On Thu, Mar 04, 2004 at 05:44:46PM -0800, Vineet Kumar wrote: > * ([EMAIL PROTECTED])[20040304 09:19]: > > dmesg just displays what the kernel outputs during boot. Usually you > > want to type "dmesg | less" so that you can scroll through all the > > messages. > > More precisely, kernel messages are sent to a buffer known > as the "kernel ring buffer". dmesg displays the contents of > this buffer. Many messages are recorded there during boot, > but it's not only boot messages that go there. In > particular, for a long-running (or long-winded) system, > dmesg may not show messages from boot, since more recent > messages may have pushed those old boot messages out of the > ring buffer.
I've been compiling everything from the cryptographic menu into my kernels (dunno whether I need all that or not), but since then, when it boots, the output from the cryptographic tests apparently overflow the standard 16K buffer, so that, even just after a boot, part of the dmesg output is cut off. I don't know the ultimate effect - it could make the system slower- but I've added this patch to my kernel sources: --------------- cut ----------------- diff -Naur -x .config linux-2.4.23-orig/kernel/printk.c linux-2.4.23/kernel/printk.c --- linux-2.4.23-orig/kernel/printk.c 2003-12-02 18:24:01.000000000 -0600 +++ linux-2.4.23/kernel/printk.c 2003-12-08 20:29:04.000000000 -0600 @@ -37,7 +37,7 @@ #elif defined(CONFIG_SMP) #define LOG_BUF_LEN (32768) #else -#define LOG_BUF_LEN (16384) /* This must be a power of two */ +#define LOG_BUF_LEN (32768) /*FIXED! This must be a power of two */ #endif #else /* CONFIG_LOG_BUF_SHIFT */ #define LOG_BUF_LEN (1 << CONFIG_LOG_BUF_SHIFT) ----------------- cut ---------------- It could be cleaner - I could have eliminated the elif condition and simply had a single define, but this works. Oh, and you have to use the "-s32768" option with dmesg to display the whole output. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
