On 11/02/2015 01:39 AM, Hugh Dickins wrote: <> >> This patch is not correct! > > I think you have actually confirmed that the patch is correct: > why bother to test PageDirty or PageWriteback when PageUptodate > already tells you what you need? > > Or do these filesystems do something unusual with PageUptodate > when PageDirty is set? I didn't find it. > This is kind of delicate stuff. It took me a while to get it right when I did it. I don't remember all the details. But consider this option: exofs_write_begin on a full PAGE_CACHE_SIZE, the page is instantiated new in page-cache is that PageUptodate(page) then? I thought not. (exofs does not set that) Now that page I do not want to read in. The latest data is in memory. (Same when this page is in writeback, dirty-bit is cleared) So for sure if page is dirty or writeback then we surly do not need a read. only if not then we need to consider the PageUptodate(page) state. Do you think the code is actually wrong as is? BTW: Very similar code is in fs/nfs/objlayout/objio_osd.c::__r4w_get_page > Thanks, > Hugh > <> Thanks Boaz -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>