Oops, it's already fixed in 2.6.8-rc2, forget it.
Luciano
Luciano A. Stertz wrote:
I guess I found a bug in the readahead code, kernel 2.6.7.
In filemap_nopage, if the memory area is not marked as sequential (VM_SEQ_READ isn't set) and the page is not in the page cache, the following code is executed:
1 ra_pages = max_sane_readahead(file->f_ra.ra_pages);
2 if (ra_pages) {
3 long start;
4
5 start = pgoff - ra_pages / 2;
6 if (pgoff < 0)
7 pgoff = 0;
8 do_page_cache_readahead(mapping, file, pgoff, ra_pages);
9 }
Seems that the author wanted to read ra_pages around pgoff. Shouldn't it be using 'start' instead of 'pgoff' in lines 6 to 8?!? Start is calculated and never used. Instead of reading pages from start to pgoff + ra_pages/2, it's reading ra_pages from pgoff.
Luciano
-- Luciano A. Stertz luciano@xxxxxxxxxxxx T&T Engenheiros Associados Ltda http://www.tteng.com.br Fone/Fax (51) 3224 8425
-- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
