2.4.0 i386: Why free_xxx_slow() instead free_xxx_fast()?

From: Ulrich Kunitz (gefm21@uumail.de)
Date: Sat Jan 20 2001 - 10:33:41 EST

Hello,

in include/asm-i386/pgalloc.h the {pgd,pmd,pte}_free* functions are
defined as the functions free_{pgd,pmd,pte}_slow (see the patch below).
IMO this leaves the quick lists pretty useless for i386.
Have I overlooked something?

There is a linux-kernel mail from Martin Schwidefsky
(schwidefsky@de.ibm.com) about 2.4.0-test10 hinting on bad_pmd_table()
and bad_pgd_table() landing in the quick lists. Using the _slow
functions would fix any problem with the quick lists for now.

Changing the defines as in the patch seems to work here on a
two uniprocessor machines (Pentium III, Athlon). I've checked with
2.4.1-pre9 too. But what are the risks?

Ciao,

Uli Kunitz

--- linux-2.4.0/include/asm-i386/pgalloc.h Tue Jan 9 15:00:08 2001
+++ linux/include/asm-i386/pgalloc.h Thu Jan 11 17:30:55 2001
@@ -92,9 +92,9 @@
         free_page((unsigned long)pte);
 }
 
-#define pte_free_kernel(pte) free_pte_slow(pte)
-#define pte_free(pte) free_pte_slow(pte)
-#define pgd_free(pgd) free_pgd_slow(pgd)
+#define pte_free_kernel(pte) free_pte_fast(pte)
+#define pte_free(pte) free_pte_fast(pte)
+#define pgd_free(pgd) free_pgd_fast(pgd)
 #define pgd_alloc() get_pgd_fast()
 
 extern inline pte_t * pte_alloc_kernel(pmd_t * pmd, unsigned long address)
@@ -145,7 +145,7 @@
  * inside the pgd, so has no extra memory associated with it.
  * (In the PAE case we free the page.)
  */
-#define pmd_free(pmd) free_pmd_slow(pmd)
+#define pmd_free(pmd) free_pmd_fast(pmd)
 
 #define pmd_free_kernel pmd_free
 #define pmd_alloc_kernel pmd_alloc 

-- 
Ulrich Kunitz (gefm21@uumail.de)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

This archive was generated by hypermail 2b29 : Tue Jan 23 2001 - 21:00:22 EST