| Message ID | 1335245671-20544-1-git-send-email-stefan.kristiansson@saunalahti.fi |
|---|---|
| State | Rejected |
| Delegated to: | Mike Frysinger |
| Headers | show |
On Tuesday 24 April 2012 01:34:31 Stefan Kristiansson wrote: > invalidating the memory area that is being executed > while icache is on is a bit adventurous, > make sure it's turned off during this operation sounds like a bug in the chip. you should be able to icache invalidate the active insn all you like. -mike
diff --git a/arch/openrisc/cpu/cache.c b/arch/openrisc/cpu/cache.c index 2a73a4f..fedfd89 100644 --- a/arch/openrisc/cpu/cache.c +++ b/arch/openrisc/cpu/cache.c @@ -44,11 +44,15 @@ void invalidate_dcache_range(unsigned long addr, unsigned long stop) static void invalidate_icache_range(unsigned long addr, unsigned long stop) { ulong block_size = (mfspr(SPR_ICCFGR) & SPR_ICCFGR_CBS) ? 32 : 16; + ulong ie = icache_status(); + icache_disable(); while (addr < stop) { mtspr(SPR_ICBIR, addr); addr += block_size; } + if (ie) + icache_enable(); } void flush_cache(unsigned long addr, unsigned long size)
invalidating the memory area that is being executed while icache is on is a bit adventurous, make sure it's turned off during this operation Signed-off-by: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi> --- arch/openrisc/cpu/cache.c | 4 ++++ 1 files changed, 4 insertions(+), 0 deletions(-)