[lkml]   [1998]   [Jan]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: shared anonymous memory
>>>>> "RJ" == Richard Jones <> writes:


RJ> But the point is that if we check inode->i_nlink and
RJ> it's 0, then we will probably never write those blocks
RJ> out to the backing store (unless we run out of memory),
RJ> so the performance of NFS isn't an issue.

Assuming small blocks, that exist for a short time. For large ones,
or long lasting ones it is an issue. Worst case is your application
has one shared page paged out. And then the NFS server crashes. Your
app is frozen until NFS comes back up. It is an issue.

RJ> Andreas Schwab <> wrote:
inode-> i_links == 0 ?

RJ> (i_nlink?) I'll have to take a look at the 2.1 code
RJ> tonight to see if it already checks this when it does
RJ> a sync, or if it's possible to patch it ...

The common case is sys_sync, or fsync(dev_fd);
For which it more efficient to totally ignore files.

Why would anyone call fsync(shared_area_fd)??


 \ /
  Last update: 2005-03-22 13:41    [W:0.034 / U:2.632 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site