Messages in this thread | | | Date | Mon, 11 Aug 2014 11:32:48 +0800 | From | "Zhang, Yanmin" <> | Subject | Re: [PATCH] export the function kmap_flush_unused. |
| |
On 2014/8/11 9:26, Sha, Ruibin wrote: > Hi Chintan, > Thank you very much for your timely and kindly response and comments. > > Here is more detail about our Scenario: > > We have a big driver on Android product. The driver allocates lots of > DDR pages. When applications mmap a file exported from the driver, > driver would mmap the pages to the application space, usually with > uncachable prot. > On ia32/x86_64 arch, we have to avoid page cache alias issue. When > driver allocates the pages, it would change page original mapping in > page table with uncachable prot. Sometimes, the allocated page was > used by kmap/kunmap. After kunmap, the page is still mapped in KMAP > space. The entries in KMAP page table are not cleaned up until a > kernel thread flushes the freed KMAP pages(usually it is woken up by kunmap). > It means the driver need force to flush the KMAP page table entries before mapping pages to > application space to be used. Otherwise, there is a race to create > cache alias. > > To resolve this issue, we need export function kmap_flush_unused as > the driver is compiled as module. Then, the driver calls > kmap_flush_unused if the allocated pages are in HIGHMEM and being > used by kmap. > > Thanks again! > > Best Regards > --------------------------------------------------------------- > Sha, Rui bin ( Robin ) > +86 13817890945 > Android System Integration Shanghai > > -----Original Message----- > From: Chintan Pandya [mailto:cpandya@codeaurora.org] > Sent: Friday, August 8, 2014 9:40 PM > To: Sha, Ruibin > Cc: linux-kernel@vger.kernel.org; linux-mm@kvack.org; mel@csn.ul.ie; a.p.zijlstra@chello.nl; mgorman@suse.de; mingo@redhat.com; Zhang, Yanmin; He, Bo > Subject: Re: [PATCH] export the function kmap_flush_unused. > > On 08/08/2014 02:46 PM, Sha, Ruibin wrote: >> export the function kmap_flush_unused. >> >> Scenario: When graphic driver need high memory spece, we use >> alloc_pages() to allocate. But if the allocated page has just been >> mapped in the KMAP space(like first kmap then kunmap) and no flush >> page happened on PKMAP, the page virtual address is not NULL.Then when >> we get that page and set page attribute like set_memory_uc and >> set_memory_wc, we hit error. > Could you explain your scenario with more details ? set_memory_* should be applied on mapped address. And in attempt to map your page (which was just kmap and kunmap'ed), it will overwrite the previous mappings. > > Moreover, in my view, kmap_flush_unused is just helping us in keeping the cache clean for kmap virtual addresses if they are unmapped. Is it serving any more purpose here ?
It depends on how to define 'clean' here. It resets pkmap_count[i] to 0, and cleans up page table entries used by PKMAP. Here, our scenario is caused by the late page table entry cleanup as driver need avoid page cache alias.
> >> fix: For that scenario,when we get the allocated page and its virtual >> address is not NULL, we would like first flush that page. >> So need export that function kmap_flush_unused. >> >> Signed-off-by: sha, ruibin <ruibin.sha@intel.com> >> >> --- >> mm/highmem.c | 1 + >> 1 file changed, 1 insertion(+) >> >> diff --git a/mm/highmem.c b/mm/highmem.c >> index b32b70c..511299b 100644 >> --- a/mm/highmem.c >> +++ b/mm/highmem.c >> @@ -156,6 +156,7 @@ void kmap_flush_unused(void) >> flush_all_zero_pkmaps(); >> unlock_kmap(); >> } >> +EXPORT_SYMBOL(kmap_flush_unused); > This symbol is already extern'ed. Is it not sufficient for your case ?
We want to call it in driver module. extern is not enough.
Thanks, Yanmin
| |