Skip to contentSkip to conversation

m68k: Add infrastructure for machine-specific get_cycles()

PATCH Superseded Geert Uytterhoeven 2 participants 4 messages Sep 9, 2013 m68k architecture

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Geert Uytterhoeven

author maintainer
#0 lore raw

PATCH 1/2 m68k: Add infrastructure for machine-specific get_cycles()

Currently get_cycles() always returns zero. As get_cycles() is called to add entropy to the random number generator pool, security can be increased by adding a hook where platforms can provide their own implementation.

---
 arch/m68k/include/asm/machdep.h |    1 +
 arch/m68k/include/asm/timex.h   |    5 +----
 arch/m68k/kernel/time.c         |   13 +++++++++++++
 3 files changed, 15 insertions(+), 4 deletions(-)
arch/m68k/include/asm/machdep.h +1 -0
index 953ca21..334f751 100644
@@ -22,6 +22,7 @@ extern unsigned int (*mach_get_ss)(void);
 extern int (*mach_get_rtc_pll)(struct rtc_pll_info *);
 extern int (*mach_set_rtc_pll)(struct rtc_pll_info *);
 extern int (*mach_set_clock_mmss)(unsigned long);
+extern cycles_t (*mach_get_cycles)(void);
 extern void (*mach_reset)( void );
 extern void (*mach_halt)( void );
 extern void (*mach_power_off)( void );
arch/m68k/include/asm/timex.h +1 -4
index 6759dad..0fdc5d2 100644
@@ -23,9 +23,6 @@
 
 typedef unsigned long cycles_t;
 
-static inline cycles_t get_cycles(void)
-{
-	return 0;
-}
+extern cycles_t get_cycles(void);
 
 #endif
arch/m68k/kernel/time.c +13 -0
index 7eb9792..65b1663 100644
@@ -28,6 +28,19 @@
 #include <linux/timex.h>
 #include <linux/profile.h>
 
+
+cycles_t (*mach_get_cycles)(void);
+
+cycles_t get_cycles(void)
+{
+	if (mach_get_cycles)
+		return mach_get_cycles();
+
+	return 0;
+}
+EXPORT_SYMBOL(get_cycles);
+
+
 /*
  * timer_interrupt() needs to keep up the real-time clock,
  * as well as call the "xtime_update()" routine every clocktick
Signature · 1 line
1.7.9.5
author maintainer
#1 lore raw

PATCH 2/2 m68k/amiga: Provide mach_get_cycles()

Use the 24-bit Time-Of-Day clock in CIA B, which is clocked by HSYNC.

---
1. Completely untested! It does compile ;-)

2. Is a 24-bit counter running at 15-31 kHz good enough?
Two cascaded 16-bit CIA timers running from the 700 kHz E-clock would be better, but as jiffies use CIA B Timer A and the floppy driver uses CIA A Timer B, we don't have two available timers in the same CIA without some code shuffling.

3. What are the semantics of get_cycles()?
Does it have to be a "nice" counter, or can there be some irregularities?
I.e. can I add one more line

x = (x << 8) | ciab.talo; // low byte of 700 kHz timer

to get 32 bits?

 arch/m68k/amiga/config.c |   16 ++++++++++++++++
 1 file changed, 16 insertions(+)
arch/m68k/amiga/config.c +16 -0
index acd9c16..914e056 100644
@@ -358,6 +358,20 @@ static void __init amiga_identify(void)
 #undef AMIGAHW_ANNOUNCE
 }
 
+
+static cycles_t amiga_get_cycles(void)
+{
+	cycles_t x;
+
+	/* CIA B 24-bit TOD is clocked by HSYNC */
+	x = ciab.todhi;
+	x = (x << 8) | ciab.todmid;
+	x = (x << 8) | ciab.todlo;
+
+	return x;
+}
+
+
     /*
      *  Setup the Amiga configuration info
      */
@@ -395,6 +409,8 @@ void __init config_amiga(void)
 	mach_heartbeat = amiga_heartbeat;
 #endif
 
+	mach_get_cycles = amiga_get_cycles;
+
 	/* Fill in the clock value (based on the 700 kHz E-Clock) */
 	amiga_colorclock = 5*amiga_eclock;	/* 3.5 MHz */
 
Signature · 1 line
1.7.9.5
author maintainer
#2 lore raw

PATCH 2/2 m68k/amiga: Provide mach_get_cycles()

In reply to Geert Uytterhoeven

Use the 24-bit Time-Of-Day clock in CIA B, which is clocked by HSYNC.

---
1. Completely untested! It does compile ;-)

2. Is a 24-bit counter running at 15-31 kHz good enough?
Two cascaded 16-bit CIA timers running from the 700 kHz E-clock would be better, but as jiffies use CIA B Timer A and the floppy driver uses CIA A Timer B, we don't have two available timers in the same CIA without some code shuffling.

3. What are the semantics of get_cycles()?
Does it have to be a "nice" counter, or can there be some irregularities?
I.e. can I add one more line

x = (x << 8) | ciab.talo; // low byte of 700 kHz timer

to get 32 bits?

 arch/m68k/amiga/config.c |   16 ++++++++++++++++
 1 file changed, 16 insertions(+)
arch/m68k/amiga/config.c +16 -0
index acd9c16..914e056 100644
@@ -358,6 +358,20 @@ static void __init amiga_identify(void)
 #undef AMIGAHW_ANNOUNCE
 }
 
+
+static cycles_t amiga_get_cycles(void)
+{
+	cycles_t x;
+
+	/* CIA B 24-bit TOD is clocked by HSYNC */
+	x = ciab.todhi;
+	x = (x << 8) | ciab.todmid;
+	x = (x << 8) | ciab.todlo;
+
+	return x;
+}
+
+
     /*
      *  Setup the Amiga configuration info
      */
@@ -395,6 +409,8 @@ void __init config_amiga(void)
 	mach_heartbeat = amiga_heartbeat;
 #endif
 
+	mach_get_cycles = amiga_get_cycles;
+
 	/* Fill in the clock value (based on the 700 kHz E-Clock) */
 	amiga_colorclock = 5*amiga_eclock;	/* 3.5 MHz */
 
Signature · 1 line
1.7.9.5
#3 lore raw
In reply to Geert Uytterhoeven

Geert Uytterhoeven dixit:

Quoted 2 lines

Currently get_cycles() always returns zero. As get_cycles() is called to add entropy to the random number generator pool, security can be increased

You rock for having something before MIPS ;-) (Note that other architectures like CRIS are also affected… just, MIPS probably has the broadest impact… even I have a FreeWRT WLAN AP.)

Sorry I can’t provide any feedback because I’m out of my depth on both the hardware and the requirements aspect.

bye,
//mirabilos

Signature · 4 lines
Sometimes they [people] care too much: pretty printers [and syntax highligh-
ting, d.A.] mechanically produce pretty output that accentuates irrelevant
detail in the program, which is as sensible as putting all the prepositions
in English text in bold font.	-- Rob Pike in "Notes on Programming in C"