powerpc: fix register usage in some inline assembly code
PATCH 1/1 powerpc: fix register usage in some inline assembly code
In some usages of inline assembly, hard-coded registers were specified when a scratch register should have been used instead.
---
I can't test the kgdb changes, but the rest appears to be okay.
arch/powerpc/lib/kgdb.c | 25 +++++++++++++++---------- arch/powerpc/lib/time.c | 5 ++++- 2 files changed, 19 insertions(+), 11 deletions(-)
arch/powerpc/lib/kgdb.c +15 -10
@@ -12,11 +12,13 @@ void breakinst(void);int kgdb_setjmp (long *buf) { - asm ("mflr 0; stw 0,0(%0);" -"stw 1,4(%0); stw 2,8(%0);" -"mfcr 0; stw 0,12(%0);" -"stmw 13,16(%0)" - : :"r" (buf)); +unsigned long temp; + + asmvolatile ("mflr %0; stw %0,0(%1);" +"stw %%r1,4(%1); stw %%r2,8(%1);" +"mfcr %0; stw %0,12(%1);" +"stmw %%r13,16(%1)" + :"=&r" (temp) :"r" (buf));/* XXX should save fp regs as well */ return 0 ; }
@@ -24,13 +26,16 @@ kgdb_setjmp(long *buf)void kgdb_longjmp (long *buf,int val) { +unsigned long temp; +if (val ==0 ) val =1 ; - asm ("lmw 13,16(%0);" -"lwz 0,12(%0); mtcrf 0x38,0;" -"lwz 0,0(%0); lwz 1,4(%0); lwz 2,8(%0);" -"mtlr 0; mr 3,%1" - : :"r" (buf),"r" (val)); + + asmvolatile ("lmw %%r13,16(%1);" +"lwz %0,12(%1); mtcrf 0x38,%0;" +"lwz %0,0(%1); lwz %%r1,4(%1); lwz %%r2,8(%1);" +"mtlr %0; mr %%r3,%2" + :"=&r" (temp) :"r" (buf),"r" (val)); } static inline unsigned long
arch/powerpc/lib/time.c +4 -1
@@ -78,6 +78,8 @@ unsigned long ticks2usec(unsigned long ticks)int init_timebase (void ) { +unsigned long temp; +#if defined(CONFIG_5xx) || defined(CONFIG_8xx)volatile immap_t *immap = (immap_t *)CONFIG_SYS_IMMR ;
@@ -86,7 +88,8 @@ int init_timebase (void)#endif /* reset */ - asm ("li 3,0 ; mttbu 3 ; mttbl 3 ;" ); + asmvolatile ("li %0,0 ; mttbu %0 ; mttbl %0;" + :"=&r" (temp) );#if defined(CONFIG_5xx) || defined(CONFIG_8xx)/* enable */
Signature · 1 line
1.7.2.3
PATCH 1/1 powerpc: fix register usage in some inline assembly code
Dear Timur Tabi,
Quoted 2 lines · In message <[email protected]> you wrote:
In some usages of inline assembly, hard-coded registers were specified when a scratch register should have been used instead.
Is this just a cosmetic improvement, or are you fixing any real bug? If so, what exactly is the problem?
Best regards,
Wolfgang Denk
Signature · 6 lines
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de Conceptual integrity in turn dictates that the design must proceed from one mind, or from a very small number of agreeing resonant minds. - Frederick Brooks Jr., "The Mythical Man Month"
PATCH 1/1 powerpc: fix register usage in some inline assembly code
Quoted 6 lines · Wolfgang Denk
Quoted 2 lines · In message<[email protected]> you wrote:
Quoted 2 lines
In some usages of inline assembly, hard-coded registers were specified when a scratch register should have been used instead.
Is this just a cosmetic improvement, or are you fixing any real bug? If so, what exactly is the problem?
Well, I'm not a expert on inline assembly, so I could be missing something, but IMHO it is fixing a real bug, although the current code does technically work.
The problem is that the code was modifying registers and gcc was unaware of that. So there is a possibility that gcc could have put something important in those registers. It appears that the authors of that code knew which registers were unused, and just made sure to use those.
Signature · 2 lines
Timur Tabi Linux kernel developer
PATCH 1/1 powerpc: fix register usage in some inline assembly code
Quoted 11 lines · On Dec 15, 2010, at 6:32 AM, Tabi Timur-B04825 wrote:
Quoted 6 lines · Wolfgang Denk
Quoted 2 lines · In message<[email protected]> you wrote:
Quoted 2 lines
In some usages of inline assembly, hard-coded registers were specified when a scratch register should have been used instead.
Is this just a cosmetic improvement, or are you fixing any real bug? If so, what exactly is the problem?
Well, I'm not a expert on inline assembly, so I could be missing something, but IMHO it is fixing a real bug, although the current code does technically work.
The problem is that the code was modifying registers and gcc was unaware of that. So there is a possibility that gcc could have put something important in those registers. It appears that the authors of that code knew which registers were unused, and just made sure to use those.
I doubt we disagree that the change is a good one, just trying to understand if it was because of an actual issue you where seeing or just because of some code review that you made it.
ie, is this needed in v2010.12 or goes into 'next'
- k
PATCH 1/1 powerpc: fix register usage in some inline assembly code
Quoted 3 lines · Kumar Gala
I doubt we disagree that the change is a good one, just trying to understand if it was because of an actual issue you where seeing or just because of some code review that you made it.
Just a code review.
Quoted 2 lines
ie, is this needed in v2010.12 or goes into 'next'
Definitely 'next'.
Signature · 2 lines
Timur Tabi Linux kernel developer at Freescale
Wolfgang Denk
AppliedPATCH 1/1 powerpc: fix register usage in some inline assembly code
In reply to Timur TabiDear Timur Tabi,
Quoted 11 lines · In message <[email protected]> you wrote:
In some usages of inline assembly, hard-coded registers were specified when a scratch register should have been used instead.
Signed-off-by Timur Tabi <[email protected]>---
I can't test the kgdb changes, but the rest appears to be okay.
arch/powerpc/lib/kgdb.c | 25 +++++++++++++++---------- arch/powerpc/lib/time.c | 5 ++++- 2 files changed, 19 insertions(+), 11 deletions(-)
Applied to "next", thanks.
Best regards,
Wolfgang Denk
Signature · 6 lines
DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de Every program has at least one bug and can be shortened by at least one instruction -- from which, by induction, one can deduce that every program can be reduced to one instruction which doesn't work.