On Mon, 10 May 1999 10:33:59 -0700 (PDT), kanoj@google.engr.sgi.com
(Kanoj Sarcar) said:
> There are probably a lot of problems with the code as it stands
> today. Reviewers, please let me know of any possible improvements.
> Any ideas on how to improve the uaccess performance will also be
> greatly appreciated. Testers, your input will be most valuable.
On a first scan one thing in particular jumped out:
+/*
+ * validate in a user page, so that the kernel can use the kernel direct
+ * mapped vaddr for the physical page to access user data. This locking
+ * relies on the fact that the caller has kernel_lock held, which restricts
+ * kswapd (or anyone else looking for a free page) from running and stealing
+ * pages. By the same token, grabbing mmap_sem is not needed.
+ */
Unfortunately, mmap_sem _is_ needed here. Both find_extend_vma and
handle_mm_fault need it. You can't modify or block while scanning the
vma list without it, or you risk breaking things in threaded
applications (for example, taking a page fault in handle_mm_fault
without it can be nasty if you are in the middle of a munmap at the
time).
--Stephen
-
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/