lkml.org 
[lkml]   [2002]   [Feb]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectLDT_ENTRIES in ldt.h: why 8192?
I'm getting "ldt allocation failures" on one of my systems.  I know of Manfred
Spraul's patch for this for 2.4.17. I'm using 2.4.12, though, and was
wondering about an alternate solution. What would happen if LDT_ENTRIES was
reduced, to, say, 4096, or 512, instead of 8192?

.../include/asm-i386/ldt.h specifies:

/* Maximum number of LDT entries supported. */
#define LDT_ENTRIES 8192

This was set to 8192 in version 0.99.15, in December, 1993, according to what I
could find by retrieving different revisions of the kernel.

I instrumented the kernel by adding a printk to the write_ldt subroutine.
Here's what I did, in the form of a patch:

diff -Naur linux-2.4.12.orig/arch/i386/kernel/ldt.c linux-2.4.12/arch/i386/kernel/ldt.c
--- linux-2.4.12.orig/arch/i386/kernel/ldt.c Thu Jul 26 13:29:45 2001
+++ linux-2.4.12/arch/i386/kernel/ldt.c Fri Feb 1 16:50:19 2002
@@ -59,6 +59,8 @@
if (copy_from_user(&ldt_info, ptr, sizeof(ldt_info)))
goto out;

+ printk(KERN_DEBUG "write_ldt: ldt_info.entry_number: %d\n", ldt_info.entry_number);
+
error = -EINVAL;
if (ldt_info.entry_number >= LDT_ENTRIES)
goto out;
---end of patch----
Yes, this slows down the system, and it fills up /var. And I would never think
about doing this on a production system. And there's probably a more efficient
way to figure out the "high water mark" of LDT entry usage using /proc, or a
static variable.

In any case, the highest entry we got was 4. We weren't using WINE.
It seems somewhat wasteful to have many processes allocating 64 kbyte
of memory for LDT's until no more processes can run, when each process
uses only 4*8=32 bytes of the 64 kbyte LDT.

I tried to find the answer to this in the lkml archives. I couldn't
find an lkml archive going back to 1993. uwsg's only goes back to 1995,
and the other lkml archives don't appear to go back even that far.

Thanks,

--Seth Alford
setha@plaza.ds.adp.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:24    [W:0.085 / U:39.252 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site