* [PATCH 0/4][Diskdump]Update patches
@ 2004-06-16 12:39 Takao Indoh
2004-06-16 12:45 ` [PATCH 1/4][Diskdump]Update patches Takao Indoh
` (4 more replies)
0 siblings, 5 replies; 9+ messages in thread
From: Takao Indoh @ 2004-06-16 12:39 UTC (permalink / raw)
To: linux-kernel
Hi!
I fixed diskdump patches except timer problem.
- Fix some codes which Arjan van de Ven pointed out
- Replace dev_t with block_device
Source code of tool(diskdumptuils) can be downloaded from
http://sourceforge.net/projects/lkdump
Regarding timer problem, please see previous mail I sent.
http://marc.theaimsgroup.com/?l=linux-kernel&m=108722344204595&w=2
In short, Timer problem is as follows.
> What is a problem? Scsi driver uses timer/tasklet to complete I/O.
> But, diskdump disables interrupt, so timer/taslket doesn't work!
There are three ways to solve this problem so far.
(1) Redefine timer/taslket routines. This method was adopted in the
first patch.
+#if defined(CONFIG_DISKDUMP) || defined(CONFIG_DISKDUMP_MODULE)
+#undef add_timer
+#define add_timer diskdump_add_timer
+#undef del_timer_sync
+#define del_timer_sync diskdump_del_timer
+#undef del_timer
+#define del_timer diskdump_del_timer
+#undef mod_timer
+#define mod_timer diskdump_mod_timer
+
+#define tasklet_schedule diskdump_tasklet_schedule
+#endif
(2) Add new code into the timer/taslket routines themselves.
static inline void add_timer(struct timer_list * timer)
{
- __mod_timer(timer, timer->expires);
+ if(crashdump_mode())
+ diskdump_add_timer(timer);
+ else
+ __mod_timer(timer, timer->expires);
}
int del_timer_sync(struct timer_list *timer)
{
tvec_base_t *base;
int i, ret = 0;
+ if(crashdump_mode()) {
+ diskdump_del_timer(timer);
+ return 0;
+ }
+
check_timer(timer);
int mod_timer(struct timer_list *timer, unsigned long expires)
{
BUG_ON(!timer->function);
+ if(crashdump_mode()) {
+ diskdump_mod_timer(timer, expires);
+ return 0;
+ }
+
check_timer(timer);
static inline void tasklet_schedule(struct tasklet_struct *t)
{
+ if(crashdump_mode()) {
+ diskdump_tasklet_schedule(t);
+ return;
+ }
+
if (!test_and_set_bit(TASKLET_STATE_SCHED, &t->state))
__tasklet_schedule(t);
}
(3) Change polling handler of driver not to use timer/tasklet.
The method (1) was already rejected because of its ugliness. The
method (3) needs many extra codes and makes handler of driver too big
and complex. I think the method (2) is the simplest though it make
common codes dirty.
Please feel free to comment.
Best Regards,
Takao Indoh
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH 1/4][Diskdump]Update patches
2004-06-16 12:39 [PATCH 0/4][Diskdump]Update patches Takao Indoh
@ 2004-06-16 12:45 ` Takao Indoh
2004-06-16 12:46 ` [PATCH 2/4][Diskdump]Update patches Takao Indoh
` (3 subsequent siblings)
4 siblings, 0 replies; 9+ messages in thread
From: Takao Indoh @ 2004-06-16 12:45 UTC (permalink / raw)
To: linux-kernel
This is a patch for diskdump common layer.
diff -Nur linux-2.6.6.org/arch/i386/kernel/nmi.c linux-2.6.6/arch/i386/kernel/nmi.c
--- linux-2.6.6.org/arch/i386/kernel/nmi.c 2004-06-04 21:21:51.000000000 +0900
+++ linux-2.6.6/arch/i386/kernel/nmi.c 2004-06-16 19:34:16.000000000 +0900
@@ -517,3 +517,4 @@
EXPORT_SYMBOL(release_lapic_nmi);
EXPORT_SYMBOL(disable_timer_nmi_watchdog);
EXPORT_SYMBOL(enable_timer_nmi_watchdog);
+EXPORT_SYMBOL(touch_nmi_watchdog);
diff -Nur linux-2.6.6.org/arch/i386/kernel/reboot.c linux-2.6.6/arch/i386/kernel/reboot.c
--- linux-2.6.6.org/arch/i386/kernel/reboot.c 2004-06-04 21:21:51.000000000 +0900
+++ linux-2.6.6/arch/i386/kernel/reboot.c 2004-06-16 19:34:16.000000000 +0900
@@ -252,7 +252,8 @@
* Stop all CPUs and turn off local APICs and the IO-APIC, so
* other OSs see a clean IRQ state.
*/
- smp_send_stop();
+ if (!crashdump_mode())
+ smp_send_stop();
#elif defined(CONFIG_X86_LOCAL_APIC)
if (cpu_has_apic) {
local_irq_disable();
diff -Nur linux-2.6.6.org/arch/i386/kernel/traps.c linux-2.6.6/arch/i386/kernel/traps.c
--- linux-2.6.6.org/arch/i386/kernel/traps.c 2004-06-04 21:21:51.000000000 +0900
+++ linux-2.6.6/arch/i386/kernel/traps.c 2004-06-16 19:34:16.000000000 +0900
@@ -258,7 +258,8 @@
int nl = 0;
console_verbose();
- spin_lock_irq(&die_lock);
+ if (!crashdump_mode())
+ spin_lock_irq(&die_lock);
bust_spinlocks(1);
handle_BUG(regs);
printk("%s: %04lx [#%d]\n", str, err & 0xffff, ++die_counter);
@@ -277,6 +278,7 @@
if (nl)
printk("\n");
show_registers(regs);
+ try_crashdump(regs);
bust_spinlocks(0);
spin_unlock_irq(&die_lock);
if (in_interrupt())
diff -Nur linux-2.6.6.org/arch/i386/mm/init.c linux-2.6.6/arch/i386/mm/init.c
--- linux-2.6.6.org/arch/i386/mm/init.c 2004-06-04 21:21:52.000000000 +0900
+++ linux-2.6.6/arch/i386/mm/init.c 2004-06-16 19:34:16.000000000 +0900
@@ -168,7 +168,7 @@
extern int is_available_memory(efi_memory_desc_t *);
-static inline int page_is_ram(unsigned long pagenr)
+int page_is_ram(unsigned long pagenr)
{
int i;
unsigned long addr, end;
@@ -205,6 +205,7 @@
}
return 0;
}
+EXPORT_SYMBOL(page_is_ram);
#ifdef CONFIG_HIGHMEM
pte_t *kmap_pte;
diff -Nur linux-2.6.6.org/drivers/block/Kconfig linux-2.6.6/drivers/block/Kconfig
--- linux-2.6.6.org/drivers/block/Kconfig 2004-06-04 21:22:15.000000000 +0900
+++ linux-2.6.6/drivers/block/Kconfig 2004-06-16 19:34:16.000000000 +0900
@@ -346,6 +346,11 @@
your machine, or if you want to have a raid or loopback device
bigger than 2TB. Otherwise say N.
+config DISKDUMP
+ tristate "Disk dump support"
+ ---help---
+ Disk dump support.
+
source "drivers/s390/block/Kconfig"
endmenu
diff -Nur linux-2.6.6.org/drivers/block/Makefile linux-2.6.6/drivers/block/Makefile
--- linux-2.6.6.org/drivers/block/Makefile 2004-06-04 21:22:15.000000000 +0900
+++ linux-2.6.6/drivers/block/Makefile 2004-06-16 19:34:16.000000000 +0900
@@ -43,3 +43,4 @@
obj-$(CONFIG_VIODASD) += viodasd.o
obj-$(CONFIG_BLK_DEV_CARMEL) += carmel.o
+obj-$(CONFIG_DISKDUMP) += diskdump.o diskdumplib.o
diff -Nur linux-2.6.6.org/drivers/block/diskdump.c linux-2.6.6/drivers/block/diskdump.c
--- linux-2.6.6.org/drivers/block/diskdump.c 1970-01-01 09:00:00.000000000 +0900
+++ linux-2.6.6/drivers/block/diskdump.c 2004-06-16 19:34:16.000000000 +0900
@@ -0,0 +1,973 @@
+/*
+ * linux/drivers/block/diskdump.c
+ *
+ * Copyright (C) 2004 FUJITSU LIMITED
+ * Copyright (C) 2002 Red Hat, Inc.
+ * Written by Nobuhiro Tachino (ntachino@jp.fujitsu.com)
+ *
+ * Some codes were derived from netdump and copyright belongs to
+ * Red Hat, Inc.
+ */
+/*
+ * 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, 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, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ */
+
+#include <linux/mm.h>
+#include <linux/init.h>
+#include <linux/delay.h>
+#include <linux/file.h>
+#include <linux/reboot.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/kernel.h>
+#include <linux/highmem.h>
+#include <linux/utsname.h>
+#include <linux/console.h>
+#include <linux/smp_lock.h>
+#include <linux/nmi.h>
+#include <linux/genhd.h>
+#include <linux/crc32.h>
+#include <linux/proc_fs.h>
+#include <linux/seq_file.h>
+#include <linux/slab.h>
+#include <linux/blkdev.h>
+#include <linux/diskdump.h>
+#include <linux/diskdumplib.h>
+#include <asm/diskdump.h>
+
+#define Dbg(x, ...) pr_debug("disk_dump: " x "\n", ## __VA_ARGS__)
+#define Err(x, ...) pr_err ("disk_dump: " x "\n", ## __VA_ARGS__)
+#define Warn(x, ...) pr_warn ("disk_dump: " x "\n", ## __VA_ARGS__)
+#define Info(x, ...) pr_info ("disk_dump: " x "\n", ## __VA_ARGS__)
+
+#ifndef TRUE
+#define TRUE !0
+#endif
+#ifndef FALSE
+#define FALSE 0
+#endif
+
+#define ROUNDUP(x, y) (((x) + ((y)-1))/(y))
+
+/* 512byte sectors to blocks */
+#define SECTOR_BLOCK(s) ((s) >> (DUMP_BLOCK_SHIFT - 9))
+
+static unsigned int fallback_on_err = 1;
+static unsigned int allow_risky_dumps = 1;
+static unsigned int block_order = 2;
+static int sample_rate = 8;
+module_param(fallback_on_err, uint, 0);
+module_param(allow_risky_dumps, uint, 0);
+module_param(block_order, uint, 0);
+module_param(sample_rate, int, 0);
+
+static unsigned long timestamp_1sec;
+static uint32_t module_crc;
+static char *scratch;
+static struct disk_dump_header dump_header;
+static struct disk_dump_sub_header dump_sub_header;
+
+/* Registered dump devices */
+static LIST_HEAD(disk_dump_devices);
+
+/* Registered dump types, e.g. SCSI, ... */
+static LIST_HEAD(disk_dump_types);
+
+static spinlock_t disk_dump_lock = SPIN_LOCK_UNLOCKED;
+
+static unsigned int header_blocks; /* The size of all headers */
+static unsigned int bitmap_blocks; /* The size of bitmap header */
+static unsigned int total_ram_blocks; /* The size of memory */
+static unsigned int total_blocks; /* The sum of above */
+
+struct notifier_block *disk_dump_notifier_list;
+
+extern int panic_timeout;
+extern unsigned long max_pfn;
+
+#if CONFIG_SMP
+static void freeze_cpu(void *dummy)
+{
+ unsigned int cpu = smp_processor_id();
+
+ dump_header.tasks[cpu] = current;
+
+ platform_freeze_cpu();
+}
+#endif
+
+static int lapse = 0; /* 200msec unit */
+
+static inline unsigned long eta(unsigned long nr, unsigned long maxnr)
+{
+ unsigned long long eta;
+
+ eta = ((maxnr << 8) / nr) * (unsigned long long)lapse;
+
+ return (unsigned long)(eta >> 8) - lapse;
+}
+
+static inline void print_status(unsigned int nr, unsigned int maxnr)
+{
+ static char *spinner = "/|\\-";
+ static unsigned long long prev_timestamp = 0;
+ unsigned long long timestamp;
+
+ platform_timestamp(timestamp);
+
+ if (timestamp - prev_timestamp > (timestamp_1sec/5)) {
+ prev_timestamp = timestamp;
+ lapse++;
+ printk("%u/%u %lu ETA %c \r",
+ nr, maxnr, eta(nr, maxnr) / 5, spinner[lapse & 3]);
+ }
+}
+
+static inline void clear_status(int nr, int maxnr)
+{
+ printk(" \r");
+ lapse = 0;
+}
+
+/*
+ * Checking the signature on a block. The format is as follows.
+ *
+ * 1st word = 'disk'
+ * 2nd word = 'dump'
+ * 3rd word = block number
+ * 4th word = ((block number + 7) * 11) & 0xffffffff
+ * 5th word = ((4th word + 7)* 11) & 0xffffffff
+ * ..
+ *
+ * Return TRUE if the signature is correct, else return FALSE
+ */
+static int check_block_signature(void *buf, unsigned int block_nr)
+{
+ int word_nr = PAGE_SIZE / sizeof(int);
+ int *words = buf;
+ unsigned int val;
+ int i;
+
+ if (memcmp(buf, DUMP_PARTITION_SIGNATURE, sizeof(*words)))
+ return FALSE;
+
+ val = block_nr;
+ for (i = 2; i < word_nr; i++) {
+ if (words[i] != val)
+ return FALSE;
+ val = (val + 7) * 11;
+ }
+
+ return TRUE;
+}
+
+/*
+ * Read one block into the dump partition
+ */
+static int read_blocks(struct disk_dump_partition *dump_part, unsigned int nr, char *buf, int len)
+{
+ struct disk_dump_device *device = dump_part->device;
+ int ret;
+
+ touch_nmi_watchdog();
+ ret = device->ops.rw_block(dump_part, READ, nr, buf, len);
+ if (ret < 0) {
+ Err("read error on block %u", nr);
+ return ret;
+ }
+ return 0;
+}
+
+static int write_blocks(struct disk_dump_partition *dump_part, unsigned int offs, char *buf, int len)
+{
+ struct disk_dump_device *device = dump_part->device;
+ int ret;
+
+ touch_nmi_watchdog();
+ ret = device->ops.rw_block(dump_part, WRITE, offs, buf, len);
+ if (ret < 0) {
+ Err("write error on block %u", offs);
+ return ret;
+ }
+ return 0;
+}
+
+/*
+ * Initialize the common header
+ */
+
+/*
+ * Write the common header
+ */
+static int write_header(struct disk_dump_partition *dump_part)
+{
+ memset(scratch, '\0', PAGE_SIZE);
+ memcpy(scratch, &dump_header, sizeof(dump_header));
+
+ return write_blocks(dump_part, 1, scratch, 1);
+}
+
+/*
+ * Check the signaures in all blocks of the dump partition
+ * Return TRUE if the signature is correct, else return FALSE
+ */
+static int check_dump_partition(struct disk_dump_partition *dump_part, unsigned int partition_size)
+{
+ unsigned int blk;
+ int ret;
+ unsigned int chunk_blks, skips;
+ int i;
+
+ if (sample_rate < 0) /* No check */
+ return TRUE;
+
+ /*
+ * If the device has limitations of transfer size, use it.
+ */
+ chunk_blks = 1 << block_order;
+ if (dump_part->device->max_blocks)
+ chunk_blks = min(chunk_blks, dump_part->device->max_blocks);
+ skips = chunk_blks << sample_rate;
+
+ lapse = 0;
+ for (blk = 0; blk < partition_size; blk += skips) {
+ unsigned int len;
+redo:
+ len = min(chunk_blks, partition_size - blk);
+ if ((ret = read_blocks(dump_part, blk, scratch, len)) < 0)
+ return FALSE;
+ print_status(blk + 1, partition_size);
+ for (i = 0; i < len; i++)
+ if (!check_block_signature(scratch + i * DUMP_BLOCK_SIZE, blk + i)) {
+ Err("bad signature in block %u", blk + i);
+ return FALSE;
+ }
+ }
+ /* Check the end of the dump partition */
+ if (blk - skips + chunk_blks < partition_size) {
+ blk = partition_size - chunk_blks;
+ goto redo;
+ }
+ clear_status(blk, partition_size);
+ return TRUE;
+}
+
+/*
+ * Write memory bitmap after location of dump headers.
+ */
+#define IDX2PAGENR(nr, byte, bit) (((nr) * PAGE_SIZE + (byte)) * 8 + (bit))
+static int write_bitmap(struct disk_dump_partition *dump_part, unsigned int bitmap_offset, unsigned int bitmap_blocks)
+{
+ unsigned int nr;
+ int bit, byte;
+ int ret = 0;
+ unsigned char val;
+
+ for (nr = 0; nr < bitmap_blocks; nr++) {
+ for (byte = 0; byte < PAGE_SIZE; byte++) {
+ val = 0;
+ for (bit = 0; bit < 8; bit++)
+ if (page_is_ram(IDX2PAGENR(nr, byte, bit)))
+ val |= (1 << bit);
+ scratch[byte] = (char)val;
+ }
+ if ((ret = write_blocks(dump_part, bitmap_offset + nr, scratch, 1)) < 0) {
+ Err("I/O error %d on block %u", ret, bitmap_offset + nr);
+ break;
+ }
+ }
+ return ret;
+}
+
+/*
+ * Write whole memory to dump partition.
+ * Return value is the number of writen blocks.
+ */
+static int write_memory(struct disk_dump_partition *dump_part, int offset, unsigned int max_blocks_written, unsigned int *blocks_written)
+{
+ char *kaddr;
+ unsigned int blocks = 0;
+ struct page *page;
+ unsigned int nr;
+ int ret = 0;
+ int blk_in_chunk = 0;
+
+ for (nr = 0; nr < max_pfn; nr++) {
+ print_status(blocks, max_blocks_written);
+
+ if (!page_is_ram(nr))
+ continue;
+
+ if (blocks >= max_blocks_written) {
+ Warn("dump device is too small. %lu pages were not saved", max_pfn - blocks);
+ goto out;
+ }
+ page = pfn_to_page(nr);
+ kaddr = (char *)kmap_atomic(page, KM_DISKDUMP);
+ /*
+ * need to copy because adapter drivers use virt_to_bus()
+ */
+ memcpy(scratch + blk_in_chunk * PAGE_SIZE, kaddr, PAGE_SIZE);
+ blk_in_chunk++;
+ blocks++;
+ kunmap_atomic(kaddr, KM_DISKDUMP);
+
+ if (blk_in_chunk >= (1 << block_order)) {
+ ret = write_blocks(dump_part, offset, scratch, blk_in_chunk);
+ if (ret < 0) {
+ Err("I/O error %d on block %u", ret, offset);
+ break;
+ }
+ offset += blk_in_chunk;
+ blk_in_chunk = 0;
+ }
+ }
+ if (ret >= 0 && blk_in_chunk > 0) {
+ ret = write_blocks(dump_part, offset, scratch, blk_in_chunk);
+ if (ret < 0)
+ Err("I/O error %d on block %u", ret, offset);
+ }
+
+out:
+ clear_status(nr, max_blocks_written);
+
+ *blocks_written = blocks;
+ return ret;
+}
+
+/*
+ * Select most suitable dump device. sanity_check() returns the state
+ * of each dump device. 0 means OK, negative value means NG, and
+ * positive value means it maybe work. select_dump_partition() first
+ * try to select a sane device and if it has no sane device and
+ * allow_risky_dumps is set, it select one from maybe OK devices.
+ *
+ * XXX We cannot handle multiple partitions yet.
+ */
+static struct disk_dump_partition *select_dump_partition(void)
+{
+ struct disk_dump_device *dump_device;
+ struct disk_dump_partition *dump_part;
+ int sanity;
+ int strict_check = 1;
+
+redo:
+ /*
+ * Select a sane polling driver.
+ */
+ list_for_each_entry(dump_device, &disk_dump_devices, list) {
+ sanity = 0;
+ if (dump_device->ops.sanity_check)
+ sanity = dump_device->ops.sanity_check(dump_device);
+ if (sanity < 0 || (sanity > 0 && strict_check))
+ continue;
+ list_for_each_entry(dump_part, &dump_device->partitions, list)
+ return dump_part;
+ }
+ if (allow_risky_dumps && strict_check) {
+ strict_check = 0;
+ goto redo;
+ }
+ return NULL;
+}
+
+static void disk_dump(struct pt_regs *regs, void *platform_arg)
+{
+ unsigned long flags;
+ int ret = -EIO;
+ struct pt_regs myregs;
+ unsigned int max_written_blocks, written_blocks;
+ int i;
+ struct disk_dump_device *dump_device = NULL;
+ struct disk_dump_partition *dump_part = NULL;
+
+ /* Inhibit interrupt and stop other CPUs */
+ local_save_flags(flags);
+ local_irq_disable();
+
+ diskdump_lib_init();
+
+ /*
+ * Check the checksum of myself
+ */
+ if (!spin_trylock(&disk_dump_lock)) {
+ Err("disk_dump_lock failed.");
+ goto done;
+ }
+
+ spin_trylock(&disk_dump_lock);
+ if (!check_crc_module()) {
+ Err("checksum error. diskdump common module may be compromised.");
+ goto done;
+ }
+
+ diskdump_mode = 1;
+
+ Dbg("notify dump start.");
+ notifier_call_chain(&disk_dump_notifier_list, 0, NULL);
+
+ dump_header.tasks[smp_processor_id()] = current;
+#if CONFIG_SMP
+ smp_call_function(freeze_cpu, NULL, 1, 0);
+ mdelay(3000);
+ printk("CPU frozen: ");
+ for (i = 0; i < NR_CPUS; i++) {
+ if (dump_header.tasks[i] != NULL)
+ printk("#%d", i);
+
+ }
+ printk("\n");
+ printk("CPU#%d is executing diskdump.\n", smp_processor_id());
+#else
+ mdelay(1000);
+#endif
+
+ platform_fix_regs();
+
+ if (list_empty(&disk_dump_devices)) {
+ Err("adapter driver is not registered.");
+ goto done;
+ }
+
+ printk("start dumping\n");
+
+ if (!(dump_part = select_dump_partition())) {
+ Err("No sane dump device found");
+ goto done;
+ }
+ dump_device = dump_part->device;
+
+ /*
+ * Stop ongoing I/O with polling driver and make the shift to I/O mode
+ * for dump
+ */
+ Dbg("do quiesce");
+ if (dump_device->ops.quiesce)
+ if ((ret = dump_device->ops.quiesce(dump_device)) < 0) {
+ Err("quiesce failed. error %d", ret);
+ goto done;
+ }
+
+ if (SECTOR_BLOCK(dump_part->nr_sects) < header_blocks + bitmap_blocks) {
+ Warn("dump partition is too small. Aborted");
+ goto done;
+ }
+
+ /* Check dump partition */
+ printk("check dump partition...\n");
+ if (!check_dump_partition(dump_part, total_blocks)) {
+ Err("check partition failed.");
+ goto done;
+ }
+
+ /*
+ * Write the common header
+ */
+ memcpy(dump_header.signature, DISK_DUMP_SIGNATURE, sizeof(dump_header.signature));
+ dump_header.utsname = system_utsname;
+ dump_header.timestamp = xtime;
+ dump_header.status = DUMP_HEADER_INCOMPLETED;
+ dump_header.block_size = PAGE_SIZE;
+ dump_header.sub_hdr_size = size_of_sub_header();
+ dump_header.bitmap_blocks = bitmap_blocks;
+ dump_header.max_mapnr = max_pfn;
+ dump_header.total_ram_blocks = total_ram_blocks;
+ dump_header.device_blocks = SECTOR_BLOCK(dump_part->nr_sects);
+ dump_header.current_cpu = smp_processor_id();
+ dump_header.nr_cpus = num_online_cpus();
+ dump_header.written_blocks = 2;
+
+ write_header(dump_part);
+
+ /*
+ * Write the architecture dependent header
+ */
+ Dbg("write sub header");
+ if ((ret = write_sub_header()) < 0) {
+ Err("writing sub header failed. error %d", ret);
+ goto done;
+ }
+
+ Dbg("writing memory bitmaps..");
+ if ((ret = write_bitmap(dump_part, header_blocks, bitmap_blocks)) < 0)
+ goto done;
+
+ max_written_blocks = total_ram_blocks;
+ if (dump_header.device_blocks < total_blocks) {
+ Warn("dump partition is too small. actual blocks %u. expected blocks %u. whole memory will not be saved",
+ dump_header.device_blocks, total_blocks);
+ max_written_blocks -= (total_blocks - dump_header.device_blocks);
+ }
+
+ dump_header.written_blocks += dump_header.sub_hdr_size;
+ dump_header.written_blocks += dump_header.bitmap_blocks;
+ write_header(dump_part);
+
+ printk("dumping memory..\n");
+ if ((ret = write_memory(dump_part, header_blocks + bitmap_blocks,
+ max_written_blocks, &written_blocks)) < 0)
+ goto done;
+
+ /*
+ * Set the number of block that is written into and write it
+ * into partition again.
+ */
+ dump_header.written_blocks += written_blocks;
+ dump_header.status = DUMP_HEADER_COMPLETED;
+ write_header(dump_part);
+
+ ret = 0;
+
+done:
+ Dbg("do adapter shutdown.");
+ if (dump_device && dump_device->ops.shutdown)
+ if (dump_device->ops.shutdown(dump_device))
+ Err("adapter shutdown failed.");
+
+ /*
+ * If diskdump failed and fallback_on_err is set,
+ * We just return and leave panic to netdump.
+ */
+ if (fallback_on_err && ret != 0)
+ return;
+
+ Dbg("notify panic.");
+ notifier_call_chain(&panic_notifier_list, 0, NULL);
+
+ diskdump_lib_exit();
+
+ if (panic_timeout > 0) {
+ int i;
+
+ printk(KERN_EMERG "Rebooting in %d second%s..",
+ panic_timeout, "s" + (panic_timeout == 1));
+ for (i = 0; i < panic_timeout; i++) {
+ touch_nmi_watchdog();
+ mdelay(1000);
+ }
+ printk("\n");
+ machine_restart(NULL);
+ }
+ printk(KERN_EMERG "halt\n");
+ for (;;) {
+ touch_nmi_watchdog();
+ machine_halt();
+ mdelay(1000);
+ }
+}
+
+static struct disk_dump_partition *find_dump_partition(dev_t dev)
+{
+ struct disk_dump_device *dump_device;
+ struct disk_dump_partition *dump_part;
+
+ list_for_each_entry(dump_device, &disk_dump_devices, list)
+ list_for_each_entry(dump_part, &dump_device->partitions, list)
+ if (dump_part->dentry->d_inode->i_rdev == dev)
+ return dump_part;
+ return NULL;
+}
+
+static struct disk_dump_device *find_dump_device(void *real_device)
+{
+ struct disk_dump_device *dump_device;
+
+ list_for_each_entry(dump_device, &disk_dump_devices, list)
+ if (real_device == dump_device->device)
+ return dump_device;
+ return NULL;
+}
+
+static void *find_real_device(struct block_device *bdev, struct disk_dump_type **_dump_type)
+{
+ void *real_device;
+ struct disk_dump_type *dump_type;
+
+ list_for_each_entry(dump_type, &disk_dump_types, list)
+ if ((real_device = dump_type->probe(bdev)) != NULL) {
+ *_dump_type = dump_type;
+ return real_device;
+ }
+ return NULL;
+}
+
+/*
+ * Add dump partition structure corresponding to file to the dump device
+ * structure.
+ */
+static int add_dump_partition(struct disk_dump_device *dump_device, struct file *file)
+{
+ struct disk_dump_partition *dump_part;
+ struct inode *inode = file->f_dentry->d_inode;
+ dev_t dev = inode->i_rdev;
+ struct block_device *bdev = inode->i_bdev;
+ char buffer[32];
+
+ if (!(dump_part = kmalloc(sizeof(*dump_part), GFP_KERNEL)))
+ return -ENOMEM;
+
+ dump_part->device = dump_device;
+ dump_part->vfsmount = mntget(file->f_vfsmnt);
+ dump_part->dentry = dget(file->f_dentry);
+
+ if (!bdev || !bdev->bd_part)
+ return -EINVAL;
+ dump_part->nr_sects = bdev->bd_part->nr_sects;
+ dump_part->start_sect = bdev->bd_part->start_sect;
+
+ if (SECTOR_BLOCK(dump_part->nr_sects) < total_blocks) {
+ format_dev_t(buffer, dev);
+ Warn("%s is too small to save whole system memory\n", buffer);
+ }
+
+ list_add(&dump_part->list, &dump_device->partitions);
+
+ return 0;
+}
+
+/*
+ * Add dump partition corresponding to file.
+ * Must be called with disk_dump_lock held.
+ */
+static int add_dump(struct file *file)
+{
+ struct disk_dump_type *dump_type = NULL;
+ struct disk_dump_device *dump_device;
+ void *real_device;
+ struct inode *dump_inode = file->f_dentry->d_inode;
+ dev_t dev = dump_inode->i_rdev;
+ int ret;
+
+ /* Check whether this inode is already registered */
+ if (find_dump_partition(dev))
+ return -EEXIST;
+
+ /* find dump_type and real device for this inode */
+ if (!(real_device = find_real_device(dump_inode->i_bdev, &dump_type)))
+ return -ENXIO;
+
+ dump_device = find_dump_device(real_device);
+ if (dump_device == NULL) {
+ /* real_device is not registered. create new dump_device */
+ if (!(dump_device = kmalloc(sizeof(*dump_device), GFP_KERNEL)))
+ return -ENOMEM;
+
+ memset(dump_device, 0, sizeof(*dump_device));
+ INIT_LIST_HEAD(&dump_device->partitions);
+
+ dump_device->dump_type = dump_type;
+ dump_device->device = real_device;
+ if ((ret = dump_type->add_device(dump_device)) < 0) {
+ kfree(dump_device);
+ return ret;
+ }
+ if (!try_module_get(dump_type->owner))
+ return -EINVAL;
+ list_add(&dump_device->list, &disk_dump_devices);
+ }
+
+ ret = add_dump_partition(dump_device, file);
+ if (ret < 0) {
+ dump_type->remove_device(dump_device);
+ module_put(dump_type->owner);
+ list_del(&dump_device->list);
+ kfree(dump_device);
+ }
+
+ return ret;
+}
+
+/*
+ * Remove dump partition corresponding to file.
+ * Must be called with disk_dump_lock held.
+ */
+static int remove_dump(struct file *file)
+{
+ struct disk_dump_device *dump_device;
+ struct disk_dump_partition *dump_part;
+ struct disk_dump_type *dump_type;
+ dev_t dev = file->f_dentry->d_inode->i_rdev;
+
+ if (!(dump_part = find_dump_partition(dev)))
+ return -ENOENT;
+
+ dump_device = dump_part->device;
+
+ list_del(&dump_part->list);
+ mntput(dump_part->vfsmount);
+ dput(dump_part->dentry);
+ kfree(dump_part);
+
+ if (list_empty(&dump_device->partitions)) {
+ dump_type = dump_device->dump_type;
+ dump_type->remove_device(dump_device);
+ module_put(dump_type->owner);
+ module_put(dump_type->owner);
+ list_del(&dump_device->list);
+ kfree(dump_device);
+ }
+
+ return 0;
+}
+
+#ifdef CONFIG_PROC_FS
+static int proc_ioctl(struct inode *inode, struct file *file, unsigned int cmd, unsigned long param)
+{
+ int fd = (int)param;
+ int ret;
+ struct file *dump_file;
+ struct inode *dump_inode;
+
+ if (!capable(CAP_SYS_ADMIN))
+ return -EPERM;
+ dump_file = fget(fd);
+ if (!dump_file)
+ return -EBADF;
+ dump_inode = dump_file->f_dentry->d_inode;
+ if (!dump_inode->i_bdev) {
+ fput(dump_file);
+ return -EBADF;
+ }
+
+ spin_lock(&disk_dump_lock);
+ switch (cmd) {
+ case BLKADDDUMPDEVICE:
+ ret = add_dump(dump_file);
+ break;
+ case BLKREMOVEDUMPDEVICE:
+ ret = remove_dump(dump_file);
+ break;
+ default:
+ ret = -EINVAL;
+ }
+
+ set_crc_modules();
+ spin_unlock(&disk_dump_lock);
+
+ fput(dump_file);
+
+ return ret;
+}
+
+static void *disk_dump_seq_start(struct seq_file *seq, loff_t *pos)
+{
+ struct disk_dump_device *dump_device;
+ struct disk_dump_partition *dump_part;
+ loff_t n = *pos;
+
+ spin_lock(&disk_dump_lock);
+ list_for_each_entry(dump_device, &disk_dump_devices, list) {
+ seq->private = dump_device;
+ list_for_each_entry(dump_part, &dump_device->partitions, list) {
+ if (!n--)
+ return dump_part;
+ }
+ }
+ return NULL;
+}
+
+static void *disk_dump_seq_next(struct seq_file *seq, void *v, loff_t *pos)
+{
+ struct list_head *partition = v;
+ struct list_head *device = seq->private;
+ struct disk_dump_device *dump_device;
+
+ dump_device = list_entry(device, struct disk_dump_device, list);
+
+ (*pos)++;
+ partition = partition->next;
+ if (partition != &dump_device->partitions)
+ return partition;
+
+ device = device->next;
+ seq->private = device;
+ if (device == &disk_dump_devices)
+ return NULL;
+
+ dump_device = list_entry(device, struct disk_dump_device, list);
+
+ return dump_device->partitions.next;
+}
+
+static void disk_dump_seq_stop(struct seq_file *seq, void *v)
+{
+ spin_unlock(&disk_dump_lock);
+}
+
+static int disk_dump_seq_show(struct seq_file *seq, void *v)
+{
+ struct disk_dump_partition *dump_part = v;
+ char *page;
+ char *path;
+
+ if (!(page = (char *)__get_free_page(GFP_KERNEL)))
+ return -ENOMEM;
+ path = d_path(dump_part->dentry, dump_part->vfsmount, page, PAGE_SIZE);
+ seq_printf(seq, "%s %lu %lu\n",
+ path, dump_part->start_sect, dump_part->nr_sects);
+ free_page((unsigned long)page);
+ return 0;
+}
+
+static struct seq_operations disk_dump_seq_ops = {
+ .start = disk_dump_seq_start,
+ .next = disk_dump_seq_next,
+ .stop = disk_dump_seq_stop,
+ .show = disk_dump_seq_show,
+};
+
+static int disk_dump_open(struct inode *inode, struct file *file)
+{
+ return seq_open(file, &disk_dump_seq_ops);
+}
+
+static struct file_operations disk_dump_fops = {
+ .owner = THIS_MODULE,
+ .open = disk_dump_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = seq_release,
+ .ioctl = proc_ioctl,
+};
+#endif
+
+
+int register_disk_dump_type(struct disk_dump_type *dump_type)
+{
+ spin_lock(&disk_dump_lock);
+ list_add(&dump_type->list, &disk_dump_types);
+ set_crc_modules();
+ spin_unlock(&disk_dump_lock);
+
+ return 0;
+}
+
+int unregister_disk_dump_type(struct disk_dump_type *dump_type)
+{
+ spin_lock(&disk_dump_lock);
+ list_del(&dump_type->list);
+ set_crc_modules();
+ spin_unlock(&disk_dump_lock);
+
+ return 0;
+}
+
+EXPORT_SYMBOL(register_disk_dump_type);
+EXPORT_SYMBOL(unregister_disk_dump_type);
+EXPORT_SYMBOL(disk_dump_notifier_list);
+
+
+static void compute_total_blocks(void)
+{
+ unsigned int nr;
+
+ /*
+ * the number of block of the common header and the header
+ * that is depend on the architecture
+ *
+ * block 0: dump partition header
+ * block 1: dump header
+ * block 2: dump subheader
+ * block 3..n: memory bitmap
+ * block (n + 1)...: saved memory
+ *
+ * We never overwrite block 0
+ */
+ header_blocks = 2 + size_of_sub_header();
+
+ total_ram_blocks = 0;
+ for (nr = 0; nr < max_pfn; nr++) {
+ if (page_is_ram(nr))
+ total_ram_blocks++;
+ }
+
+ bitmap_blocks = ROUNDUP(max_pfn, 8 * PAGE_SIZE);
+
+ /*
+ * The necessary size of area for dump is:
+ * 1 block for common header
+ * m blocks for architecture dependent header
+ * n blocks for memory bitmap
+ * and whole memory
+ */
+ total_blocks = header_blocks + bitmap_blocks + total_ram_blocks;
+
+ Info("total blocks required: %u (header %u + bitmap %u + memory %u)",
+ total_blocks, header_blocks, bitmap_blocks, total_ram_blocks);
+}
+
+static int init_diskdump(void)
+{
+ unsigned long long t0;
+ unsigned long long t1;
+ struct page *page;
+
+ if (!platform_supports_diskdump) {
+ Err("platform does not support diskdump.");
+ return -1;
+ }
+
+ /* Allocate one block that is used temporally */
+ do {
+ page = alloc_pages(GFP_KERNEL, block_order);
+ if (page != NULL)
+ break;
+ } while (--block_order >= 0);
+ if (!page) {
+ Err("alloc_pages failed.");
+ return -1;
+ }
+ scratch = page_address(page);
+ Info("Maximum block size: %lu", PAGE_SIZE << block_order);
+
+ if (diskdump_register_hook(disk_dump)) {
+ Err("failed to register hooks.");
+ return -1;
+ }
+
+ compute_total_blocks();
+
+ platform_timestamp(t0);
+ mdelay(1);
+ platform_timestamp(t1);
+ timestamp_1sec = (unsigned long)(t1 - t0) * 1000;
+
+#ifdef CONFIG_PROC_FS
+ {
+ struct proc_dir_entry *p;
+
+ p = create_proc_entry("diskdump", S_IRUGO|S_IWUSR, NULL);
+ if (p)
+ p->proc_fops = &disk_dump_fops;
+ }
+#endif
+
+ return 0;
+}
+
+static void cleanup_diskdump(void)
+{
+ Info("shut down.");
+ diskdump_unregister_hook();
+ free_pages((unsigned long)scratch, block_order);
+#ifdef CONFIG_PROC_FS
+ remove_proc_entry("diskdump", NULL);
+#endif
+}
+
+module_init(init_diskdump);
+module_exit(cleanup_diskdump);
+MODULE_LICENSE("GPL");
diff -Nur linux-2.6.6.org/drivers/block/diskdumplib.c linux-2.6.6/drivers/block/diskdumplib.c
--- linux-2.6.6.org/drivers/block/diskdumplib.c 1970-01-01 09:00:00.000000000 +0900
+++ linux-2.6.6/drivers/block/diskdumplib.c 2004-06-16 19:34:16.000000000 +0900
@@ -0,0 +1,203 @@
+/*
+ * linux/drivers/block/diskdumplib.c
+ *
+ * Copyright (C) 2004 FUJITSU LIMITED
+ * Written by Nobuhiro Tachino (ntachino@jp.fujitsu.com)
+ *
+ */
+/*
+ * 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, 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, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ */
+#include <linux/mm.h>
+#include <linux/init.h>
+#include <linux/delay.h>
+#include <linux/file.h>
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/utsname.h>
+#include <linux/smp_lock.h>
+#include <linux/genhd.h>
+#include <linux/slab.h>
+#include <linux/blkdev.h>
+#include <linux/diskdump.h>
+#include <linux/diskdumplib.h>
+#include <asm/diskdump.h>
+
+/*
+ * timer list and tasklet_struct holder
+ */
+unsigned long volatile diskdump_base_jiffies;
+static unsigned long long timestamp_base;
+static unsigned long timestamp_hz;
+
+#define DISKDUMP_NUM_TASKLETS 8
+
+/*
+ * We can't use next field of tasklet because it breaks the original
+ * tasklets chain and we have no way to know which chain the tasklet is
+ * linked.
+ */
+static struct tasklet_struct *diskdump_tasklets[DISKDUMP_NUM_TASKLETS];
+
+static LIST_HEAD(diskdump_timers);
+static LIST_HEAD(diskdump_workq);
+
+
+static int store_tasklet(struct tasklet_struct *tasklet)
+{
+ int i;
+
+ for (i = 0; i < DISKDUMP_NUM_TASKLETS; i++)
+ if (diskdump_tasklets[i] == NULL) {
+ diskdump_tasklets[i] = tasklet;
+ return 0;
+ }
+ return -1;
+}
+
+static struct tasklet_struct *find_tasklet(struct tasklet_struct *tasklet)
+{
+ int i;
+
+ for (i = 0; i < DISKDUMP_NUM_TASKLETS; i++)
+ if (diskdump_tasklets[i] == tasklet)
+ return diskdump_tasklets[i];
+ return NULL;
+}
+
+void _diskdump_tasklet_schedule(struct tasklet_struct *tasklet)
+{
+ if (!find_tasklet(tasklet))
+ if (store_tasklet(tasklet))
+ printk(KERN_ERR "diskdumplib: too many tasklet. Ignored\n");
+ set_bit(TASKLET_STATE_SCHED, &tasklet->state);
+}
+
+int _diskdump_schedule_work(struct work_struct *work)
+{
+ list_add_tail(&work->entry, &diskdump_workq);
+ return 1;
+}
+
+void _diskdump_add_timer(struct timer_list *timer)
+{
+ timer->base = (void *)1;
+ list_add(&timer->entry, &diskdump_timers);
+}
+
+int _diskdump_del_timer(struct timer_list *timer)
+{
+ if (timer->base != NULL) {
+ list_del(&timer->entry);
+ return 1;
+ } else {
+ timer->base = NULL;
+ return 0;
+ }
+}
+
+int _diskdump_mod_timer(struct timer_list *timer, unsigned long expires)
+{
+ int ret;
+
+ ret = _diskdump_del_timer(timer);
+ timer->expires = expires;
+ _diskdump_add_timer(timer);
+
+ return ret;
+}
+
+static void update_jiffies(void)
+{
+ unsigned long long t;
+
+ platform_timestamp(t);
+ while (t > timestamp_base + timestamp_hz) {
+ timestamp_base += timestamp_hz;
+ jiffies++;
+ platform_timestamp(t);
+ }
+}
+
+void diskdump_update(void)
+{
+ struct tasklet_struct *tasklet;
+ struct work_struct *work;
+ struct timer_list *timer;
+ struct list_head *t, *n, head;
+ int i;
+
+ update_jiffies();
+
+ /* run timers */
+ list_for_each_safe(t, n, &diskdump_timers) {
+ timer = list_entry(t, struct timer_list, entry);
+ if (time_before_eq(timer->expires, jiffies)) {
+ list_del(t);
+ timer->function(timer->data);
+ }
+ }
+
+ /* run tasklet */
+ for (i = 0; i < DISKDUMP_NUM_TASKLETS; i++)
+ if ((tasklet = diskdump_tasklets[i]))
+ if (!atomic_read(&tasklet->count))
+ if (test_and_clear_bit(TASKLET_STATE_SCHED, &tasklet->state))
+ tasklet->func(tasklet->data);
+
+ /* run work queue */
+ list_add(&head, &diskdump_workq);
+ list_del_init(&diskdump_workq);
+ n = head.next;
+ while (n != &head) {
+ work = list_entry(t, struct work_struct, entry);
+ n = n->next;
+ if (work->func)
+ work->func(work->wq_data);
+ }
+}
+
+void diskdump_lib_init(void)
+{
+ unsigned long long t;
+
+ /* Save original jiffies value */
+ diskdump_base_jiffies = jiffies;
+
+ platform_timestamp(timestamp_base);
+ udelay(1000000/HZ);
+ platform_timestamp(t);
+ timestamp_hz = (unsigned long)(t - timestamp_base);
+
+ diskdump_update();
+}
+
+void diskdump_lib_exit(void)
+{
+ /* Resotre original jiffies. */
+ jiffies = diskdump_base_jiffies;
+}
+
+EXPORT_SYMBOL(diskdump_lib_init);
+EXPORT_SYMBOL(diskdump_lib_exit);
+EXPORT_SYMBOL(diskdump_update);
+EXPORT_SYMBOL(_diskdump_add_timer);
+EXPORT_SYMBOL(_diskdump_del_timer);
+EXPORT_SYMBOL(_diskdump_mod_timer);
+EXPORT_SYMBOL(_diskdump_tasklet_schedule);
+EXPORT_SYMBOL(_diskdump_schedule_work);
+
+MODULE_LICENSE("GPL");
diff -Nur linux-2.6.6.org/drivers/block/genhd.c linux-2.6.6/drivers/block/genhd.c
--- linux-2.6.6.org/drivers/block/genhd.c 2004-06-04 21:22:15.000000000 +0900
+++ linux-2.6.6/drivers/block/genhd.c 2004-06-16 19:34:16.000000000 +0900
@@ -629,3 +629,28 @@
}
EXPORT_SYMBOL(invalidate_partition);
+
+/*
+ * Dump stuff.
+ */
+void (*diskdump_func) (struct pt_regs *regs, void *platform_arg) = NULL;
+EXPORT_SYMBOL_GPL(diskdump_func);
+
+int diskdump_register_hook(void (*dump_func) (struct pt_regs *, void *))
+{
+ if (diskdump_func)
+ return -EEXIST;
+
+ diskdump_func = dump_func;
+
+ return 0;
+}
+
+EXPORT_SYMBOL_GPL(diskdump_register_hook);
+
+void diskdump_unregister_hook(void)
+{
+ diskdump_func = NULL;
+}
+
+EXPORT_SYMBOL_GPL(diskdump_unregister_hook);
diff -Nur linux-2.6.6.org/drivers/char/sysrq.c linux-2.6.6/drivers/char/sysrq.c
--- linux-2.6.6.org/drivers/char/sysrq.c 2004-06-04 21:22:29.000000000 +0900
+++ linux-2.6.6/drivers/char/sysrq.c 2004-06-16 19:34:16.000000000 +0900
@@ -107,6 +107,19 @@
.action_msg = "Resetting",
};
+/* crash sysrq handler */
+static void sysrq_handle_crash(int key, struct pt_regs *pt_regs,
+ struct tty_struct *tty)
+{
+ *( (char *) 0) = 0;
+}
+
+static struct sysrq_key_op sysrq_crash_op = {
+ .handler = sysrq_handle_crash,
+ .help_msg = "Crash",
+ .action_msg = "Crashing the kernel by request",
+};
+
static void sysrq_handle_sync(int key, struct pt_regs *pt_regs,
struct tty_struct *tty)
{
@@ -235,7 +248,7 @@
it is handled specially on the sparc
and will never arrive */
/* b */ &sysrq_reboot_op,
-/* c */ NULL,
+/* c */ &sysrq_crash_op,
/* d */ NULL,
/* e */ &sysrq_term_op,
/* f */ NULL,
diff -Nur linux-2.6.6.org/include/asm-i386/diskdump.h linux-2.6.6/include/asm-i386/diskdump.h
--- linux-2.6.6.org/include/asm-i386/diskdump.h 1970-01-01 09:00:00.000000000 +0900
+++ linux-2.6.6/include/asm-i386/diskdump.h 2004-06-16 19:34:16.000000000 +0900
@@ -0,0 +1,82 @@
+#ifndef _ASM_I386_DISKDUMP_H
+#define _ASM_I386_DISKDUMP_H
+
+/*
+ * linux/include/asm-i386/diskdump.h
+ *
+ * Copyright (c) 2004 FUJITSU LIMITED
+ * Copyright (c) 2003 Red Hat, Inc. All rights reserved.
+ */
+/*
+ * 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, 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, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ */
+
+#ifdef __KERNEL__
+
+#include <linux/elf.h>
+
+extern int page_is_ram(unsigned long);
+const static int platform_supports_diskdump = 1;
+
+#define platform_fix_regs() \
+{ \
+ unsigned long esp; \
+ unsigned short ss; \
+ esp = (unsigned long) ((char *)regs + sizeof (struct pt_regs)); \
+ ss = __KERNEL_DS; \
+ if (regs->xcs & 3) { \
+ esp = regs->esp; \
+ ss = regs->xss & 0xffff; \
+ } \
+ myregs = *regs; \
+ myregs.esp = esp; \
+ myregs.xss = (myregs.xss & 0xffff0000) | ss; \
+}
+
+struct disk_dump_sub_header {
+ elf_gregset_t elf_regs;
+};
+
+#define platform_timestamp(x) rdtscll(x)
+
+#define size_of_sub_header() ((sizeof(struct disk_dump_sub_header) + PAGE_SIZE - 1) / DUMP_BLOCK_SIZE)
+
+#define write_sub_header() \
+({ \
+ int ret; \
+ \
+ ELF_CORE_COPY_REGS(dump_sub_header.elf_regs, (&myregs)); \
+ clear_page(scratch); \
+ memcpy(scratch, &dump_sub_header, sizeof(dump_sub_header)); \
+ \
+ if ((ret = write_blocks(dump_part, 2, scratch, 1)) >= 0) \
+ ret = 1; /* size of sub header in page */; \
+ ret; \
+})
+
+#define platform_freeze_cpu() \
+{ \
+ for (;;) local_irq_disable(); \
+}
+
+#define platform_start_diskdump(func, regs) \
+{ \
+ func(regs, NULL); \
+}
+
+#endif /* __KERNEL__ */
+
+#endif /* _ASM_I386_DISKDUMP_H */
diff -Nur linux-2.6.6.org/include/asm-i386/kmap_types.h linux-2.6.6/include/asm-i386/kmap_types.h
--- linux-2.6.6.org/include/asm-i386/kmap_types.h 2004-06-04 21:22:02.000000000 +0900
+++ linux-2.6.6/include/asm-i386/kmap_types.h 2004-06-16 19:34:16.000000000 +0900
@@ -24,7 +24,8 @@
D(11) KM_IRQ1,
D(12) KM_SOFTIRQ0,
D(13) KM_SOFTIRQ1,
-D(14) KM_TYPE_NR
+D(14) KM_DISKDUMP,
+D(15) KM_TYPE_NR
};
#undef D
diff -Nur linux-2.6.6.org/include/linux/diskdump.h linux-2.6.6/include/linux/diskdump.h
--- linux-2.6.6.org/include/linux/diskdump.h 1970-01-01 09:00:00.000000000 +0900
+++ linux-2.6.6/include/linux/diskdump.h 2004-06-16 19:34:16.000000000 +0900
@@ -0,0 +1,157 @@
+#ifndef _LINUX_DISKDUMP_H
+#define _LINUX_DISKDUMP_H
+
+/*
+ * linux/include/linux/diskdump.h
+ *
+ * Copyright (c) 2004 FUJITSU LIMITED
+ *
+ * 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, 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, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ */
+
+#include <linux/list.h>
+#include <linux/mount.h>
+#include <linux/dcache.h>
+#include <linux/blkdev.h>
+#include <linux/utsname.h>
+
+/* The minimum Dump I/O unit. Must be the same of PAGE_SIZE */
+#define DUMP_BLOCK_SIZE PAGE_SIZE
+#define DUMP_BLOCK_SHIFT PAGE_SHIFT
+
+/* Dump ioctls */
+#define BLKADDDUMPDEVICE 0xdf00 /* Add a dump device */
+#define BLKREMOVEDUMPDEVICE 0xdf01 /* Delete a dump device */
+
+int diskdump_register_hook(void (*dump_func)(struct pt_regs *, void *));
+void diskdump_unregister_hook(void);
+
+/*
+ * The handler that adapter driver provides for the common module of
+ * dump
+ */
+struct disk_dump_partition;
+struct disk_dump_device;
+
+struct disk_dump_type {
+ void *(*probe)(struct block_device *);
+ int (*add_device)(struct disk_dump_device *);
+ void (*remove_device)(struct disk_dump_device *);
+ struct module *owner;
+ struct list_head list;
+};
+
+struct disk_dump_device_ops {
+ int (*sanity_check)(struct disk_dump_device *);
+ int (*quiesce)(struct disk_dump_device *);
+ int (*shutdown)(struct disk_dump_device *);
+ int (*rw_block)(struct disk_dump_partition *, int rw, unsigned long block_nr, void *buf, int len);
+};
+
+/* The data structure for a dump device */
+struct disk_dump_device {
+ struct list_head list;
+ struct disk_dump_device_ops ops;
+ struct disk_dump_type *dump_type;
+ void *device;
+ unsigned int max_blocks;
+ struct list_head partitions;
+};
+
+/* The data structure for a dump partition */
+struct disk_dump_partition {
+ struct list_head list;
+ struct disk_dump_device *device;
+ struct vfsmount *vfsmount;
+ struct dentry *dentry;
+ unsigned long start_sect;
+ unsigned long nr_sects;
+};
+
+
+int register_disk_dump_type(struct disk_dump_type *);
+int unregister_disk_dump_type(struct disk_dump_type *);
+
+
+/*
+ * Architecture-independent dump header
+ */
+
+/* The signature which is written in each block in the dump partition */
+#define DISK_DUMP_SIGNATURE "DISKDUMP"
+#define DISK_DUMP_HEADER_VERSION 1
+
+#define DUMP_PARTITION_SIGNATURE "diskdump"
+
+#define DUMP_HEADER_COMPLETED 0
+#define DUMP_HEADER_INCOMPLETED 1
+
+struct disk_dump_header {
+ char signature[8]; /* = "DISKDUMP" */
+ int header_version; /* Dump header version */
+ struct new_utsname utsname; /* copy of system_utsname */
+ struct timespec timestamp; /* Time stamp */
+ unsigned int status; /* Above flags */
+ int block_size; /* Size of a block in byte */
+ int sub_hdr_size; /* Size of arch dependent
+ header in blocks */
+ unsigned int bitmap_blocks; /* Size of Memory bitmap in
+ block */
+ unsigned int max_mapnr; /* = max_mapnr */
+ unsigned int total_ram_blocks;/* Size of Memory in block */
+ unsigned int device_blocks; /* Number of total blocks in
+ * the dump device */
+ unsigned int written_blocks; /* Number of written blocks */
+ unsigned int current_cpu; /* CPU# which handles dump */
+ int nr_cpus; /* Number of CPUs */
+ struct task_struct *tasks[NR_CPUS];
+};
+
+/*
+ * Calculate the check sum of the whole module
+ */
+#define get_crc_module() \
+({ \
+ struct module *module = &__this_module; \
+ crc32_le(0, (char *)(module->module_core), \
+ ((unsigned long)module - (unsigned long)(module->module_core))); \
+})
+
+/* Calculate the checksum of the whole module */
+#define set_crc_modules() \
+({ \
+ module_crc = 0; \
+ module_crc = get_crc_module(); \
+})
+
+/*
+ * Compare the checksum value that is stored in module_crc to the check
+ * sum of current whole module. Must be called with holding disk_dump_lock.
+ * Return TRUE if they are the same, else return FALSE
+ *
+ */
+#define check_crc_module() \
+({ \
+ uint32_t orig_crc, cur_crc; \
+ \
+ orig_crc = module_crc; module_crc = 0; \
+ cur_crc = get_crc_module(); \
+ module_crc = orig_crc; \
+ orig_crc == cur_crc; \
+})
+
+
+#endif /* _LINUX_DISKDUMP_H */
diff -Nur linux-2.6.6.org/include/linux/diskdumplib.h linux-2.6.6/include/linux/diskdumplib.h
--- linux-2.6.6.org/include/linux/diskdumplib.h 1970-01-01 09:00:00.000000000 +0900
+++ linux-2.6.6/include/linux/diskdumplib.h 2004-06-16 19:34:16.000000000 +0900
@@ -0,0 +1,67 @@
+#ifndef _LINUX_DISKDUMPLIB_H
+#define _LINUX_DISKDUMPLIB_H
+
+#include <linux/interrupt.h>
+#include <linux/timer.h>
+
+void diskdump_lib_init(void);
+void diskdump_lib_exit(void);
+void diskdump_update(void);
+
+void _diskdump_add_timer(struct timer_list *);
+int _diskdump_del_timer(struct timer_list *);
+int _diskdump_mod_timer(struct timer_list *, unsigned long);
+void _diskdump_tasklet_schedule(struct tasklet_struct *);
+int _diskdump_schedule_work(struct work_struct *);
+
+static inline void diskdump_add_timer(struct timer_list *timer)
+{
+ if (crashdump_mode())
+ _diskdump_add_timer(timer);
+ else
+ add_timer(timer);
+}
+
+static inline int diskdump_del_timer(struct timer_list *timer)
+{
+ if (crashdump_mode())
+ return _diskdump_del_timer(timer);
+ else
+ return del_timer(timer);
+}
+
+static inline int diskdump_mod_timer(struct timer_list *timer, unsigned long expires)
+{
+ if (crashdump_mode())
+ return _diskdump_mod_timer(timer, expires);
+ else
+ return mod_timer(timer, expires);
+}
+
+static inline void diskdump_tasklet_schedule(struct tasklet_struct *tasklet)
+{
+ if (crashdump_mode())
+ return _diskdump_tasklet_schedule(tasklet);
+ else
+ return tasklet_schedule(tasklet);
+}
+
+static inline int diskdump_schedule_work(struct work_struct *work)
+{
+ if (crashdump_mode())
+ return _diskdump_schedule_work(work);
+ else
+ return schedule_work(work);
+}
+
+static inline long diskdump_schedule_timeout(signed long timeout)
+{
+ if (crashdump_mode()) {
+ mdelay(timeout);
+ set_current_state(TASK_RUNNING);
+ return timeout;
+ } else
+ return schedule_timeout(timeout);
+}
+
+#endif /* _LINUX_DISKDUMPLIB_H */
diff -Nur linux-2.6.6.org/include/linux/kernel.h linux-2.6.6/include/linux/kernel.h
--- linux-2.6.6.org/include/linux/kernel.h 2004-06-04 21:22:08.000000000 +0900
+++ linux-2.6.6/include/linux/kernel.h 2004-06-16 19:34:16.000000000 +0900
@@ -112,6 +112,11 @@
extern int system_state; /* See values below */
extern int tainted;
extern const char *print_tainted(void);
+struct pt_regs;
+extern void try_crashdump(struct pt_regs *);
+extern void (*diskdump_func) (struct pt_regs *regs, void *platform_arg);
+extern int diskdump_mode;
+#define crashdump_mode() (diskdump_mode)
/* Values used for system_state */
#define SYSTEM_BOOTING 0
@@ -136,6 +141,12 @@
#define pr_info(fmt,arg...) \
printk(KERN_INFO fmt,##arg)
+#define pr_err(fmt,arg...) \
+ printk(KERN_ERR fmt,##arg)
+
+#define pr_warn(fmt,arg...) \
+ printk(KERN_WARNING fmt,##arg)
+
/*
* Display an IP address in readable format.
*/
diff -Nur linux-2.6.6.org/kernel/panic.c linux-2.6.6/kernel/panic.c
--- linux-2.6.6.org/kernel/panic.c 2004-06-04 21:21:58.000000000 +0900
+++ linux-2.6.6/kernel/panic.c 2004-06-16 19:34:16.000000000 +0900
@@ -23,8 +23,10 @@
int panic_timeout;
int panic_on_oops;
int tainted;
+int diskdump_mode = 0;
EXPORT_SYMBOL(panic_timeout);
+EXPORT_SYMBOL_GPL(diskdump_mode);
struct notifier_block *panic_notifier_list;
@@ -60,6 +62,8 @@
vsnprintf(buf, sizeof(buf), fmt, args);
va_end(args);
printk(KERN_EMERG "Kernel panic: %s\n",buf);
+ if (diskdump_func)
+ BUG();
if (in_interrupt())
printk(KERN_EMERG "In interrupt handler - not syncing\n");
else if (!current->pid)
@@ -134,3 +138,23 @@
snprintf(buf, sizeof(buf), "Not tainted");
return(buf);
}
+
+/*
+ * Try crashdump. Diskdump is first, netdump is second.
+ * We clear diskdump_func before call of diskdump_func, so
+ * If double panic would occur in diskdump, netdump can handle
+ * it.
+ */
+#include <asm/diskdump.h>
+void try_crashdump(struct pt_regs *regs)
+{
+ void (*func)(struct pt_regs *, void *);
+
+ if (diskdump_func) {
+ func = diskdump_func;
+ diskdump_func = NULL;
+ platform_start_diskdump(func, regs);
+ }
+ if (panic_on_oops)
+ panic("Fatal exception");
+}
diff -Nur linux-2.6.6.org/mm/bootmem.c linux-2.6.6/mm/bootmem.c
--- linux-2.6.6.org/mm/bootmem.c 2004-06-04 21:22:37.000000000 +0900
+++ linux-2.6.6/mm/bootmem.c 2004-06-16 19:34:16.000000000 +0900
@@ -16,6 +16,7 @@
#include <linux/init.h>
#include <linux/bootmem.h>
#include <linux/mmzone.h>
+#include <linux/module.h>
#include <asm/dma.h>
#include <asm/io.h>
@@ -26,6 +27,7 @@
unsigned long max_low_pfn;
unsigned long min_low_pfn;
unsigned long max_pfn;
+EXPORT_SYMBOL(max_pfn);
/* return the number of _pages_ that will be allocated for the boot bitmap */
unsigned long __init bootmem_bootmap_pages (unsigned long pages)
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH 2/4][Diskdump]Update patches
2004-06-16 12:39 [PATCH 0/4][Diskdump]Update patches Takao Indoh
2004-06-16 12:45 ` [PATCH 1/4][Diskdump]Update patches Takao Indoh
@ 2004-06-16 12:46 ` Takao Indoh
2004-06-16 12:48 ` [PATCH 3/4][Diskdump]Update patches Takao Indoh
` (2 subsequent siblings)
4 siblings, 0 replies; 9+ messages in thread
From: Takao Indoh @ 2004-06-16 12:46 UTC (permalink / raw)
To: linux-kernel
This is a patch for scsi common layer.
diff -Nur linux-2.6.6.org/drivers/scsi/Kconfig linux-2.6.6/drivers/scsi/Kconfig
--- linux-2.6.6.org/drivers/scsi/Kconfig 2004-06-04 21:22:20.000000000 +0900
+++ linux-2.6.6/drivers/scsi/Kconfig 2004-06-16 19:34:16.000000000 +0900
@@ -55,6 +55,12 @@
In this case, do not compile the driver for your SCSI host adapter
(below) as a module either.
+config SCSI_DUMP
+ tristate "SCSI dump support"
+ depends on DISKDUMP && SCSI
+ help
+ SCSI dump support
+
config CHR_DEV_ST
tristate "SCSI tape support"
depends on SCSI
diff -Nur linux-2.6.6.org/drivers/scsi/Makefile linux-2.6.6/drivers/scsi/Makefile
--- linux-2.6.6.org/drivers/scsi/Makefile 2004-06-04 21:22:21.000000000 +0900
+++ linux-2.6.6/drivers/scsi/Makefile 2004-06-16 19:34:16.000000000 +0900
@@ -133,6 +133,8 @@
obj-$(CONFIG_BLK_DEV_SR) += sr_mod.o
obj-$(CONFIG_CHR_DEV_SG) += sg.o
+obj-$(CONFIG_SCSI_DUMP) += scsi_dump.o
+
scsi_mod-y += scsi.o hosts.o scsi_ioctl.o constants.o \
scsicam.o scsi_error.o scsi_lib.o \
scsi_scan.o scsi_syms.o scsi_sysfs.o \
diff -Nur linux-2.6.6.org/drivers/scsi/scsi.c linux-2.6.6/drivers/scsi/scsi.c
--- linux-2.6.6.org/drivers/scsi/scsi.c 2004-06-04 21:22:20.000000000 +0900
+++ linux-2.6.6/drivers/scsi/scsi.c 2004-06-16 19:34:16.000000000 +0900
@@ -691,6 +691,9 @@
{
unsigned long flags;
+ if (crashdump_mode())
+ return;
+
/*
* We don't have to worry about this one timing out any more.
* If we are unable to remove the timer, then the command
diff -Nur linux-2.6.6.org/drivers/scsi/scsi_dump.c linux-2.6.6/drivers/scsi/scsi_dump.c
--- linux-2.6.6.org/drivers/scsi/scsi_dump.c 1970-01-01 09:00:00.000000000 +0900
+++ linux-2.6.6/drivers/scsi/scsi_dump.c 2004-06-16 19:34:16.000000000 +0900
@@ -0,0 +1,579 @@
+/*
+ * linux/drivers/scsi/scsi_dump.c
+ *
+ * Copyright (C) 2004 FUJITSU LIMITED
+ * Written by Nobuhiro Tachino (ntachino@jp.fujitsu.com)
+ *
+ * Some codes are derived from drivers/scsi/sd.c
+ */
+/*
+ * 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, 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, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ */
+
+#include <linux/module.h>
+
+#include <linux/fs.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+
+#include <linux/blkdev.h>
+#include <linux/blkpg.h>
+
+#include <linux/genhd.h>
+#include <linux/utsname.h>
+#include <linux/crc32.h>
+#include <linux/delay.h>
+#include <linux/diskdump.h>
+#include <linux/diskdumplib.h>
+
+#include <scsi/scsi.h>
+#include <scsi/scsi_ioctl.h>
+#include <scsi/scsi_host.h>
+#include <scsi/scsi_cmnd.h>
+#include <scsi/scsi_device.h>
+#include <scsi/scsi_dump.h>
+
+#include "scsi.h"
+#include "scsi_priv.h"
+
+#define MAX_RETRIES 5
+#define SD_TIMEOUT (60 * HZ)
+
+#define Dbg(x, ...) pr_debug("scsi_dump: " x "\n", ## __VA_ARGS__)
+#define Err(x, ...) pr_err ("scsi_dump: " x "\n", ## __VA_ARGS__)
+#define Warn(x, ...) pr_warn ("scsi_dump: " x "\n", ## __VA_ARGS__)
+#define Info(x, ...) pr_info ("scsi_dump: " x "\n", ## __VA_ARGS__)
+
+/* blocks to 512byte sectors */
+#define BLOCK_SECTOR(s) ((s) << (DUMP_BLOCK_SHIFT - 9))
+
+static int quiesce_ok = 0;
+static struct scsi_cmnd scsi_dump_cmnd;
+static struct request scsi_dump_req;
+static uint32_t module_crc;
+
+static void rw_intr(struct scsi_cmnd * scmd)
+{
+ scmd->done = NULL;
+}
+
+/*
+ * Common code to make Scsi_Cmnd
+ */
+static void init_scsi_command(struct scsi_device *sdev, struct scsi_cmnd *scmd,
+ void *buf, int len, unsigned char direction,
+ int set_lun)
+{
+ scmd->request = &scsi_dump_req;
+ scmd->sc_magic = SCSI_CMND_MAGIC;
+ scmd->owner = SCSI_OWNER_MIDLEVEL;
+ scmd->device = sdev;
+ scmd->buffer = scmd->request_buffer = buf;
+ scmd->bufflen = scmd->request_bufflen = len;
+
+
+ scmd->sc_data_direction = direction;
+
+ memcpy(scmd->data_cmnd, scmd->cmnd, sizeof(scmd->cmnd));
+ scmd->cmd_len = COMMAND_SIZE(scmd->cmnd[0]);
+ scmd->old_cmd_len = scmd->cmd_len;
+
+
+ if (set_lun)
+ scmd->cmnd[1] |= (sdev->scsi_level <= SCSI_2) ?
+ ((sdev->lun << 5) & 0xe0) : 0;
+
+ scmd->transfersize = sdev->sector_size;
+ if (direction == SCSI_DATA_WRITE)
+ scmd->underflow = len;
+
+ scmd->allowed = MAX_RETRIES;
+ scmd->timeout_per_command = SD_TIMEOUT;
+
+ /*
+ * This is the completion routine we use. This is matched in terms
+ * of capability to this function.
+ */
+ scmd->done = rw_intr;
+}
+
+/* MODE SENSE */
+static void init_mode_sense_command(struct scsi_device *sdev,
+ struct scsi_cmnd *scmd, void *buf)
+{
+ memset(scmd, 0, sizeof(*scmd));
+ scmd->cmnd[0] = MODE_SENSE;
+ scmd->cmnd[1] = 0x00; /* DBD=0 */
+ scmd->cmnd[2] = 0x08; /* PCF=0 Page 8(Cache) */
+ scmd->cmnd[4] = 255;
+
+ init_scsi_command(sdev, scmd, buf, 256, SCSI_DATA_READ, 1);
+}
+
+/* MODE SELECT */
+static void init_mode_select_command(struct scsi_device *sdev,
+ struct scsi_cmnd *scmd, void *buf, int len)
+{
+ memset(scmd, 0, sizeof(*scmd));
+ scmd->cmnd[0] = MODE_SELECT;
+ scmd->cmnd[1] = 0x10; /* PF=1 SP=0 */
+ scmd->cmnd[4] = len;
+
+ init_scsi_command(sdev, scmd, buf, len, SCSI_DATA_WRITE, 1);
+}
+
+/* SYNCHRONIZE CACHE */
+static void init_sync_command(struct scsi_device *sdev, struct scsi_cmnd * scmd)
+{
+ memset(scmd, 0, sizeof(*scmd));
+ scmd->cmnd[0] = SYNCHRONIZE_CACHE;
+
+ init_scsi_command(sdev, scmd, NULL, 0, SCSI_DATA_NONE, 0);
+}
+
+/* REQUEST SENSE */
+static void init_sense_command(struct scsi_device *sdev, struct scsi_cmnd *scmd,
+ void *buf)
+{
+ memset(scmd, 0, sizeof(*scmd));
+ scmd->cmnd[0] = REQUEST_SENSE;
+ scmd->cmnd[4] = 255;
+
+ init_scsi_command(sdev, scmd, buf, 256, SCSI_DATA_READ, 1);
+}
+
+/* READ/WRITE */
+static int init_rw_command(struct disk_dump_partition *dump_part,
+ struct scsi_device *sdev, struct scsi_cmnd * scmd,
+ int rw, int block, void *buf, unsigned int len)
+{
+ int this_count = len >> 9;
+
+ memset(scmd, 0, sizeof(*scmd));
+
+ if (block + this_count > dump_part->nr_sects) {
+ Err("block number %d is larger than %lu",
+ block + this_count, dump_part->nr_sects);
+ return -EFBIG;
+ }
+
+ block += dump_part->start_sect;
+
+ /*
+ * If we have a 1K hardware sectorsize, prevent access to single
+ * 512 byte sectors. In theory we could handle this - in fact
+ * the scsi cdrom driver must be able to handle this because
+ * we typically use 1K blocksizes, and cdroms typically have
+ * 2K hardware sectorsizes. Of course, things are simpler
+ * with the cdrom, since it is read-only. For performance
+ * reasons, the filesystems should be able to handle this
+ * and not force the scsi disk driver to use bounce buffers
+ * for this.
+ */
+ if (sdev->sector_size == 1024) {
+ block = block >> 1;
+ this_count = this_count >> 1;
+ }
+ if (sdev->sector_size == 2048) {
+ block = block >> 2;
+ this_count = this_count >> 2;
+ }
+ if (sdev->sector_size == 4096) {
+ block = block >> 3;
+ this_count = this_count >> 3;
+ }
+ switch (rw) {
+ case WRITE:
+ if (!sdev->writeable) {
+ Err("writable media");
+ return 0;
+ }
+ scmd->cmnd[0] = WRITE_10;
+ break;
+ case READ:
+ scmd->cmnd[0] = READ_10;
+ break;
+ default:
+ Err("Unknown command %d", rw);
+ return -EINVAL;
+ }
+
+ if (this_count > 0xffff)
+ this_count = 0xffff;
+
+ scmd->cmnd[2] = (unsigned char) (block >> 24) & 0xff;
+ scmd->cmnd[3] = (unsigned char) (block >> 16) & 0xff;
+ scmd->cmnd[4] = (unsigned char) (block >> 8) & 0xff;
+ scmd->cmnd[5] = (unsigned char) block & 0xff;
+ scmd->cmnd[7] = (unsigned char) (this_count >> 8) & 0xff;
+ scmd->cmnd[8] = (unsigned char) this_count & 0xff;
+
+ init_scsi_command(sdev, scmd, buf, len,
+ (rw == WRITE ? SCSI_DATA_WRITE : SCSI_DATA_READ), 1);
+ return 0;
+}
+
+/*
+ * Check the status of scsi command and determine whether it is
+ * success, fail, or retriable.
+ *
+ * Return code
+ * > 0: should retry
+ * = 0: success
+ * < 0: fail
+ */
+static int cmd_result(struct scsi_cmnd *scmd)
+{
+ int status;
+
+ status = status_byte(scmd->result);
+
+ switch (scsi_decide_disposition(scmd)) {
+ case FAILED:
+ break;
+ case NEEDS_RETRY:
+ case ADD_TO_MLQUEUE:
+ return 1 /* retry */;
+ case SUCCESS:
+ if (host_byte(scmd->result) != DID_OK)
+ break;
+
+ if (status == GOOD || status == INTERMEDIATE_GOOD
+ || status == INTERMEDIATE_C_GOOD)
+ return 0;
+ if (status == CHECK_CONDITION &&
+ scmd->sense_buffer[2] == RECOVERED_ERROR)
+ return 0;
+ break;
+ default:
+ Err("bad disposition: %d", scmd->result);
+ return -EIO;
+ }
+
+ Err("command %x failed with 0x%x", scmd->cmnd[0], scmd->result);
+ return -EIO;
+}
+
+static int send_command(struct scsi_cmnd *scmd)
+{
+ struct Scsi_Host *host = scmd->device->host;
+ struct scsi_device *sdev = scmd->device;
+ int ret;
+
+ do {
+ if (!scsi_device_online(sdev)) {
+ Err("Scsi disk is not online");
+ return -EIO;
+ }
+ if (sdev->changed) {
+ Err("SCSI disk has been changed. Prohibiting further I/O");
+ return -EIO;
+ }
+
+ spin_lock(host->host_lock);
+ host->hostt->queuecommand(scmd, rw_intr);
+ spin_unlock(host->host_lock);
+
+ while (scmd->done != NULL) {
+ host->hostt->dump_poll(scmd->device);
+ udelay(100);
+ diskdump_update();
+ }
+ scmd->done = rw_intr;
+ } while ((ret = cmd_result(scmd)) > 0);
+
+ return ret;
+}
+
+/*
+ * If Write Cache Enable of disk device is not set, write I/O takes
+ * long long time. So enable WCE temporary and issue SYNCHRONIZE CACHE
+ * after all write I/Os are done, Following system reboot will reset
+ * WCE bit to original value.
+ */
+static void
+enable_write_cache(struct scsi_device *sdev)
+{
+ char buf[256];
+ int ret;
+ int data_len;
+
+ Dbg("enable write cache");
+ memset(buf, 0, 256);
+
+ init_mode_sense_command(sdev, &scsi_dump_cmnd, buf);
+ if ((ret = send_command(&scsi_dump_cmnd)) < 0) {
+ Warn("MODE SENSE failed");
+ return;
+ }
+
+ if (buf[14] & 0x04) /* WCE is already set */
+ return;
+
+ data_len = buf[0] + 1; /* Data length in mode parameter header */
+ buf[0] = 0;
+ buf[1] = 0;
+ buf[2] = 0;
+ buf[12] &= 0x7f; /* clear PS */
+ buf[14] |= 0x04; /* set WCE */
+
+ init_mode_select_command(sdev, &scsi_dump_cmnd, buf, data_len);
+ if ((ret = send_command(&scsi_dump_cmnd)) < 0) {
+ Warn("MODE SELECT failed");
+
+ init_sense_command(sdev, &scsi_dump_cmnd, buf);
+ if ((ret = send_command(&scsi_dump_cmnd)) < 0) {
+ Err("sense failed");
+ }
+ }
+}
+
+/*
+ * Check whether the dump device is sane enough to handle I/O.
+ *
+ * Return value:
+ * 0: the device is ok
+ * < 0: the device is not ok
+ * > 0: Cannot determine
+ */
+static int
+scsi_dump_sanity_check(struct disk_dump_device *dump_device)
+{
+ struct scsi_device *sdev = dump_device->device;
+ struct Scsi_Host *host = sdev->host;
+ int adapter_sanity = 0;
+ int sanity = 0;
+
+ if (!check_crc_module()) {
+ Err("checksum error. scsi dump module may be compromised.");
+ return -EINVAL;
+ }
+ /*
+ * If host's spinlock is already taken, assume it's part
+ * of crash and skip it.
+ */
+ if (!scsi_device_online(sdev)) {
+ Warn("device not online: host %d channel %d id %d lun %d",
+ host->host_no, sdev->channel, sdev->id, sdev->lun);
+ return -EIO;
+ }
+ if (sdev->changed) {
+ Err("SCSI disk has been changed. Prohibiting further I/O: host %d channel %d id %d lun %d",
+ host->host_no, sdev->channel, sdev->id, sdev->lun);
+ return -EIO;
+ }
+
+ if (host->hostt->dump_sanity_check) {
+ adapter_sanity = host->hostt->dump_sanity_check(sdev);
+ if (adapter_sanity < 0) {
+ Warn("adapter status is not sane");
+ return adapter_sanity;
+ }
+ }
+
+ if (!spin_is_locked(host->host_lock)) {
+ sanity = 0;
+ } else {
+ Warn("host_lock is held: host %d channel %d id %d lun %d",
+ host->host_no, sdev->channel, sdev->id, sdev->lun);
+ return -EIO;
+ }
+ return sanity + adapter_sanity;
+}
+
+/*
+ * Try to reset the host adapter. If the adapter does not have its host reset
+ * handler, try to use its bus device reset handler.
+ */
+static int scsi_dump_reset(struct scsi_device *sdev)
+{
+ struct Scsi_Host *host = sdev->host;
+ struct scsi_host_template *hostt = host->hostt;
+ char buf[256];
+ int ret, i;
+
+ init_sense_command(sdev, &scsi_dump_cmnd, buf);
+
+ if (hostt->eh_host_reset_handler) {
+ spin_lock(host->host_lock);
+ ret = hostt->eh_host_reset_handler(&scsi_dump_cmnd);
+ } else if (hostt->eh_bus_reset_handler) {
+ spin_lock(host->host_lock);
+ ret = hostt->eh_bus_reset_handler(&scsi_dump_cmnd);
+ } else
+ return 0;
+ spin_unlock(host->host_lock);
+
+ if (ret != SUCCESS) {
+ Err("adapter reset failed");
+ return -EIO;
+ }
+
+ /* bus reset settle time. 5sec for old disk devices */
+ for (i = 0; i < 5000; i++) {
+ diskdump_update();
+ mdelay(1);
+ }
+
+ Dbg("request sense");
+ if ((ret = send_command(&scsi_dump_cmnd)) < 0) {
+ Err("sense failed");
+ return -EIO;
+ }
+ return 0;
+}
+
+static int
+scsi_dump_quiesce(struct disk_dump_device *dump_device)
+{
+ struct scsi_device *sdev = dump_device->device;
+ struct Scsi_Host *host = sdev->host;
+ int ret;
+
+ if (host->hostt->dump_quiesce) {
+ ret = host->hostt->dump_quiesce(sdev);
+ if (ret < 0)
+ return ret;
+ }
+
+ Dbg("do bus reset");
+ if ((ret = scsi_dump_reset(sdev)) < 0)
+ return ret;
+
+ if (sdev->scsi_level >= SCSI_2)
+ enable_write_cache(sdev);
+
+ quiesce_ok = 1;
+ return 0;
+}
+
+static int scsi_dump_rw_block(struct disk_dump_partition *dump_part, int rw,
+ unsigned long dump_block_nr, void *buf, int len)
+{
+ struct disk_dump_device *dump_device = dump_part->device;
+ struct scsi_device *sdev = dump_device->device;
+ int block_nr = BLOCK_SECTOR(dump_block_nr);
+ int ret;
+
+ if (!quiesce_ok) {
+ Err("quiesce not called");
+ return -EIO;
+ }
+
+ ret = init_rw_command(dump_part, sdev, &scsi_dump_cmnd, rw,
+ block_nr, buf, DUMP_BLOCK_SIZE * len);
+ if (ret < 0) {
+ Err("init_rw_command failed");
+ return ret;
+ }
+ return send_command(&scsi_dump_cmnd);
+}
+
+static int
+scsi_dump_shutdown(struct disk_dump_device *dump_device)
+{
+ struct scsi_device *sdev = dump_device->device;
+ struct Scsi_Host *host = sdev->host;
+
+ if (sdev->scsi_level >= SCSI_2) {
+ init_sync_command(sdev, &scsi_dump_cmnd);
+ send_command(&scsi_dump_cmnd);
+ }
+
+ if (host->hostt->dump_shutdown)
+ return host->hostt->dump_shutdown(sdev);
+
+ return 0;
+}
+
+static void *scsi_dump_probe(struct block_device *bdev)
+{
+ struct scsi_device *sdev;
+
+ sdev = sd_find_scsi_device(bdev);
+ if (sdev == NULL)
+ return NULL;
+ if (!sdev->host->hostt->dump_poll)
+ return NULL;
+
+ return sdev;
+}
+
+
+struct disk_dump_device_ops scsi_dump_device_ops = {
+ .sanity_check = scsi_dump_sanity_check,
+ .rw_block = scsi_dump_rw_block,
+ .quiesce = scsi_dump_quiesce,
+ .shutdown = scsi_dump_shutdown,
+};
+
+static int scsi_dump_add_device(struct disk_dump_device *dump_device)
+{
+ struct scsi_device *sdev;
+ int error;
+
+ sdev = dump_device->device;
+ if (!sdev->host->hostt->dump_poll)
+ return -ENOTSUPP;
+
+ if ((error = scsi_device_get(sdev)) != 0)
+ return error;
+
+ memcpy(&dump_device->ops, &scsi_dump_device_ops,
+ sizeof(scsi_dump_device_ops));
+ if (sdev->host->max_sectors) {
+ dump_device->max_blocks =
+ (sdev->sector_size * sdev->host->max_sectors)
+ >> DUMP_BLOCK_SHIFT;
+ }
+ return 0;
+}
+
+static void scsi_dump_remove_device(struct disk_dump_device *dump_device)
+{
+ struct scsi_device *sdev = dump_device->device;
+
+ scsi_device_put(sdev);
+}
+
+static struct disk_dump_type scsi_dump_type = {
+ .probe = scsi_dump_probe,
+ .add_device = scsi_dump_add_device,
+ .remove_device = scsi_dump_remove_device,
+ .owner = THIS_MODULE,
+};
+
+static int init_scsi_dump(void)
+{
+ int ret;
+
+ if ((ret = register_disk_dump_type(&scsi_dump_type)) < 0) {
+ Err("register failed");
+ return ret;
+ }
+ set_crc_modules();
+ return ret;
+}
+
+static void cleanup_scsi_dump(void)
+{
+ if (unregister_disk_dump_type(&scsi_dump_type) < 0)
+ Err("register failed");
+}
+
+module_init(init_scsi_dump);
+module_exit(cleanup_scsi_dump);
+MODULE_LICENSE("GPL");
diff -Nur linux-2.6.6.org/drivers/scsi/scsi_error.c linux-2.6.6/drivers/scsi/scsi_error.c
--- linux-2.6.6.org/drivers/scsi/scsi_error.c 2004-06-04 21:22:20.000000000 +0900
+++ linux-2.6.6/drivers/scsi/scsi_error.c 2004-06-16 19:34:16.000000000 +0900
@@ -402,6 +402,9 @@
**/
static void scsi_eh_done(struct scsi_cmnd *scmd)
{
+ if (crashdump_mode())
+ return;
+
/*
* if the timeout handler is already running, then just set the
* flag which says we finished late, and return. we have no
diff -Nur linux-2.6.6.org/drivers/scsi/scsi_syms.c linux-2.6.6/drivers/scsi/scsi_syms.c
--- linux-2.6.6.org/drivers/scsi/scsi_syms.c 2004-06-04 21:22:21.000000000 +0900
+++ linux-2.6.6/drivers/scsi/scsi_syms.c 2004-06-16 19:34:16.000000000 +0900
@@ -25,6 +25,7 @@
#include "scsi.h"
#include "scsi_logging.h"
+#include "scsi_priv.h"
/*
@@ -107,3 +108,5 @@
*/
EXPORT_SYMBOL(scsi_add_timer);
EXPORT_SYMBOL(scsi_delete_timer);
+
+EXPORT_SYMBOL(scsi_decide_disposition);
diff -Nur linux-2.6.6.org/drivers/scsi/sd.c linux-2.6.6/drivers/scsi/sd.c
--- linux-2.6.6.org/drivers/scsi/sd.c 2004-06-04 21:22:22.000000000 +0900
+++ linux-2.6.6/drivers/scsi/sd.c 2004-06-16 19:34:16.000000000 +0900
@@ -680,6 +680,24 @@
.revalidate_disk = sd_revalidate_disk,
};
+#if defined(CONFIG_DISKDUMP) || defined(CONFIG_DISKDUMP_MODULE)
+struct scsi_device *sd_find_scsi_device(struct block_device *bdev)
+{
+ struct gendisk *disk = bdev->bd_disk;
+
+ /* Check whether dev is scsi or not */
+ if (!disk || (disk->fops != &sd_fops))
+ return NULL;
+
+ if(disk->private_data)
+ return scsi_disk(disk)->device;
+ else
+ return NULL;
+}
+
+EXPORT_SYMBOL(sd_find_scsi_device);
+#endif
+
/**
* sd_rw_intr - bottom half handler: called when the lower level
* driver has completed (successfully or otherwise) a scsi command.
diff -Nur linux-2.6.6.org/include/scsi/scsi_dump.h linux-2.6.6/include/scsi/scsi_dump.h
--- linux-2.6.6.org/include/scsi/scsi_dump.h 1970-01-01 09:00:00.000000000 +0900
+++ linux-2.6.6/include/scsi/scsi_dump.h 2004-06-16 19:34:16.000000000 +0900
@@ -0,0 +1,31 @@
+#ifndef _SCSI_DUMP_H
+#define _SCSI_DUMP_H
+
+/*
+ * linux/include/scsi/scsi_dump.h
+ *
+ * Copyright (C) 2004 FUJITSU LIMITED
+ * Written by Nobuhiro Tachino (ntachino@jp.fujitsu.com)
+ *
+ */
+/*
+ * 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, 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, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ */
+
+/* sd.c function */
+extern struct scsi_device *sd_find_scsi_device(struct block_device *);
+
+#endif /* _SCSI_DUMP_H */
diff -Nur linux-2.6.6.org/include/scsi/scsi_host.h linux-2.6.6/include/scsi/scsi_host.h
--- linux-2.6.6.org/include/scsi/scsi_host.h 2004-06-04 21:22:02.000000000 +0900
+++ linux-2.6.6/include/scsi/scsi_host.h 2004-06-16 19:34:16.000000000 +0900
@@ -345,6 +345,47 @@
* module_init/module_exit.
*/
struct list_head legacy_hosts;
+
+#if defined(CONFIG_SCSI_DUMP) || defined(CONFIG_SCSI_DUMP_MODULE)
+ /* operations for dump */
+
+ /*
+ * dump_sanity_check() checks if the selected device works normally.
+ * A device which returns an error status will not be selected as
+ * the dump device.
+ *
+ * Status: OPTIONAL
+ */
+ int (* dump_sanity_check)(struct scsi_device *);
+
+ /*
+ * dump_quiesce() is called after the device is selected as the
+ * dump device. Usually, host reset is executed and Write Cache
+ * Enable bit of the disk device is temporarily set for the
+ * dump operation.
+ *
+ * Status: OPTIONAL
+ */
+ int (* dump_quiesce)(struct scsi_device *);
+
+ /*
+ * dump_shutdown() is called after dump is completed. Usually
+ * "SYNCHRONIZE CACHE" command is issued to the disk.
+ *
+ * Status: OPTIONAL
+ */
+ int (* dump_shutdown)(struct scsi_device *);
+
+ /*
+ * dump_poll() should call the interrupt handler. It is called
+ * repeatedly after queuecommand() is issued, and until the command
+ * is completed. If the low level device driver support crash dump,
+ * it must have this routine.
+ *
+ * Status: OPTIONAL
+ */
+ void (* dump_poll)(struct scsi_device *);
+#endif
};
/*
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH 3/4][Diskdump]Update patches
2004-06-16 12:39 [PATCH 0/4][Diskdump]Update patches Takao Indoh
2004-06-16 12:45 ` [PATCH 1/4][Diskdump]Update patches Takao Indoh
2004-06-16 12:46 ` [PATCH 2/4][Diskdump]Update patches Takao Indoh
@ 2004-06-16 12:48 ` Takao Indoh
2004-06-16 12:50 ` [PATCH 4/4][Diskdump]Update patches Takao Indoh
2004-06-16 17:25 ` [PATCH 0/4][Diskdump]Update patches Jeff Garzik
4 siblings, 0 replies; 9+ messages in thread
From: Takao Indoh @ 2004-06-16 12:48 UTC (permalink / raw)
To: linux-kernel
This is a patch for aic7xxx driver.
diff -Nur linux-2.6.6.org/drivers/scsi/aic7xxx/aic7xxx_osm.c linux-2.6.6/drivers/scsi/aic7xxx/aic7xxx_osm.c
--- linux-2.6.6.org/drivers/scsi/aic7xxx/aic7xxx_osm.c 2004-06-04 21:22:20.000000000 +0900
+++ linux-2.6.6/drivers/scsi/aic7xxx/aic7xxx_osm.c 2004-06-16 19:34:16.000000000 +0900
@@ -774,6 +774,10 @@
static int ahc_linux_bus_reset(Scsi_Cmnd *);
static int ahc_linux_dev_reset(Scsi_Cmnd *);
static int ahc_linux_abort(Scsi_Cmnd *);
+#if defined(CONFIG_SCSI_DUMP) || defined(CONFIG_SCSI_DUMP_MODULE)
+static int ahc_linux_sanity_check(struct scsi_device *);
+static void ahc_linux_poll(struct scsi_device *);
+#endif
/*
* Calculate a safe value for AHC_NSEG (as expressed through ahc_linux_nseg).
@@ -1313,6 +1317,10 @@
.slave_alloc = ahc_linux_slave_alloc,
.slave_configure = ahc_linux_slave_configure,
.slave_destroy = ahc_linux_slave_destroy,
+#if defined(CONFIG_SCSI_DUMP) || defined(CONFIG_SCSI_DUMP_MODULE)
+ .dump_sanity_check = ahc_linux_sanity_check,
+ .dump_poll = ahc_linux_poll,
+#endif
};
/**************************** Tasklet Handler *********************************/
@@ -3863,6 +3871,42 @@
return IRQ_RETVAL(ours);
}
+static int
+ahc_linux_sanity_check(struct scsi_device *device)
+{
+ struct ahc_softc *ahc;
+ struct ahc_linux_device *dev;
+
+ ahc = *(struct ahc_softc **)device->host->hostdata;
+ dev = ahc_linux_get_device(ahc, device->channel,
+ device->id, device->lun,
+ /*alloc*/FALSE);
+ if (dev == NULL)
+ return -ENXIO;
+ if (ahc->platform_data->qfrozen || dev->qfrozen)
+ return -EBUSY;
+ if (spin_is_locked(&ahc->platform_data->spin_lock))
+ return -EBUSY;
+ return 0;
+}
+
+#if defined(CONFIG_SCSI_DUMP) || defined(CONFIG_SCSI_DUMP_MODULE)
+static void
+ahc_linux_poll(struct scsi_device *device)
+{
+ struct ahc_softc *ahc;
+ struct ahc_linux_device *dev;
+
+ ahc = *(struct ahc_softc **)device->host->hostdata;
+ ahc_intr(ahc);
+ while ((dev = ahc_linux_next_device_to_run(ahc)) != NULL) {
+ TAILQ_REMOVE(&ahc->platform_data->device_runq, dev, links);
+ dev->flags &= ~AHC_DEV_ON_RUN_LIST;
+ ahc_linux_check_device_queue(ahc, dev);
+ }
+ ahc_linux_run_complete_queue(ahc);
+}
+
void
ahc_platform_flushwork(struct ahc_softc *ahc)
{
@@ -3870,6 +3914,7 @@
while (ahc_linux_run_complete_queue(ahc) != NULL)
;
}
+#endif
static struct ahc_linux_target*
ahc_linux_alloc_target(struct ahc_softc *ahc, u_int channel, u_int target)
diff -Nur linux-2.6.6.org/drivers/scsi/aic7xxx/aic7xxx_osm.h linux-2.6.6/drivers/scsi/aic7xxx/aic7xxx_osm.h
--- linux-2.6.6.org/drivers/scsi/aic7xxx/aic7xxx_osm.h 2004-06-04 21:22:20.000000000 +0900
+++ linux-2.6.6/drivers/scsi/aic7xxx/aic7xxx_osm.h 2004-06-16 19:34:16.000000000 +0900
@@ -67,6 +67,7 @@
#include <linux/smp_lock.h>
#include <linux/version.h>
#include <linux/module.h>
+#include <linux/diskdumplib.h>
#include <asm/byteorder.h>
#include <asm/io.h>
@@ -82,6 +83,8 @@
#include <linux/malloc.h>
#endif
+#include <scsi/scsi_dump.h>
+
/* Core SCSI definitions */
#define AIC_LIB_PREFIX ahc
#include "scsi.h"
@@ -253,6 +256,22 @@
*/
#define ahc_dmamap_sync(ahc, dma_tag, dmamap, offset, len, op)
+
+/******************************** Disk dump ***********************************/
+#if defined(CONFIG_DISKDUMP) || defined(CONFIG_DISKDUMP_MODULE)
+#undef add_timer
+#define add_timer diskdump_add_timer
+#undef del_timer_sync
+#define del_timer_sync diskdump_del_timer
+#undef del_timer
+#define del_timer diskdump_del_timer
+#undef mod_timer
+#define mod_timer diskdump_mod_timer
+
+#define tasklet_schedule diskdump_tasklet_schedule
+#endif
+
+
/************************** Timer DataStructures ******************************/
typedef struct timer_list ahc_timer_t;
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH 4/4][Diskdump]Update patches
2004-06-16 12:39 [PATCH 0/4][Diskdump]Update patches Takao Indoh
` (2 preceding siblings ...)
2004-06-16 12:48 ` [PATCH 3/4][Diskdump]Update patches Takao Indoh
@ 2004-06-16 12:50 ` Takao Indoh
2004-06-16 17:25 ` [PATCH 0/4][Diskdump]Update patches Jeff Garzik
4 siblings, 0 replies; 9+ messages in thread
From: Takao Indoh @ 2004-06-16 12:50 UTC (permalink / raw)
To: linux-kernel
This is a patch for aic79xx driver.
diff -Nur linux-2.6.6.org/drivers/scsi/aic7xxx/aic79xx_osm.c linux-2.6.6/drivers/scsi/aic7xxx/aic79xx_osm.c
--- linux-2.6.6.org/drivers/scsi/aic7xxx/aic79xx_osm.c 2004-06-04 21:22:20.000000000 +0900
+++ linux-2.6.6/drivers/scsi/aic7xxx/aic79xx_osm.c 2004-06-16 19:34:16.000000000 +0900
@@ -786,6 +786,10 @@
static int ahd_linux_bus_reset(Scsi_Cmnd *);
static int ahd_linux_dev_reset(Scsi_Cmnd *);
static int ahd_linux_abort(Scsi_Cmnd *);
+#if defined(CONFIG_SCSI_DUMP) || defined(CONFIG_SCSI_DUMP_MODULE)
+static int ahd_linux_sanity_check(struct scsi_device *);
+static void ahd_linux_poll(struct scsi_device *);
+#endif
/*
* Calculate a safe value for AHD_NSEG (as expressed through ahd_linux_nseg).
@@ -1684,6 +1688,10 @@
.slave_alloc = ahd_linux_slave_alloc,
.slave_configure = ahd_linux_slave_configure,
.slave_destroy = ahd_linux_slave_destroy,
+#if defined(CONFIG_SCSI_DUMP) || defined(CONFIG_SCSI_DUMP_MODULE)
+ .dump_sanity_check = ahd_linux_sanity_check,
+ .dump_poll = ahd_linux_poll,
+#endif
};
/**************************** Tasklet Handler *********************************/
@@ -4190,6 +4198,41 @@
return IRQ_RETVAL(ours);
}
+#if defined(CONFIG_SCSI_DUMP) || defined(CONFIG_SCSI_DUMP_MODULE)
+static int
+ahd_linux_sanity_check(struct scsi_device *device)
+{
+ struct ahd_softc *ahd;
+ struct ahd_linux_device *dev;
+
+ ahd = *(struct ahd_softc **)device->host->hostdata;
+ dev = ahd_linux_get_device(ahd, device->channel,
+ device->id, device->lun,
+ /*alloc*/FALSE);
+
+ if (dev == NULL)
+ return -ENXIO;
+ if (ahd->platform_data->qfrozen || dev->qfrozen)
+ return -EBUSY;
+ if (spin_is_locked(&ahd->platform_data->spin_lock))
+ return -EBUSY;
+ return 0;
+}
+
+static void
+ahd_linux_poll(struct scsi_device *device)
+{
+ struct ahd_softc *ahd;
+ int ours;
+
+ ahd = *(struct ahd_softc **)device->host->hostdata;
+ ours = ahd_intr(ahd);
+ if (ahd_linux_next_device_to_run(ahd) != NULL)
+ ahd_schedule_runq(ahd);
+ ahd_linux_run_complete_queue(ahd);
+}
+#endif
+
void
ahd_platform_flushwork(struct ahd_softc *ahd)
{
diff -Nur linux-2.6.6.org/drivers/scsi/aic7xxx/aic79xx_osm.h linux-2.6.6/drivers/scsi/aic7xxx/aic79xx_osm.h
--- linux-2.6.6.org/drivers/scsi/aic7xxx/aic79xx_osm.h 2004-06-04 21:22:20.000000000 +0900
+++ linux-2.6.6/drivers/scsi/aic7xxx/aic79xx_osm.h 2004-06-16 19:34:16.000000000 +0900
@@ -50,6 +50,9 @@
#include <linux/smp_lock.h>
#include <linux/version.h>
#include <linux/module.h>
+#if defined(CONFIG_SCSI_DUMP) || defined(CONFIG_SCSI_DUMP_MODULE)
+#include <linux/diskdumplib.h>
+#endif
#include <asm/byteorder.h>
#include <asm/io.h>
@@ -65,6 +68,10 @@
#include <linux/malloc.h>
#endif
+#if defined(CONFIG_SCSI_DUMP) || defined(CONFIG_SCSI_DUMP_MODULE)
+#include <scsi/scsi_dump.h>
+#endif
+
/* Core SCSI definitions */
#define AIC_LIB_PREFIX ahd
#include "scsi.h"
@@ -96,6 +103,20 @@
/* No debugging code. */
#endif
+/********************************** Disk Dump *********************************/
+#if defined(CONFIG_SCSI_DUMP) || defined(CONFIG_SCSI_DUMP_MODULE)
+#undef add_timer
+#define add_timer diskdump_add_timer
+#undef del_timer_sync
+#define del_timer_sync diskdump_del_timer
+#undef del_timer
+#define del_timer diskdump_del_timer
+#undef mod_timer
+#define mod_timer diskdump_mod_timer
+
+#define tasklet_schedule diskdump_tasklet_schedule
+#endif
+
/********************************** Misc Macros *******************************/
#define roundup(x, y) ((((x)+((y)-1))/(y))*(y))
#define powerof2(x) ((((x)-1)&(x))==0)
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH 0/4][Diskdump]Update patches
2004-06-16 12:39 [PATCH 0/4][Diskdump]Update patches Takao Indoh
` (3 preceding siblings ...)
2004-06-16 12:50 ` [PATCH 4/4][Diskdump]Update patches Takao Indoh
@ 2004-06-16 17:25 ` Jeff Garzik
4 siblings, 0 replies; 9+ messages in thread
From: Jeff Garzik @ 2004-06-16 17:25 UTC (permalink / raw)
To: Takao Indoh; +Cc: linux-kernel
Takao Indoh wrote:
> static inline void tasklet_schedule(struct tasklet_struct *t)
> {
> + if(crashdump_mode()) {
> + diskdump_tasklet_schedule(t);
> + return;
> + }
> +
I would suggest
if (unlikely(crashdump_mode()))
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH 0/4][Diskdump]Update patches
@ 2004-06-22 13:47 Takao Indoh
0 siblings, 0 replies; 9+ messages in thread
From: Takao Indoh @ 2004-06-22 13:47 UTC (permalink / raw)
To: linux-kernel; +Cc: Ingo Molnar, Christoph Hellwig, Andi Kleen
Hi!
I update the Diskdump patches!
- fix timer problem
- support kernel 2.6.7
Source code of tool(diskdumptuils) can be downloaded from
http://sourceforge.net/projects/lkdump
TODO:
- Replace proc interface with sysfs.
- Merge scsi_dump with scsi_mod.
I solved the timer problem by the method which Ingo Molnar proposed.
On Thu, 17 Jun 2004 14:13:56 +0200, Ingo Molnar wrote:
>but there's another possible method (suggested by Alan Cox) that
>requires no changes to the timer API hotpaths: 'clear' all timer lists
>upon a crash [once all CPUs have stopped and irqs are disabled] and just
>let the drivers use the normal timer APIs. Drive timer execution via a
>polling method.
>
>this basically approximates your polling based implementation but uses
>the existing kernel timer data structures and timer mechanism so should
>be robust and compatible. It doesnt rely on any previous state (because
>all currently pending timers are discarded) so it's as crash-safe as
>possible.
The following is core part of patches related to timer problem.
The complete patch is posted later.
diff -Nur linux-2.6.7.org/include/linux/interrupt.h linux-2.6.7/include/linux/interrupt.h
--- linux-2.6.7.org/include/linux/interrupt.h 2004-06-22 10:27:34.000000000 +0900
+++ linux-2.6.7/include/linux/interrupt.h 2004-06-22 22:26:39.000000000 +0900
@@ -246,4 +246,8 @@
extern int probe_irq_off(unsigned long); /* returns 0 or negative on failure */
extern unsigned int probe_irq_mask(unsigned long); /* returns mask of ISA interrupts */
+
+extern void dump_clear_tasklet(void);
+extern void dump_run_tasklet(void);
+
#endif
diff -Nur linux-2.6.7.org/include/linux/timer.h linux-2.6.7/include/linux/timer.h
--- linux-2.6.7.org/include/linux/timer.h 2004-06-22 10:27:31.000000000 +0900
+++ linux-2.6.7/include/linux/timer.h 2004-06-22 22:26:39.000000000 +0900
@@ -99,4 +99,7 @@
extern void run_local_timers(void);
extern void it_real_fn(unsigned long);
+extern void dump_clear_timers(void);
+extern void dump_run_timers(void);
+
#endif
diff -Nur linux-2.6.7.org/include/linux/workqueue.h linux-2.6.7/include/linux/workqueue.h
--- linux-2.6.7.org/include/linux/workqueue.h 2004-06-22 10:27:35.000000000 +0900
+++ linux-2.6.7/include/linux/workqueue.h 2004-06-22 22:26:39.000000000 +0900
@@ -84,4 +84,7 @@
return ret;
}
+extern void dump_clear_workqueue(void);
+extern void dump_run_workqueue(void);
+
#endif
diff -Nur linux-2.6.7.org/kernel/softirq.c linux-2.6.7/kernel/softirq.c
--- linux-2.6.7.org/kernel/softirq.c 2004-06-22 10:27:25.000000000 +0900
+++ linux-2.6.7/kernel/softirq.c 2004-06-22 22:26:39.000000000 +0900
@@ -314,6 +314,38 @@
EXPORT_SYMBOL(tasklet_kill);
+struct tasklet_head saved_tasklet;
+
+void dump_clear_tasklet(void)
+{
+ saved_tasklet.list = __get_cpu_var(tasklet_vec).list;
+ __get_cpu_var(tasklet_vec).list = NULL;
+}
+
+EXPORT_SYMBOL(dump_clear_tasklet);
+
+void dump_run_tasklet(void)
+{
+ struct tasklet_struct *list;
+
+ list = __get_cpu_var(tasklet_vec).list;
+ __get_cpu_var(tasklet_vec).list = NULL;
+
+ while (list) {
+ struct tasklet_struct *t = list;
+ list = list->next;
+
+ if (!atomic_read(&t->count) &&
+ (test_and_clear_bit(TASKLET_STATE_SCHED, &t->state)))
+ t->func(t->data);
+
+ t->next = __get_cpu_var(tasklet_vec).list;
+ __get_cpu_var(tasklet_vec).list = t;
+ }
+}
+
+EXPORT_SYMBOL(dump_run_tasklet);
+
void __init softirq_init(void)
{
open_softirq(TASKLET_SOFTIRQ, tasklet_action, NULL);
diff -Nur linux-2.6.7.org/kernel/timer.c linux-2.6.7/kernel/timer.c
--- linux-2.6.7.org/kernel/timer.c 2004-06-22 10:27:25.000000000 +0900
+++ linux-2.6.7/kernel/timer.c 2004-06-22 22:26:39.000000000 +0900
@@ -31,6 +31,7 @@
#include <linux/time.h>
#include <linux/jiffies.h>
#include <linux/cpu.h>
+#include <linux/delay.h>
#include <asm/uaccess.h>
#include <asm/unistd.h>
@@ -424,7 +425,6 @@
{
struct timer_list *timer;
- spin_lock_irq(&base->lock);
while (time_after_eq(jiffies, base->timer_jiffies)) {
struct list_head work_list = LIST_HEAD_INIT(work_list);
struct list_head *head = &work_list;
@@ -460,6 +460,12 @@
}
}
set_running_timer(base, NULL);
+}
+
+static inline void _run_timers(tvec_base_t *base)
+{
+ spin_lock_irq(&base->lock);
+ __run_timers(base);
spin_unlock_irq(&base->lock);
}
@@ -909,7 +915,7 @@
tvec_base_t *base = &__get_cpu_var(tvec_bases);
if (time_after_eq(jiffies, base->timer_jiffies))
- __run_timers(base);
+ _run_timers(base);
}
/*
@@ -1105,6 +1111,12 @@
struct timer_list timer;
unsigned long expire;
+ if (unlikely(crashdump_mode())) {
+ mdelay(timeout);
+ set_current_state(TASK_RUNNING);
+ return timeout;
+ }
+
switch (timeout)
{
case MAX_SCHEDULE_TIMEOUT:
@@ -1308,7 +1320,7 @@
return 0;
}
-static void __devinit init_timers_cpu(int cpu)
+static void /* __devinit */ init_timers_cpu(int cpu)
{
int j;
tvec_base_t *base;
@@ -1327,6 +1339,27 @@
base->timer_jiffies = jiffies;
}
+static tvec_base_t saved_tvec_base;
+
+void dump_clear_timers(void)
+{
+ tvec_base_t *base = &per_cpu(tvec_bases, smp_processor_id());
+
+ memcpy(&saved_tvec_base, base, sizeof(saved_tvec_base));
+ init_timers_cpu(smp_processor_id());
+}
+
+EXPORT_SYMBOL(dump_clear_timers);
+
+void dump_run_timers(void)
+{
+ tvec_base_t *base = &__get_cpu_var(tvec_bases);
+
+ __run_timers(base);
+}
+
+EXPORT_SYMBOL(dump_run_timers);
+
#ifdef CONFIG_HOTPLUG_CPU
static int migrate_timer_list(tvec_base_t *new_base, struct list_head *head)
{
diff -Nur linux-2.6.7.org/kernel/workqueue.c linux-2.6.7/kernel/workqueue.c
--- linux-2.6.7.org/kernel/workqueue.c 2004-06-22 10:27:25.000000000 +0900
+++ linux-2.6.7/kernel/workqueue.c 2004-06-22 22:26:39.000000000 +0900
@@ -424,6 +424,37 @@
}
+struct cpu_workqueue_struct saved_cwq;
+
+void dump_clear_workqueue(void)
+{
+ int cpu = smp_processor_id();
+ struct cpu_workqueue_struct *cwq = keventd_wq->cpu_wq + cpu;
+
+ memcpy(&saved_cwq, cwq, sizeof(saved_cwq));
+ spin_lock_init(&cwq->lock);
+ INIT_LIST_HEAD(&cwq->worklist);
+ init_waitqueue_head(&cwq->more_work);
+ init_waitqueue_head(&cwq->work_done);
+}
+
+void dump_run_workqueue(void)
+{
+ struct cpu_workqueue_struct *cwq;
+
+ cwq = keventd_wq->cpu_wq + smp_processor_id();
+ while (!list_empty(&cwq->worklist)) {
+ struct work_struct *work = list_entry(cwq->worklist.next,
+ struct work_struct, entry);
+ void (*f) (void *) = work->func;
+ void *data = work->data;
+
+ list_del_init(cwq->worklist.next);
+ clear_bit(0, &work->pending);
+ f(data);
+ }
+}
+
#ifdef CONFIG_HOTPLUG_CPU
/* Take the work from this (downed) CPU. */
static void take_over_work(struct workqueue_struct *wq, unsigned int cpu)
@@ -507,3 +538,6 @@
EXPORT_SYMBOL(schedule_delayed_work);
EXPORT_SYMBOL(flush_scheduled_work);
+EXPORT_SYMBOL(dump_clear_workqueue);
+EXPORT_SYMBOL(dump_run_workqueue);
+
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH 0/4][Diskdump]Update patches
@ 2004-07-07 8:26 Takao Indoh
0 siblings, 0 replies; 9+ messages in thread
From: Takao Indoh @ 2004-07-07 8:26 UTC (permalink / raw)
To: linux-kernel
Hi!
I release new diskdump patches.
- Support sysfs interface instead of proc
- Fix some bugs
Source code can be downloaded from
http://sourceforge.net/projects/lkdump
The main change from previous version is replacing proc interface with
sysfs.
If you want to add a new dump device /dev/sda3,
# echo 3 > /sys/block/sda/device/dump
Like this, you need write the partition number into the appropriate
sysfs entry. You can also do the same thing using diskdumpctl command,
which is included in the diskdumputils.
# diskdumpctl /dev/sda3
If you want to remove the dump device /dev/sda3,
# echo -3 > /sys/block/sda/device/dump
or
# diskdumpctl -u /dev/sda1
Here is a part of patch of sysfs interface.
I added a new attribute "dump" to the /sys/block/sdX/device. The handler
of show/store operation calls a function which is registered via
sdev_dump_handler_register(). When the scsi_dump module is installed,
sysfs handler of scsi_dump is registered using
sdev_dump_handler_register().
Please feel free to comment!
diff -Nur linux-2.6.7.org/drivers/scsi/scsi_sysfs.c linux-2.6.7/drivers/scsi/scsi_sysfs.c
--- linux-2.6.7.org/drivers/scsi/scsi_sysfs.c 2004-06-22 10:27:50.000000000 +0900
+++ linux-2.6.7/drivers/scsi/scsi_sysfs.c 2004-07-07 17:01:26.146353152 +0900
@@ -375,6 +375,66 @@
DEVICE_ATTR(state, S_IRUGO | S_IWUSR, show_state_field, store_state_field);
+static DECLARE_MUTEX(sdev_dump_mutex);
+
+static struct device_attribute sdev_dump_attr = {
+ .show = NULL,
+ .store = NULL,
+};
+
+int sdev_dump_handler_register(struct device_attribute* attr)
+{
+ if (sdev_dump_attr.show || sdev_dump_attr.store)
+ return -EEXIST;
+
+ down(&sdev_dump_mutex);
+ sdev_dump_attr = *attr;
+ up(&sdev_dump_mutex);
+
+ return 0;
+}
+
+EXPORT_SYMBOL(sdev_dump_handler_register);
+
+void sdev_dump_handler_unregister(void)
+{
+ down(&sdev_dump_mutex);
+ sdev_dump_attr.show = NULL;
+ sdev_dump_attr.store = NULL;
+ up(&sdev_dump_mutex);
+}
+
+EXPORT_SYMBOL(sdev_dump_handler_unregister);
+
+static ssize_t
+sdev_store_dump(struct device *dev, const char *buf, size_t count)
+{
+ ssize_t ret = count;
+
+ down(&sdev_dump_mutex);
+ if (sdev_dump_attr.store)
+ ret = sdev_dump_attr.store(dev, buf, count);
+ up(&sdev_dump_mutex);
+
+ return ret;
+}
+
+static ssize_t
+sdev_show_dump(struct device *dev, char *buf)
+{
+ ssize_t ret;
+
+ down(&sdev_dump_mutex);
+ if (sdev_dump_attr.show)
+ ret = sdev_dump_attr.show(dev, buf);
+ else
+ ret = snprintf(buf, 20, "handler not found\n");
+ up(&sdev_dump_mutex);
+
+ return ret;
+}
+
+DEVICE_ATTR(dump, S_IRUGO | S_IWUSR, sdev_show_dump, sdev_store_dump);
/* Default template for device attributes. May NOT be modified */
static struct device_attribute *scsi_sysfs_sdev_attrs[] = {
@@ -389,6 +449,7 @@
&dev_attr_delete,
&dev_attr_state,
&dev_attr_timeout,
+ &dev_attr_dump,
NULL
};
Best Regards,
Takao Indoh
^ permalink raw reply [flat|nested] 9+ messages in thread
* [PATCH 0/4][Diskdump]Update patches
@ 2004-07-09 7:15 Takao Indoh
0 siblings, 0 replies; 9+ messages in thread
From: Takao Indoh @ 2004-07-09 7:15 UTC (permalink / raw)
To: linux-kernel; +Cc: arjanv, hch
Hi all,
I improved the patches which I posted on 7th Jul.
- Improve sysfs interface code
- Bug fix related to eh_timeout of scsi_cmd
Source code can be downloaded from
http://sourceforge.net/projects/lkdump
On Mon, 21 Jun 2004 10:01:29 +0200, Arjan van de Ven wrote:
>On Mon, Jun 21, 2004 at 04:59:52PM +0900, Takao Indoh wrote:
>> Hi,
>>
>> Now I am fixing diskdump according to comments by you and Christoph.
>>
>> On Fri, 11 Jun 2004 13:50:45 +0200, Arjan van de Ven wrote:
>>
>> >> +#ifdef CONFIG_PROC_FS
>> >> +static int proc_ioctl(struct inode *inode, struct file *file, unsigned
>> >> int cmd, unsigned long param)
>> >
>> >
>> >ehhh this looks evil
>>
>> Do you mean I should use not ioctl but the following style?
>>
>> echo "add /dev/hda1" > /proc/diskdump
>> echo "delete /dev/hda1" > /proc/diskdump
>
>well no since /dev/hda is pointless; major/minor pairs maybe.
>But why in /proc???? it sounds like a sysfs job to me, where you probably
>want to represent a dump relationship with a symlink, and use "rm" to remove
>an entry..
As I wrote in the previous mail, I replaced proc interface with
sysfs.
If you want to add a new dump device /dev/sda3,
# echo 3 > /sys/block/sda/device/dump
If you want to remove the dump device /dev/sda3,
# echo -3 > /sys/block/sda/device/dump
I added a new attribute "dump" to the /sys/block/sdX/device.
diff -Nur linux-2.6.7.org/drivers/scsi/scsi_sysfs.c linux-2.6.7/drivers/scsi/scsi_sysfs.c
--- linux-2.6.7.org/drivers/scsi/scsi_sysfs.c 2004-06-22 10:27:50.000000000 +0900
+++ linux-2.6.7/drivers/scsi/scsi_sysfs.c 2004-07-09 11:36:41.451983216 +0900
@@ -11,6 +11,7 @@
#include <linux/init.h>
#include <linux/blkdev.h>
#include <linux/device.h>
+#include <linux/diskdump.h>
#include <scsi/scsi_host.h>
#include <scsi/scsi_transport.h>
@@ -375,6 +376,7 @@
DEVICE_ATTR(state, S_IRUGO | S_IWUSR, show_state_field, store_state_field);
+DEVICE_ATTR(dump, S_IRUGO | S_IWUSR, diskdump_sysfs_show, diskdump_sysfs_store);
/* Default template for device attributes. May NOT be modified */
static struct device_attribute *scsi_sysfs_sdev_attrs[] = {
@@ -389,6 +391,7 @@
&dev_attr_delete,
&dev_attr_state,
&dev_attr_timeout,
+ &dev_attr_dump,
NULL
};
diskdump_sysfs_show/store is the sysfs handler of diskdump. These
functions passes struct device to the diskdump module.
Diskdump changes "struct device" to "struct scsi_device" like this:
+static void *scsi_dump_probe(struct device *dev)
+{
+ struct scsi_device *sdev;
+
+ if ((dev->bus == NULL) || (dev->bus->name == NULL) ||
+ strncmp(dev->bus->name, "scsi", 4))
+ return NULL;
+
+ sdev = to_scsi_device(dev);
+ if (!sdev->host->hostt->dump_poll)
+ return NULL;
+
+ return sdev;
+}
I used device->bus->name to confirm device is scsi device or not.
Please comment.
Best Regards,
Takao Indoh
^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2004-07-09 7:13 UTC | newest]
Thread overview: 9+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-06-16 12:39 [PATCH 0/4][Diskdump]Update patches Takao Indoh
2004-06-16 12:45 ` [PATCH 1/4][Diskdump]Update patches Takao Indoh
2004-06-16 12:46 ` [PATCH 2/4][Diskdump]Update patches Takao Indoh
2004-06-16 12:48 ` [PATCH 3/4][Diskdump]Update patches Takao Indoh
2004-06-16 12:50 ` [PATCH 4/4][Diskdump]Update patches Takao Indoh
2004-06-16 17:25 ` [PATCH 0/4][Diskdump]Update patches Jeff Garzik
-- strict thread matches above, loose matches on Subject: below --
2004-06-22 13:47 Takao Indoh
2004-07-07 8:26 Takao Indoh
2004-07-09 7:15 Takao Indoh
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox