Conversation with #inferno at Wed Mar 31 15:07:45 2010 on powerman-asdf@irc.freenode.net (irc) (15:08:01) powerman-asdf left the room (quit: Read error: Connection reset by peer). (15:08:56) mkn left the room. (15:19:26) bvalek2 [~c11a2f4b@gateway/web/freenode/x-ippzexniuoctrmvu] entered the room. (15:38:30) mkmks [~nf@90-230-91-71-no148.tbcn.telia.com] entered the room. (15:53:36) wrtp: vsrinivas: where should the hooks be? (16:12:06) robot12 left the room. (16:17:18) mennis [~mennis@adsl-068-016-104-079.sip.asm.bellsouth.net] entered the room. (16:53:06) mkn [~c2ed8e07@gateway/web/freenode/x-pzfonwzkavhyppne] entered the room. (17:03:58) bvalek2 left the room (quit: Quit: Page closed). (17:25:19) mennis left the room (quit: Quit: mennis). (17:25:27) mennis [~mennis@adsl-068-016-104-079.sip.asm.bellsouth.net] entered the room. (17:25:52) mennis left the room (quit: Read error: Connection reset by peer). (17:26:12) mennis [~mennis@adsl-068-016-104-079.sip.asm.bellsouth.net] entered the room. (17:47:04) vsrinivas: to be honest, write a 9p server that sends TClunk for non-Excl files lazily ; see what happens. have it keep a suffix tree of open fids to dirs, you can eliminate a good number of stats and walks. (17:47:12) vsrinivas: all while remaining coherent. (17:47:34) vsrinivas: once that's exhausted, attack caching; (17:48:07) vsrinivas: you want the hooks on the 9p ops, not the syscalls that cause them... (17:54:48) mkn left the room (quit: Ping timeout: 252 seconds). (17:55:52) wrtp: the hooks are in the 9p ops, more or less (17:55:58) wrtp: they're in devmnt (17:56:25) vsrinivas: yes. they're in response to open, read, ... (17:58:21) wrtp: thing is, you're always going to have to do at least one round trip when opening a file, to see if qid.version has changed (17:58:40) wrtp: i suppose you save a walk if you keep the fid around (18:12:18) bvalek2 [~bela@unaffiliated/bvalek2] entered the room. (18:15:51) powerman: is this safe? for example, if app poll to check is file/dir exists, file initially exists (fid cached), then file removed on server, but app won't notice this because instead of sending real Twalk cached fid will be used (18:20:37) less1 [~pravin@32.97.110.64] entered the room. (18:20:58) wrtp: powerman-asdf1: you have to at least stat the file each time it's opened (18:21:35) wrtp: any caching at all is not ok if you want to play strictly by the rules (18:21:48) wrtp: you have to decide what incoherence you can live with (18:31:19) less1 left the room (quit: Ping timeout: 260 seconds). (18:38:05) mennis left the room (quit: Quit: mennis). (18:45:52) less1 [~pravin@32.97.110.63] entered the room. (19:03:09) bvalek2 left the room (quit: Quit: I've seen things you people wouldn't believe). (19:10:33) powerman: caching can be safe, but this add a lot of complexity - timeouts, full-duplex protocol so server can request cache invalidation on client, etc. (19:13:34) bvalek2 [~bela@unaffiliated/bvalek2] entered the room. (19:14:12) wrtp: powerman-asdf1: yeah, we're not talking 9p here... (19:16:14) powerman: this can be done using 9p. caching client will be 9p server as same time exporting fs with details of what is cached now, and original server which aware about caching client can connect to that client's fs and use it to control client's cache (19:27:12) wrtp: yes, i guess it's possible. but it's quite hard sending metadata out-of-band, because 9p replies and cache invalidation might get re-ordered with respect to one another. unless it's in the same 9p stream and the mount driver can somehow access the alternate fs. attach files might make that easier, be linking two separate fs's. (20:34:14) eno left the room (quit: Read error: Connection reset by peer). (20:39:14) eno [~eno@nslu2-linux/eno] entered the room. (20:49:39) vsrinivas: wrtp: doesn't need to be out-of-band; also , see oplocks (20:49:55) vsrinivas: not really great, to be fair. (21:02:50) wrtp: vsrinivas: oplocks? (22:02:18) visof [~visof@41.238.232.160] entered the room. (22:36:58) visof left the room (quit: Remote host closed the connection). (23:15:45) ericvh [~ericvh@32.97.110.63] entered the room. (23:48:03) ericvh left the room (quit: Quit: ericvh). (00:45:06) adelfino [~username@201-212-160-19.net.prima.net.ar] entered the room. (00:45:17) adelfino left the room (quit: Remote host closed the connection). (00:46:27) bvalek2 left the room (quit: Quit: I've seen things you people wouldn't believe). (02:30:21) less1 left the room (quit: Quit: Leaving.). (02:32:36) anth left the room (quit: Ping timeout: 245 seconds). (02:52:18) anth_x [~a@adsl-99-25-148-5.dsl.bcvloh.sbcglobal.net] entered the room. (04:45:51) adelfino [~username@201-212-160-19.net.prima.net.ar] entered the room. (05:02:14) adelfino left the room (quit: Remote host closed the connection). (06:17:29) mkmks left the room (quit: Ping timeout: 248 seconds). (07:07:31) robot12 [~robot12@robot12.kgts.ru] entered the room. (07:11:26) vsrinivas is now known as jokeserver (07:12:37) jokeserver left the room (quit: Disconnected by services). (08:14:01) less1 [~pravin@cpe-66-68-151-36.austin.res.rr.com] entered the room. (08:32:46) less1 left the room (quit: Quit: Leaving.). (08:55:09) rapidfx [~host666@vl-cen-ce1.avtlg.ru] entered the room. (10:17:05) bvalek2 [~c11a2f4d@gateway/web/freenode/x-klwjttbfsfqrvqfm] entered the room. (11:49:20) rapidfx left the room (quit: Ping timeout: 276 seconds). (12:02:55) rapidfx [~host666@vl-cen-ce1.avtlg.ru] entered the room. (13:48:19) The account has disconnected and you are no longer in this chat. You will be automatically rejoined in the chat when the account reconnects.