>The only reason to add a bitflag in the page->flags field is to avoid
>us to play with the pte. Now the pte is used only to set the page readonly
>to allow us to remove the clean flag at the first page fault.
Ah but I just found a problem... When we set the PG_clean flag on the page
we should set the pte readonly for that page in all process vm and not
only in the process running. But if we must play with the page table it's
easier to directly set the page as clean as I was used to do with my
previous update_shared_mappings() patch. So I think we could drop
completly my last patch and return to my old code and solve the problem to
handle the mmap_sem locking right...
To get the locking right I think we could do something like:
if (sharedmapping)
{
for_each_process_that_shares_the_mmap(p)
down(&p->mm->mmap_sem);
} else {
down(¤t->mm->mmap_sem);
}
for_each_process_that_shares_the_mmap() will return processes always in
the same order relative to how the mappings are ordered in the the inode.
comments?
Andrea Arcangeli
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/