Skip to contentSkip to conversation

reboot on out-of-file handles

PATCH In review Dr. David Alan Gilbert 2 participants 4 messages Aug 31, 2002

Subsystems (3)
Documentation VFS Proc sysctl
lore
Share & tools
lore

Dr. David Alan Gilbert

author
#0 lore raw

PATCH 1/1 reboot on out-of-file handles

Hi,
Please find below a patch that adds the ability to panic if you run
out of file handles (by setting /proc/sys/fs/file-max-panic to none-0). When combined with reboot-on-panic this means that a server might be able to get out of a service-gone-mad situation. It calls show_state before panic'ing to log what was going on - adding something similar which listed open filehandles would probably be advantageous.

The patch is against a clean 2.4.19 and was tested on sparc64.

Comments?

Dave

Documentation/sysctl/fs.txt +6 -1
diff -urN linux-2.4.19.orig/Documentation/sysctl/fs.txt linux-2.4.19/Documentation/sysctl/fs.txt
@@ -19,6 +19,7 @@
 - dquot-max
 - dquot-nr
 - file-max
+- file-max-panic
 - file-nr
 - inode-max
 - inode-nr

@@ -71,7 +72,7 @@
 
 ==============================================================
 
-file-max & file-nr:
+file-max, file-nr & file-max-panic:
 
 The kernel allocates file handles dynamically, but as yet it
 doesn't free them again.

@@ -88,6 +89,10 @@
 far behind, you've encountered a peak in your usage of file
 handles and you don't need to increase the maximum.
 
+If file-max-panic is none zero, then when your system runs out
+of file handles it will panic.  While this may seem extreme, when
+combined with rebooting on panic it can be better to have a machine
+reboot itself than sit there unable to open any files.
 ==============================================================
 
 inode-max, inode-nr & inode-state:
fs/file_table.c +12 -0
diff -urN linux-2.4.19.orig/fs/file_table.c linux-2.4.19/fs/file_table.c
@@ -5,6 +5,7 @@
  *  Copyright (C) 1997 David S. Miller ([email protected])
  */
 
+#include <linux/sched.h>
 #include <linux/string.h>
 #include <linux/slab.h>
 #include <linux/file.h>

@@ -15,6 +16,8 @@
 
 /* sysctl tunables... */
 struct files_stat_struct files_stat = {0, 0, NR_FILE};
+/* sysctl flag - if !=0 we panic when we run out of file handles */
+int files_panic_nofh=0;
 
 /* Here the new files go */
 static LIST_HEAD(anon_list);

@@ -72,6 +75,15 @@
 	} else if (files_stat.max_files > old_max) {
 		printk(KERN_INFO "VFS: file-max limit %d reached\n", files_stat.max_files);
 		old_max = files_stat.max_files;
+                if (files_panic_nofh)
+		{
+	          /* DAG: chances are that the kernel is still OK it is just a
+		   * user task that has gone mad; so lets dump the state of
+		   * all processes so the victim can find what caused it
+		   */
+	          show_state();
+                  panic("VFS: Out of file handles (files_panic_nofh!=0)\n");
+		}
 	}
 	file_list_unlock();
 	return NULL;
include/linux/fs.h +2 -0
diff -urN linux-2.4.19.orig/include/linux/fs.h linux-2.4.19/include/linux/fs.h
@@ -53,6 +53,8 @@
 	int max_files;		/* tunable */
 };
 extern struct files_stat_struct files_stat;
+extern int files_panic_nofh;
+
 
 struct inodes_stat_t {
 	int nr_inodes;
include/linux/sysctl.h +1 -0
diff -urN linux-2.4.19.orig/include/linux/sysctl.h linux-2.4.19/include/linux/sysctl.h
@@ -546,6 +546,7 @@
 	FS_LEASES=13,	/* int: leases enabled */
 	FS_DIR_NOTIFY=14,	/* int: directory notification enabled */
 	FS_LEASE_TIME=15,	/* int: maximum time to wait for a lease break */
+        FS_PANICNOFH=16,        /* int: !=0 means panic when run out of fh */
 };
 
 /* CTL_DEBUG names: */
kernel/sysctl.c +2 -0
diff -urN linux-2.4.19.orig/kernel/sysctl.c linux-2.4.19/kernel/sysctl.c
@@ -309,6 +309,8 @@
 	 sizeof(int), 0644, NULL, &proc_dointvec},
 	{FS_LEASE_TIME, "lease-break-time", &lease_break_time, sizeof(int),
 	 0644, NULL, &proc_dointvec},
+	{FS_PANICNOFH, "file-max-panic", &files_panic_nofh, sizeof(int),
+	 0644, NULL, &proc_dointvec},
 	{0}
 };

--
---------------- Have a happy GNU millennium! ----------------------

/ Dr. David Alan Gilbert    | Running GNU/Linux on Alpha,68K| Happy  \ 
\ gro.gilbert @ treblig.org | MIPS,x86,ARM, SPARC and HP-PA | In Hex /
 \ _________________________|_____ http://www.treblig.org   |_______/
#1 lore raw
Quoted 3 lines · Dr. David Alan Gilbert

Hi,
Please find below a patch that adds the ability to panic if you run
out of file handles (by setting /proc/sys/fs/file-max-panic to none-0).

You can already do this reliably in user space as part of your watchdog daemon processing.

author
#2 lore raw
Quoted 7 lines · * Alan Cox ([email protected]) wrote:
Quoted 3 lines · Dr. David Alan Gilbert

Hi,
Please find below a patch that adds the ability to panic if you run
out of file handles (by setting /proc/sys/fs/file-max-panic to none-0).

You can already do this reliably in user space as part of your watchdog daemon processing.

Ah - good idea.
The only thing that doesn't do (which my patch doesn't do much of), is log the state of the system at failure - I'd really like to know what ate all the filehandles.

Dave
---------------- Have a happy GNU millennium! ----------------------

/ Dr. David Alan Gilbert    | Running GNU/Linux on Alpha,68K| Happy  \ 
\ gro.gilbert @ treblig.org | MIPS,x86,ARM, SPARC and HP-PA | In Hex /
 \ _________________________|_____ http://www.treblig.org   |_______/
#3 lore raw
Quoted 7 lines · Dr. David Alan Gilbert
Quoted 2 lines

You can already do this reliably in user space as part of your watchdog daemon processing.

Ah - good idea.
The only thing that doesn't do (which my patch doesn't do much of), is log the state of the system at failure - I'd really like to know what ate all the filehandles.

If your daemon keeps a few open handles to reuse and the log file it can maybe do that when it spots the problem occurs and isnt bumping the watchdog