I've been searching everywhere finding a documentation that explains when to use each flags of kmalloc. I found this reference which fairly explains when to use some of the flags but I can't find the other flags like GFP_HIGHUSER_PAGECACHE, GFP_HIGHUSER_MOVABLE and other flags in gfp.h header file... Can somebody tell me when to use these other flags?
You want section Get Free Page (GFP) Flags of Understanding the Linux Virtual Memory Manager (pdf). It is for the 2.4 kernel but it should still apply.
original post:
From the 2.6.32 gfp.h:
/*
* Action modifiers - doesn't change the zoning
*
* __GFP_REPEAT: Try hard to allocate the memory, but the allocation attempt
* _might_ fail. This depends upon the particular VM implementation.
*
* __GFP_NOFAIL: The VM implementation _must_ retry infinitely: the caller
* cannot handle allocation failures.
*
* __GFP_NORETRY: The VM implementation must not retry indefinitely.
*
* __GFP_MOVABLE: Flag that this page will be movable by the page migration
* mechanism or reclaimed
*/
#define __GFP_WAIT ((__force gfp_t)0x10u) /* Can wait and reschedule? */
#define __GFP_ ((__force gfp_t)0x20u) /* Should access emergency pools? */
#define __GFP_IO ((__force gfp_t)0x40u) /* Can start physical IO? */
#define __GFP_FS ((__force gfp_t)0x80u) /* Can call down to low-level FS? */
#define __GFP_COLD ((__force gfp_t)0x100u) /* Cache-cold page required */
#define __GFP_NOWARN ((__force gfp_t)0x200u) /* Suppress page allocation failure warning */
#define __GFP_REPEAT ((__force gfp_t)0x400u) /* See above */
#define __GFP_NOFAIL ((__force gfp_t)0x800u) /* See above */
#define __GFP_NORETRY ((__force gfp_t)0x1000u)/* See above */
#define __GFP_COMP ((__force gfp_t)0x4000u)/* Add compound page metadata */
#define __GFP_ZERO ((__force gfp_t)0x8000u)/* Return zeroed page on success */
#define __GFP_NOMEMALLOC ((__force gfp_t)0x10000u) /* Don't use emergency reserves */
#define __GFP_HARDWALL ((__force gfp_t)0x20000u) /* Enforce hardwall cpuset memory allocs */
#define __GFP_THISNODE ((__force gfp_t)0x40000u)/* No fallback, no policies */
#define __GFP_RECLAIMABLE ((__force gfp_t)0x80000u) /* Page is reclaimable */
Also, from kernel archives:
To make code easier to read, a set of three GFP flags are added called GFP_PAGECACHE, GFP_NOFS_PAGECACHE and GFP_HIGHUSER_PAGECACHE.
Looking at the diff source on that page shows the flags you were asking about are a combination of existing flags:
#define GFP_HIGHUSER_MOVABLE (__GFP_WAIT | __GFP_IO | __GFP_FS | \
__GFP_HARDWALL | __GFP_HIGHMEM | \
__GFP_MOVABLE)
+#define GFP_NOFS_PAGECACHE (__GFP_WAIT | __GFP_IO | __GFP_MOVABLE)
+#define GFP_USER_PAGECACHE (__GFP_WAIT | __GFP_IO | __GFP_FS | \
+ __GFP_HARDWALL | __GFP_MOVABLE)
+#define GFP_HIGHUSER_PAGECACHE (__GFP_WAIT | __GFP_IO | __GFP_FS | \
+ __GFP_HARDWALL | __GFP_HIGHMEM | \
+ __GFP_MOVABLE)
#ifdef CONFIG_NUMA
#define GFP_THISNODE (__GFP_THISNODE | __GFP_NOWARN | __GFP_NORETRY)