Re: [PATCH] flush icache before set_pte on ia64 take3

2007-07-26 Thread Zoltan Menyhart
KAMEZAWA Hiroyuki wrote: Thank you for advise. Hmm..how about this ? == /* * synchronize icache and dcache if the hardware doesn't do it automatically * and the page is executable. */ static inline arch_sync_icache_dcache(pte_t pte, page) { } Why not? I just wanted to see if you

Re: [PATCH] flush icache before set_pte on ia64 take3

2007-07-26 Thread KAMEZAWA Hiroyuki
On Thu, 26 Jul 2007 17:13:52 +0200 Zoltan Menyhart <[EMAIL PROTECTED]> wrote: > I can agree: > - lazy_mmu_prot_update() is not the right place for a flush > - to flush I-cache before inserting PTE (at least for the > machines with physical address tagged caches) > > Can we have a separate,

Re: [PATCH] flush icache before set_pte on ia64 take3

2007-07-26 Thread Zoltan Menyhart
I do not think changing the semantics of flush_cache_page() is a good idea. Please have a look at Documentation/cachetlb.txt. This is for virtual address tagged caches. The doc. does not say much about flush_icache_page(). It is used in: install_page(): can release any previously existing

Re: [PATCH] flush icache before set_pte on ia64 take3

2007-07-26 Thread KAMEZAWA Hiroyuki
On Thu, 26 Jul 2007 17:13:23 +0900 KAMEZAWA Hiroyuki <[EMAIL PROTECTED]> wrote: > - lazy_mmu_prot_update() checked pte's executable bit. This was because old >Linux's anon page has executable bit. Current kernel doesn't set executable >bit on anonymous pages. > Linux's anon page had

[PATCH] flush icache before set_pte on ia64 take3

2007-07-26 Thread KAMEZAWA Hiroyuki
This is a patch for flush icache for ia64 take3. against 2.6.23-rc1. tested on ia64/NUMA. Comments ? Should I add CONFIG_MONTECITO ? (It seems header-file dependency is not clean..but...) Thanks, -Kame == [IA64] flush icache before setting pte Changelog: - reviewed the whole patch again and

Re: [PATCH] flush icache before set_pte on ia64 take3

2007-07-26 Thread KAMEZAWA Hiroyuki
On Thu, 26 Jul 2007 17:13:23 +0900 KAMEZAWA Hiroyuki [EMAIL PROTECTED] wrote: - lazy_mmu_prot_update() checked pte's executable bit. This was because old Linux's anon page has executable bit. Current kernel doesn't set executable bit on anonymous pages. Linux's anon page had VM_EXEC

Re: [PATCH] flush icache before set_pte on ia64 take3

2007-07-26 Thread KAMEZAWA Hiroyuki
On Thu, 26 Jul 2007 17:13:52 +0200 Zoltan Menyhart [EMAIL PROTECTED] wrote: I can agree: - lazy_mmu_prot_update() is not the right place for a flush - to flush I-cache before inserting PTE (at least for the machines with physical address tagged caches) Can we have a separate, specific

Re: [PATCH] flush icache before set_pte on ia64 take3

2007-07-26 Thread Zoltan Menyhart
KAMEZAWA Hiroyuki wrote: Thank you for advise. Hmm..how about this ? == /* * synchronize icache and dcache if the hardware doesn't do it automatically * and the page is executable. */ static inline arch_sync_icache_dcache(pte_t pte, page) { } Why not? I just wanted to see if you

[PATCH] flush icache before set_pte on ia64 take3

2007-07-26 Thread KAMEZAWA Hiroyuki
This is a patch for flush icache for ia64 take3. against 2.6.23-rc1. tested on ia64/NUMA. Comments ? Should I add CONFIG_MONTECITO ? (It seems header-file dependency is not clean..but...) Thanks, -Kame == [IA64] flush icache before setting pte Changelog: - reviewed the whole patch again and

Re: [PATCH] flush icache before set_pte on ia64 take3

2007-07-26 Thread Zoltan Menyhart
I do not think changing the semantics of flush_cache_page() is a good idea. Please have a look at Documentation/cachetlb.txt. This is for virtual address tagged caches. The doc. does not say much about flush_icache_page(). It is used in: install_page(): can release any previously existing