Skip to contentSkip to conversation

Final removal of FASTCALL()/fastcall

PATCH In review Harvey Harrison 3 participants 5 messages Feb 11, 2008 – Feb 12, 2008

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 3/3 Final removal of FASTCALL()/fastcall

All users are gone, remove definitions and comments referring to them.

---
 drivers/char/drm/i830_dma.c   |    2 +-
 include/asm-mn10300/highmem.h |    4 ++--
 include/asm-mn10300/linkage.h |    2 --
 include/linux/irq.h           |    1 -
 include/linux/linkage.h       |    5 -----
 kernel/rcupdate.c             |    2 +-
 6 files changed, 4 insertions(+), 12 deletions(-)
drivers/char/drm/i830_dma.c +1 -1
index 379cbda..9df0810 100644
@@ -36,7 +36,7 @@
 #include "i830_drm.h"
 #include "i830_drv.h"
 #include <linux/interrupt.h>	/* For task queue support */
-#include <linux/pagemap.h>	/* For FASTCALL on unlock_page() */
+#include <linux/pagemap.h>
 #include <linux/delay.h>
 #include <asm/uaccess.h>
 
include/asm-mn10300/highmem.h +2 -2
index 383c0c4..5256854 100644
@@ -42,8 +42,8 @@ extern void __init kmap_init(void);
 #define PKMAP_NR(virt)  ((virt - PKMAP_BASE) >> PAGE_SHIFT)
 #define PKMAP_ADDR(nr)  (PKMAP_BASE + ((nr) << PAGE_SHIFT))
 
-extern unsigned long __fastcall kmap_high(struct page *page);
-extern void __fastcall kunmap_high(struct page *page);
+extern unsigned long kmap_high(struct page *page);
+extern void kunmap_high(struct page *page);
 
 static inline unsigned long kmap(struct page *page)
 {
include/asm-mn10300/linkage.h +0 -2
index 29a32e4..dda3002 100644
@@ -13,8 +13,6 @@
 
 /* don't override anything */
 #define asmlinkage
-#define FASTCALL(x) x
-#define fastcall
 
 #define __ALIGN		.align 4,0xcb
 #define __ALIGN_STR	".align 4,0xcb"
include/linux/irq.h +0 -1
index bfd9efb..176e5e7 100644
@@ -285,7 +285,6 @@ extern void handle_bad_irq(unsigned int irq, struct irq_desc *desc);
 
 /*
  * Monolithic do_IRQ implementation.
- * (is an explicit fastcall, because i386 4KSTACKS calls it from assembly)
  */
 #ifndef CONFIG_GENERIC_HARDIRQS_NO__DO_IRQ
 extern unsigned int __do_IRQ(unsigned int irq);
include/linux/linkage.h +0 -5
index 3faf599..0592936 100644
@@ -73,9 +73,4 @@
 #define ATTRIB_NORET  __attribute__((noreturn))
 #define NORET_AND     noreturn,
 
-#ifndef FASTCALL
-#define FASTCALL(x)	x
-#define fastcall
-#endif
-
 #endif
kernel/rcupdate.c +1 -1
index 760dfc2..0810e78 100644
@@ -56,7 +56,7 @@ static atomic_t rcu_barrier_cpu_count;
 static DEFINE_MUTEX(rcu_barrier_mutex);
 static struct completion rcu_barrier_completion;
 
-/* Because of FASTCALL declaration of complete, we use this wrapper */
+/* FASTCALL no longer exists, this wrapper may no longer be needed */
 static void wakeme_after_rcu(struct rcu_head  *head)
 {
 	struct rcu_synchronize *rcu;
Signature · 1 line
1.5.4.1219.g65b9
#1 lore raw
Quoted 2 lines · On Mon, 11 Feb 2008 10:38:42 -0800 Harvey Harrison <[email protected]>…

All users are gone, remove definitions and comments referring to them.

I'm still showing occurrences in:

./Documentation/RCU/NMI-RCU.txt
./Documentation/kprobes.txt
./Documentation/uml/UserModeLinux-HOWTO.txt
./kernel/rcupdate.c

The last one is interesting:

/* FASTCALL no longer exists, this wrapper may no longer be needed */ static void wakeme_after_rcu(struct rcu_head *head)

{
        struct rcu_synchronize *rcu;

        rcu = container_of(head, struct rcu_synchronize, head);
        complete(&rcu->completion);
}

The comment is rather ungrammatical/meaningless. Perhaps it is trying to tell us that we can remove this function now?

#2 lore raw
Quoted 24 lines · Andrew Morton
Quoted 2 lines · On Mon, 11 Feb 2008 10:38:42 -0800 Harvey Harrison <[email protected]>…

All users are gone, remove definitions and comments referring to them.

I'm still showing occurrences in:

./Documentation/RCU/NMI-RCU.txt
./Documentation/kprobes.txt
./Documentation/uml/UserModeLinux-HOWTO.txt
./kernel/rcupdate.c

The last one is interesting:

/* FASTCALL no longer exists, this wrapper may no longer be needed */ static void wakeme_after_rcu(struct rcu_head *head)

{
        struct rcu_synchronize *rcu;

        rcu = container_of(head, struct rcu_synchronize, head);
        complete(&rcu->completion);
}
Quoted 2 lines

The comment is rather ungrammatical/meaningless. Perhaps it is trying to tell us that we can remove this function now?

Sorry, I added that as a note in my patch to look into and ended up forgetting about it.

The original comment that was there read:
/* Because of FASTCALL declaration of complete, we use this wrapper */

Sorry about that, I'll look into the last occurances and send you a patch 4/3.

Harvey

#3 lore raw

[PATCH 4/3] Documentation: Remove references to FASTCALL/fastcall

In reply to Andrew Morton

An example function prototype and description of options that modify how arguments are passed had references to fastcall.

---
Andrew, I don't have the knowledge to deal with the rcupdate.c wrapper function, hopefully Paul will be able to have a look.

Also, the UML reference is to a config option UM_FASTCALL, not the FASTCALL macro.

 Documentation/RCU/NMI-RCU.txt |    2 +-
 Documentation/kprobes.txt     |   11 +++++------
 2 files changed, 6 insertions(+), 7 deletions(-)
Documentation/RCU/NMI-RCU.txt +1 -1
index d0634a5..c64158e 100644
@@ -25,7 +25,7 @@ the NMI handler to take the default machine-specific action.
 This nmi_callback variable is a global function pointer to the current
 NMI handler.
 
-	fastcall void do_nmi(struct pt_regs * regs, long error_code)
+	void do_nmi(struct pt_regs * regs, long error_code)
 	{
 		int cpu;
 
Documentation/kprobes.txt +5 -6
index 30c1017..83f515c 100644
@@ -92,9 +92,8 @@ handler has run.  Up to MAX_STACK_SIZE bytes are copied -- e.g.,
 64 bytes on i386.
 
 Note that the probed function's args may be passed on the stack
-or in registers (e.g., for x86_64 or for an i386 fastcall function).
-The jprobe will work in either case, so long as the handler's
-prototype matches that of the probed function.
+or in registers.  The jprobe will work in either case, so long as the
+handler's prototype matches that of the probed function.
 
 1.3 Return Probes
 
@@ -270,9 +269,9 @@ Kprobes runs the handler whose address is jp->entry.
 The handler should have the same arg list and return type as the probed
 function; and just before it returns, it must call jprobe_return().
 (The handler never actually returns, since jprobe_return() returns
-control to Kprobes.)  If the probed function is declared asmlinkage,
-fastcall, or anything else that affects how args are passed, the
-handler's declaration must match.
+control to Kprobes.)  If the probed function is declared asmlinkage
+or anything else that affects how args are passed, the handler's
+declaration must match.
 
 register_jprobe() returns 0 on success, or a negative errno otherwise.
 
Signature · 1 line
1.5.4.1219.g65b9
#4 lore raw
In reply to Andrew Morton
Quoted 24 lines · Andrew Morton
Quoted 2 lines · On Mon, 11 Feb 2008 10:38:42 -0800 Harvey Harrison <[email protected]>…

All users are gone, remove definitions and comments referring to them.

I'm still showing occurrences in:

./Documentation/RCU/NMI-RCU.txt
./Documentation/kprobes.txt
./Documentation/uml/UserModeLinux-HOWTO.txt
./kernel/rcupdate.c

The last one is interesting:

/* FASTCALL no longer exists, this wrapper may no longer be needed */ static void wakeme_after_rcu(struct rcu_head *head)

{
        struct rcu_synchronize *rcu;

        rcu = container_of(head, struct rcu_synchronize, head);
        complete(&rcu->completion);
}
Quoted 2 lines

The comment is rather ungrammatical/meaningless. Perhaps it is trying to tell us that we can remove this function now?

Hmmm... A better comment might be something like "Awaken the corresponding synchronize_rcu() instance now that a grace period has completed". Given that the underlying RCU implementation provides callbacks, synchronize_rcu() will need some sort of callback function.

So, the question is whether "complete()" can be that callback function. As near as I can tell, the answer is "no", because the RCU callback function always gets a pointer to a struct rcu_head, which would not make much sense to the "complete()" function.

In short, I believe that wakeme_after_rcu() needs to stick around. A better comment for wakeme_after_rcu() would be good, perhaps as shown below.

Thanx, Paul

---

 rcupdate.c |    5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/rcupdate.c +4 -1
diff -urpNa -X dontdiff linux-2.6.24/kernel/rcupdate.c linux-2.6.24-wakeme-comment/kernel/rcupdate.c
@@ -601,7 +601,10 @@ struct rcu_synchronize {
 	struct completion completion;
 };
 
-/* Because of FASTCALL declaration of complete, we use this wrapper */
+/*
+ * Awaken the corresponding synchronize_rcu() instance now that a
+ * grace period has elapsed.
+ */
 static void wakeme_after_rcu(struct rcu_head  *head)
 {
 	struct rcu_synchronize *rcu;