| |||||||||||||
Dcache scalability and RCU-walkDcache scalability and RCU-walkPosted Dec 16, 2010 7:23 UTC (Thu) by jzbiciak (✭ supporter ✭, #5246)Parent article: Dcache scalability and RCU-walk For some reason, I envision the RCU walk on a dynamically changing filesystem as occasionally being a little like this.
(Log in to post comments)
Dcache scalability and RCU-walk Posted Dec 16, 2010 14:50 UTC (Thu) by Nick (guest, #15060) [Link]
That's pretty accurate! Sometimes the step crumbles before our hero gets to the next one. Sometimes the next step is missing completely.
Fortunately there is a safety net at the bottom, and a less crumbly but slower staircase available.
But actually operations that require a sequence count change are not so common, and path walks require sequence protection only briefly, so sequence count failure is quite rare. It's more common that the dentry is missing or some filesystem operation requires blocking, which can be handled more gracefully.
Dcache scalability and RCU-walk Posted Dec 16, 2010 15:15 UTC (Thu) by dlang (✭ supporter ✭, #313) [Link]
that link doesn't work for me
Dcache scalability and RCU-walk Posted Dec 16, 2010 15:33 UTC (Thu) by jzbiciak (✭ supporter ✭, #5246) [Link] I was afraid that would happen. Here, I re-hosted it. (Image is from Getty Images, FWIW, in case the watermark wasn't enough to clue you in.)
|
|||||||||||||
