a specific design decision, and went down a specific path, after thinking about and a similar idea as your suggestion. > The point is, doing this as a config option design is just broken. It > doesn't scale for new code, and it doesn't scale as soon as you start > throwing in more blocks of SRAM with varying locality, latencies, etc. > Littering around ifdefs for this stuff all over the code is equally > ugly. It is equally bad to force end users to maintain a patch set, to change around our best guess of what their fast paths might be in a .h file. > I'll reiterate my original suggestion, allow the user to cap the amount > of memory that's used by the kernel, then add something akin to initcall > levels that you can use for prioritizing items to be relocated in > on-chip memory, while spilling the rest to RAM. With this sort of scheme > you can also trivially spill in to additional SRAM pools and whatever > else you have handy as you go along. I don't understand how this is radically different than what we have today. You are suggesting that we take the knobs that are easy to modify, and hide them, to create pain for a user who wants to/needs to use them. I don't think this is good design either. > You can of course argue that as an embedded platform the user must make > all of these decisions on their own, but there can be a compromise if > the kernel's at least half-way intelligent about matters. I guess it is an disagreement in overall design philosophy - I thought the kernel was suppost to be flexible enough to handle the corner cases, and robust enough to ensure proper operation in all cases. What we are attempting to do (maybe poorly - jury is still out), is continue to add flexibility, putting the burden on us - the maintainers of all the files where the #ifdefs exist (all in arch/blackfin or include/asm-blackfin), to make it easier (both from use and maintenance) on end users. If your disagreement is - this doesn't make it easier, it makes it harder - then OK - but I don't think moving the controls somewhere else makes it easier. Plus I am not sure how I would explain things - "there are n levels of prioritizing if a function will be in slow or fast SRAM. Setting things to the highest priority may not mean it will go in the fastest SRAM, and you won't know until runtime, when your application doesn't run as fast as it should or by digging through the map file and looking at the address of the function". This is a worst case explanation, and I am sure that you could think of something better. I tend to be negative. What we have today - is a switch (not a knob). If you select too much, link fails. There is no "spill over" - by design. It is in on-chip SRAM, or not. > While it may not count for much, until something is done about this and > the other issues that were raised earlier in the thread, I'd have to NAK > this patch. If not wanting to manually select everything that goes in to > L1 memory makes me too unsophisticated by blackfin standards, so be it > ;-) It's not that we don't appreciate the thoughts and time that you have looking into this - we do and Thanks - I guess I would rather have something that: - hard failure, rather than soft failure. (it works, or it doesn't boot) - is maintained in .config file We are all about worrying about good design, proper code, and future maintenance/extendability. I think that this specific part of the patch meet those goals. As for the other pieces of the patch, that you pointed out - most need to be fixed per your comments. Thanks again for the time. -Robin ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 9:23 ` Paul Mundt 2007-03-05 12:32 ` Bernd Schmidt @ 2007-03-06 2:04 ` Mike Frysinger 2007-03-21 15:44 ` Mike Frysinger 2 siblings, 0 replies; 24+ messages in thread From: Mike Frysinger @ 2007-03-06 2:04 UTC (permalink / raw) To: Paul Mundt, Wu, Bryan, Andrew Morton, linux-kernel On 3/5/07, Paul Mundt <lethal@linux-sh.org> wrote: > > +config DEBUG_HUNT_FOR_ZERO > > + bool "Catch NULL pointer reads/writes" > > + default y > > Is this sane to have conditional? yes ... with the ability to change the load address of the kernel, you can create a reserved chunk of memory for users to utilize in their custom kernel/userspace apps an abuse of no-mmu to be sure ;) > > +#ifdef CONFIG_BUG > > +#define HAVE_ARCH_BUG > > +#define BUG() do { \ > > + dump_stack(); \ > > + printk(KERN_WARNING "\nkernel BUG at %s:%d!\n",\ > > + __FILE__, __LINE__); \ > > + panic("BUG!"); \ > > +} while (0) > > +#endif > > + > > +#include <asm-generic/bug.h> > > +#endif > > What do you need HAVE_ARCH_BUG for? You're not doing anything with it.. last i checked, the default BUG() does not include the dump_stack() ... if things have changed, then sure the generic bug implementation is OK, otherwise this helps us with debugging -mike ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-05 9:23 ` Paul Mundt 2007-03-05 12:32 ` Bernd Schmidt 2007-03-06 2:04 ` Mike Frysinger @ 2007-03-21 15:44 ` Mike Frysinger 2007-03-21 23:42 ` Paul Mundt 2 siblings, 1 reply; 24+ messages in thread From: Mike Frysinger @ 2007-03-21 15:44 UTC (permalink / raw) To: Paul Mundt, Wu, Bryan, Andrew Morton, linux-kernel On 3/5/07, Paul Mundt <lethal@linux-sh.org> wrote: > > +int __init oprofile_arch_init(struct oprofile_operations *ops) > > +{ > > +#ifdef CONFIG_HARDWARE_PM > [snip] > > +#else > > + return -1; > > +#endif > > +} > > Uh.. fix your dependencies. can you elaborate ? -mike ^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH -mm 1/5] Blackfin: blackfin architecture patch update 2007-03-21 15:44 ` Mike Frysinger @ 2007-03-21 23:42 ` Paul Mundt 0 siblings, 0 replies; 24+ messages in thread From: Paul Mundt @ 2007-03-21 23:42 UTC (permalink / raw) To: Mike Frysinger; +Cc: Wu, Bryan, Andrew Morton, linux-kernel On Wed, Mar 21, 2007 at 11:44:27AM -0400, Mike Frysinger wrote: > On 3/5/07, Paul Mundt <lethal@linux-sh.org> wrote: > >> +int __init oprofile_arch_init(struct oprofile_operations *ops) > >> +{ > >> +#ifdef CONFIG_HARDWARE_PM > >[snip] > >> +#else > >> + return -1; > >> +#endif > >> +} > > > >Uh.. fix your dependencies. > > can you elaborate ? This code does nothing if CONFIG_HARDWARE_PM isn't enabled, so your best bet is to simply have an op_model_null for the cases where you can't support the counters and work out the dependencies in the Makefile. Trying to hack both of these together in one driver is not the way to go. ^ permalink raw reply [flat|nested] 24+ messages in thread
end of thread, other threads:[~2007-03-21 23:46 UTC | newest] Thread overview: 24+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2007-03-01 4:14 [PATCH -mm 1/5] Blackfin: blackfin architecture patch update Wu, Bryan 2007-03-03 20:38 ` Arnd Bergmann 2007-03-05 7:13 ` Wu, Bryan 2007-03-03 22:30 ` Arnd Bergmann 2007-03-03 22:50 ` bert hubert 2007-03-03 23:05 ` Arnd Bergmann 2007-03-05 6:54 ` Aubrey Li 2007-03-05 8:47 ` Arnd Bergmann 2007-03-05 9:19 ` Wu, Bryan 2007-03-05 16:43 ` Arnd Bergmann 2007-03-05 7:34 ` Wu, Bryan 2007-03-05 8:10 ` Arnd Bergmann 2007-03-06 2:09 ` Mike Frysinger 2007-03-05 9:23 ` Paul Mundt 2007-03-05 12:32 ` Bernd Schmidt 2007-03-05 12:39 ` Paul Mundt 2007-03-05 13:26 ` Robin Getz 2007-03-05 14:00 ` Paul Mundt 2007-03-05 16:29 ` Robin Getz 2007-03-05 17:32 ` Paul Mundt 2007-03-05 22:06 ` Robin Getz 2007-03-06 2:04 ` Mike Frysinger 2007-03-21 15:44 ` Mike Frysinger 2007-03-21 23:42 ` Paul Mundt
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox