[MTD] oops and panic message logging to MTD device
authorRichard Purdie <richard@openedhand.com>
Tue, 29 May 2007 12:31:42 +0000 (13:31 +0100)
committerDavid Woodhouse <dwmw2@infradead.org>
Thu, 28 Jun 2007 19:18:02 +0000 (20:18 +0100)
Kernel oops and panic messages are invaluable when debugging crashes.
These messages often don't make it to flash based logging methods (say a
syslog on jffs2) due to the overheads involved in writing to flash.

This patch allows you to turn an MTD partition into a circular log
buffer where kernel oops and panic messages are written to. The messages
are obtained by registering a console driver and checking
oops_in_progress. Erases are performed in advance to maximise the
chances of a saving messages.

To activate it, add console=ttyMTDx to the kernel commandline (where x
is the mtd device number to use).

Signed-off-by: Richard Purdie <rpurdie@openedhand.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
drivers/mtd/Kconfig
drivers/mtd/Makefile
drivers/mtd/mtdoops.c [new file with mode: 0644]

index fbec8cd..8848e8a 100644 (file)
@@ -278,6 +278,14 @@ config SSFDC
          This enables read only access to SmartMedia formatted NAND
          flash. You can mount it with FAT file system.
 
+config MTD_OOPS
+       tristate "Log panic/oops to an MTD buffer"
+       depends on MTD
+       help
+         This enables panic and oops messages to be logged to a circular
+         buffer in a flash partition where it can be read back at some
+         later point.
+
 source "drivers/mtd/chips/Kconfig"
 
 source "drivers/mtd/maps/Kconfig"
index 451adcc..024d0e5 100644 (file)
@@ -22,6 +22,7 @@ obj-$(CONFIG_NFTL)            += nftl.o
 obj-$(CONFIG_INFTL)            += inftl.o
 obj-$(CONFIG_RFD_FTL)          += rfd_ftl.o
 obj-$(CONFIG_SSFDC)            += ssfdc.o
+obj-$(CONFIG_MTD_OOPS)         += mtdoops.o
 
 nftl-objs              := nftlcore.o nftlmount.o
 inftl-objs             := inftlcore.o inftlmount.o
diff --git a/drivers/mtd/mtdoops.c b/drivers/mtd/mtdoops.c
new file mode 100644 (file)
index 0000000..cfc28ab
--- /dev/null
@@ -0,0 +1,365 @@
+/*
+ * MTD Oops/Panic logger
+ *
+ * Copyright (C) 2007 Nokia Corporation. All rights reserved.
+ *
+ * Author: Richard Purdie <rpurdie@openedhand.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * version 2 as published by the Free Software Foundation.
+ *
+ * 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., 51 Franklin St, Fifth Floor, Boston, MA
+ * 02110-1301 USA
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/console.h>
+#include <linux/vmalloc.h>
+#include <linux/workqueue.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
+#include <linux/mtd/mtd.h>
+
+#define OOPS_PAGE_SIZE 4096
+
+static struct mtdoops_context {
+       int mtd_index;
+       struct work_struct work;
+       struct mtd_info *mtd;
+       int oops_pages;
+       int nextpage;
+       int nextcount;
+
+       void *oops_buf;
+       int ready;
+       int writecount;
+} oops_cxt;
+
+static void mtdoops_erase_callback(struct erase_info *done)
+{
+       wait_queue_head_t *wait_q = (wait_queue_head_t *)done->priv;
+       wake_up(wait_q);
+}
+
+static int mtdoops_erase_block(struct mtd_info *mtd, int offset)
+{
+       struct erase_info erase;
+       DECLARE_WAITQUEUE(wait, current);
+       wait_queue_head_t wait_q;
+       int ret;
+
+       init_waitqueue_head(&wait_q);
+       erase.mtd = mtd;
+       erase.callback = mtdoops_erase_callback;
+       erase.addr = offset;
+       if (mtd->erasesize < OOPS_PAGE_SIZE)
+               erase.len = OOPS_PAGE_SIZE;
+       else
+               erase.len = mtd->erasesize;
+       erase.priv = (u_long)&wait_q;
+
+       set_current_state(TASK_INTERRUPTIBLE);
+       add_wait_queue(&wait_q, &wait);
+
+       ret = mtd->erase(mtd, &erase);
+       if (ret) {
+               set_current_state(TASK_RUNNING);
+               remove_wait_queue(&wait_q, &wait);
+               printk (KERN_WARNING "mtdoops: erase of region [0x%x, 0x%x] "
+                                    "on \"%s\" failed\n",
+                       erase.addr, erase.len, mtd->name);
+               return ret;
+       }
+
+       schedule();  /* Wait for erase to finish. */
+       remove_wait_queue(&wait_q, &wait);
+
+       return 0;
+}
+
+static int mtdoops_inc_counter(struct mtdoops_context *cxt)
+{
+       struct mtd_info *mtd = cxt->mtd;
+       size_t retlen;
+       u32 count;
+       int ret;
+
+       cxt->nextpage++;
+       if (cxt->nextpage > cxt->oops_pages)
+               cxt->nextpage = 0;
+       cxt->nextcount++;
+       if (cxt->nextcount == 0xffffffff)
+               cxt->nextcount = 0;
+
+       ret = mtd->read(mtd, cxt->nextpage * OOPS_PAGE_SIZE, 4,
+                       &retlen, (u_char *) &count);
+       if ((retlen != 4) || (ret < 0)) {
+               printk(KERN_ERR "mtdoops: Read failure at %d (%d of 4 read)"
+                               ", err %d.\n", cxt->nextpage * OOPS_PAGE_SIZE,
+                               retlen, ret);
+               return 1;
+       }
+
+       /* See if we need to erase the next block */
+       if (count != 0xffffffff)
+               return 1;
+
+       printk(KERN_DEBUG "mtdoops: Ready %d, %d (no erase)\n",
+                       cxt->nextpage, cxt->nextcount);
+       cxt->ready = 1;
+       return 0;
+}
+
+static void mtdoops_prepare(struct mtdoops_context *cxt)
+{
+       struct mtd_info *mtd = cxt->mtd;
+       int i = 0, j, ret, mod;
+
+       /* We were unregistered */
+       if (!mtd)
+               return;
+
+       mod = (cxt->nextpage * OOPS_PAGE_SIZE) % mtd->erasesize;
+       if (mod != 0) {
+               cxt->nextpage = cxt->nextpage + ((mtd->erasesize - mod) / OOPS_PAGE_SIZE);
+               if (cxt->nextpage > cxt->oops_pages)
+                       cxt->nextpage = 0;
+       }
+
+       while (mtd->block_isbad &&
+                       mtd->block_isbad(mtd, cxt->nextpage * OOPS_PAGE_SIZE)) {
+badblock:
+               printk(KERN_WARNING "mtdoops: Bad block at %08x\n",
+                               cxt->nextpage * OOPS_PAGE_SIZE);
+               i++;
+               cxt->nextpage = cxt->nextpage + (mtd->erasesize / OOPS_PAGE_SIZE);
+               if (cxt->nextpage > cxt->oops_pages)
+                       cxt->nextpage = 0;
+               if (i == (cxt->oops_pages / (mtd->erasesize / OOPS_PAGE_SIZE))) {
+                       printk(KERN_ERR "mtdoops: All blocks bad!\n");
+                       return;
+               }
+       }
+
+       for (j = 0, ret = -1; (j < 3) && (ret < 0); j++)
+               ret = mtdoops_erase_block(mtd, cxt->nextpage * OOPS_PAGE_SIZE);
+
+       if (ret < 0) {
+               if (mtd->block_markbad)
+                       mtd->block_markbad(mtd, cxt->nextpage * OOPS_PAGE_SIZE);
+               goto badblock;
+       }
+
+       printk(KERN_DEBUG "mtdoops: Ready %d, %d \n", cxt->nextpage, cxt->nextcount);
+
+       cxt->ready = 1;
+}
+
+static void mtdoops_workfunc(struct work_struct *work)
+{
+       struct mtdoops_context *cxt =
+                       container_of(work, struct mtdoops_context, work);
+
+       mtdoops_prepare(cxt);
+}
+
+static int find_next_position(struct mtdoops_context *cxt)
+{
+       struct mtd_info *mtd = cxt->mtd;
+       int page, maxpos = 0;
+       u32 count, maxcount = 0xffffffff;
+       size_t retlen;
+
+       for (page = 0; page < cxt->oops_pages; page++) {
+               mtd->read(mtd, page * OOPS_PAGE_SIZE, 4, &retlen, (u_char *) &count);
+               if (count == 0xffffffff)
+                       continue;
+               if (maxcount == 0xffffffff) {
+                       maxcount = count;
+                       maxpos = page;
+               } else if ((count < 0x40000000) && (maxcount > 0xc0000000)) {
+                       maxcount = count;
+                       maxpos = page;
+               } else if ((count > maxcount) && (count < 0xc0000000)) {
+                       maxcount = count;
+                       maxpos = page;
+               } else if ((count > maxcount) && (count > 0xc0000000)
+                                       && (maxcount > 0x80000000)) {
+                       maxcount = count;
+                       maxpos = page;
+               }
+       }
+       if (maxcount == 0xffffffff) {
+               cxt->nextpage = 0;
+               cxt->nextcount = 1;
+               cxt->ready = 1;
+               printk(KERN_DEBUG "mtdoops: Ready %d, %d (first init)\n",
+                               cxt->nextpage, cxt->nextcount);
+               return 0;
+       }
+
+       cxt->nextpage = maxpos;
+       cxt->nextcount = maxcount;
+
+       return mtdoops_inc_counter(cxt);
+}
+
+
+static void mtdoops_notify_add(struct mtd_info *mtd)
+{
+       struct mtdoops_context *cxt = &oops_cxt;
+       int ret;
+
+       if ((mtd->index != cxt->mtd_index) || cxt->mtd_index < 0)
+               return;
+
+       if (mtd->size < (mtd->erasesize * 2)) {
+               printk(KERN_ERR "MTD partition %d not big enough for mtdoops\n",
+                               mtd->index);
+               return;
+       }
+
+       cxt->mtd = mtd;
+       cxt->oops_pages = mtd->size / OOPS_PAGE_SIZE;
+
+       ret = find_next_position(cxt);
+       if (ret == 1)
+               mtdoops_prepare(cxt);
+
+       printk(KERN_DEBUG "mtdoops: Attached to MTD device %d\n", mtd->index);
+}
+
+static void mtdoops_notify_remove(struct mtd_info *mtd)
+{
+       struct mtdoops_context *cxt = &oops_cxt;
+
+       if ((mtd->index != cxt->mtd_index) || cxt->mtd_index < 0)
+               return;
+
+       cxt->mtd = NULL;
+       flush_scheduled_work();
+}
+
+
+static void
+mtdoops_console_write(struct console *co, const char *s, unsigned int count)
+{
+       struct mtdoops_context *cxt = co->data;
+       struct mtd_info *mtd = cxt->mtd;
+       int i, ret;
+
+       if (!cxt->ready || !mtd)
+               return;
+
+       if (!oops_in_progress && cxt->writecount != 0) {
+               size_t retlen;
+               if (cxt->writecount < OOPS_PAGE_SIZE)
+                       memset(cxt->oops_buf + cxt->writecount, 0xff,
+                                       OOPS_PAGE_SIZE - cxt->writecount);
+
+               ret = mtd->write(mtd, cxt->nextpage * OOPS_PAGE_SIZE,
+                                       OOPS_PAGE_SIZE, &retlen, cxt->oops_buf);
+               cxt->ready = 0;
+               cxt->writecount = 0;
+
+               if ((retlen != OOPS_PAGE_SIZE) || (ret < 0))
+                       printk(KERN_ERR "mtdoops: Write failure at %d (%d of %d"
+                               " written), err %d.\n",
+                               cxt->nextpage * OOPS_PAGE_SIZE, retlen,
+                               OOPS_PAGE_SIZE, ret);
+
+               ret = mtdoops_inc_counter(cxt);
+               if (ret == 1)
+                       schedule_work(&cxt->work);
+       }
+
+       if (!oops_in_progress)
+               return;
+
+       if (cxt->writecount == 0) {
+               u32 *stamp = cxt->oops_buf;
+               *stamp = cxt->nextcount;
+               cxt->writecount = 4;
+       }
+
+       if ((count + cxt->writecount) > OOPS_PAGE_SIZE)
+               count = OOPS_PAGE_SIZE - cxt->writecount;
+
+       for (i = 0; i < count; i++, s++)
+               *((char *)(cxt->oops_buf) + cxt->writecount + i) = *s;
+
+       cxt->writecount = cxt->writecount + count;
+}
+
+static int __init mtdoops_console_setup(struct console *co, char *options)
+{
+       struct mtdoops_context *cxt = co->data;
+
+       if (cxt->mtd_index != -1)
+               return -EBUSY;
+       if (co->index == -1)
+               return -EINVAL;
+
+       cxt->mtd_index = co->index;
+       return 0;
+}
+
+static struct mtd_notifier mtdoops_notifier = {
+       .add    = mtdoops_notify_add,
+       .remove = mtdoops_notify_remove,
+};
+
+static struct console mtdoops_console = {
+       .name           = "ttyMTD",
+       .write          = mtdoops_console_write,
+       .setup          = mtdoops_console_setup,
+       .flags          = CON_PRINTBUFFER,
+       .index          = -1,
+       .data           = &oops_cxt,
+};
+
+static int __init mtdoops_console_init(void)
+{
+       struct mtdoops_context *cxt = &oops_cxt;
+
+       cxt->mtd_index = -1;
+       cxt->oops_buf = vmalloc(OOPS_PAGE_SIZE);
+
+       if (!cxt->oops_buf) {
+               printk(KERN_ERR "Failed to allocate oops buffer workspace\n");
+               return -ENOMEM;
+       }
+
+       INIT_WORK(&cxt->work, mtdoops_workfunc);
+
+       register_console(&mtdoops_console);
+       register_mtd_user(&mtdoops_notifier);
+       return 0;
+}
+
+static void __exit mtdoops_console_exit(void)
+{
+       struct mtdoops_context *cxt = &oops_cxt;
+
+       unregister_mtd_user(&mtdoops_notifier);
+       unregister_console(&mtdoops_console);
+       vfree(cxt->oops_buf);
+}
+
+
+subsys_initcall(mtdoops_console_init);
+module_exit(mtdoops_console_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Richard Purdie <rpurdie@openedhand.com>");
+MODULE_DESCRIPTION("MTD Oops/Panic console logger/driver");