Conversation with #inferno at Wed Mar 31 11:02:51 2010 on powerman-asdf@irc.freenode.net (irc)
(11:05:22) mjl-: hmm
(11:05:23) mjl-: http://www.vitanuova.com/inferno/man/4/dossrv.html
(11:05:33) mjl-: -f devicefile   the DOS/Windows file system image; defaults to /dev/hd0disk. 
(11:05:36) mjl-: hd0disk?!
(11:05:42) mjl-: never heard of that
(11:13:13) wrtp: mjl-: it's a mistake - i think they meant fd0disk. and in fact there is no default.
(11:20:57) robot12: obot12@darkstar:~/Inferno/appl/cmd$ cat dossrv.b | grep deffile
(11:20:58) robot12: deffile: string;
(11:20:58) robot12:  deffile = ""; # no default, for safety
(11:20:58) robot12:    deffile = arg->arg();
(11:20:58) robot12:    if(deffile == nil)
(11:20:58) robot12:  if(deffile == "" || !stdin && srvfile == "")
(11:20:58) robot12:   if(sys->mount(pipefd[0], nil, srvfile, sys->MREPL|sys->MCREATE, deffile) < 0)
(11:20:59) robot12:   t.aname = deffile;
(11:20:59) robot12: robot12@darkstar:~/Inferno/appl/cmd$  
(11:49:00) mjl-: wrtp: that makes sense.
(11:49:12) mjl-: i thought hd0disk might refer to an old interface
(11:49:46) wrtp: dossrv has historically almost always been used for floppies
(11:52:56) mjl-: ok
(11:53:07) mjl-: that also explains not supporting all the fat32 stuff
(13:07:42) bvalek2 left the room (quit: Ping timeout: 252 seconds).
(13:12:07) bvalek2 [~c11a2f4b@gateway/web/freenode/x-ijkhsknmexzwyqrj] entered the room.
(13:52:53) mjl-: hmm
(13:52:56) mjl-: i'm writing a plan9 partition
(13:53:15) mjl-: so second sector has the lines of the form "part <name> <start> <end>\n"
(13:53:37) mjl-: disk/prep tells me that i'm leaving the first two sectors unused
(13:53:46) mjl-: can't remember that from plan 9...
(13:54:05) mjl-: i can't put data in the first two sectors as the first is boot stuff and the second holds the partition table
(13:54:20) mjl-: so perhaps <start> and <end> are offset by 2 sectors?
(13:58:49) wrtp: i'm afraid it's ages since i prepped a disk - can't remember anything!
(13:59:02) bvalek2 left the room (quit: Ping timeout: 252 seconds).
(14:00:27) mjl-: i find it funny i have to put a "master boot record" on an sd card to be used on an ARM machine :)
(14:00:58) mjl-: wrtp: do you know of anything like a disk block cache for inferno (or plan 9 for that matter)
(14:01:12) mjl-: at least reading from disk could be easily cached
(14:03:32) wrtp: there's MCACHE as a mount option, but it doesn't do anything in hosted mode (see emu/port/cache.c)
(14:05:15) mjl-: i guess that's the same as os/port/nocache.c?
(14:05:55) wrtp: yeah, looks like it's stubs all round
(14:06:10) wrtp: well, the hooks are there - shouldn't be too difficult to flesh out
(14:06:13) mjl-: yeah
(14:06:23) mjl-: nice to have hooks in the right place
(14:06:25) wrtp: just have to decide how much memory you're willing to spend
(14:07:19) mjl-: yeah.  ideally all memory, with evictions when malloc() needs it
(14:07:28) wrtp: and you've got /sys/src/9/port/cache.c as a current working implementation...
(14:07:39) wrtp: mjl-: in hosted mode?
(14:07:42) wrtp: or native?
(14:07:44) mjl-: native
(14:07:57) wrtp: yeah, sure, in that case
(14:08:05) mjl-: have to be careful though.  not to cause fragmentation
(14:08:10) mjl-: but that should be doable.
(14:08:38) mjl-: but, i won't be doing it anytime soon i'm afraid.  enough more pressing code needs :)
(14:09:00) wrtp: sure, it's just an optimisation after all :-)
(14:09:39) mkn [~c2ed8e06@gateway/web/freenode/x-zertlurladivhrkf] entered the room.
(14:29:07) bvalek2 [~c11a2f4b@gateway/web/freenode/x-fwhznsamjbkqhndb] entered the room.
(14:55:53) vsrinivas: the hooks for MCACHE aren't in the greatest places.