Messages in this thread Patch in this message | | | Date | Wed, 21 Mar 2001 22:23:53 -0800 (PST) | From | Linus Torvalds <> | Subject | Re: kswapd deadlock 2.4.3-pre6 |
| |
On Wed, 21 Mar 2001, Linus Torvalds wrote: > > The deadlock implies that somebody scheduled with page_table_lock held. > Which would be really bad.
..and it is probably do_swap_page().
Despite the name, "lookup_swap_cache()" does more than a lookup - it will wait for the page that it looked up. And we call it with the page_table_lock held in do_swap_page().
Ho humm. Does the appended patch fix it for you? Looks obvious enough, but this bug is actually hidden on true SMP, and I'm too lazy to test with "num_cpus=1" or something..
Linus
----- diff -u --recursive --new-file pre6/linux/mm/memory.c linux/mm/memory.c --- pre6/linux/mm/memory.c Tue Mar 20 23:13:03 2001 +++ linux/mm/memory.c Wed Mar 21 22:21:27 2001 @@ -1031,18 +1031,20 @@ struct vm_area_struct * vma, unsigned long address, pte_t * page_table, swp_entry_t entry, int write_access) { - struct page *page = lookup_swap_cache(entry); + struct page *page; pte_t pte;
+ spin_unlock(&mm->page_table_lock); + page = lookup_swap_cache(entry); if (!page) { - spin_unlock(&mm->page_table_lock); lock_kernel(); swapin_readahead(entry); page = read_swap_cache(entry); unlock_kernel(); - spin_lock(&mm->page_table_lock); - if (!page) + if (!page) { + spin_lock(&mm->page_table_lock); return -1; + }
flush_page_to_ram(page); flush_icache_page(vma, page); @@ -1053,13 +1055,13 @@ * Must lock page before transferring our swap count to already * obtained page count. */ - spin_unlock(&mm->page_table_lock); lock_page(page); - spin_lock(&mm->page_table_lock);
/* - * Back out if somebody else faulted in this pte while we slept. + * Back out if somebody else faulted in this pte while we + * released the page table lock. */ + spin_lock(&mm->page_table_lock); if (pte_present(*page_table)) { UnlockPage(page); page_cache_release(page); - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |