#tyndur · 2009-08-22
65 Nachrichten · Seite 1 von 1 · Original-CSV herunterladen
- taljethMoin
- FreakyPenguinMoin
- taljethMoin FreakyPenguin, XanClic
- XanClicGood morning, FreakyPenguin, taljeth
- XanClicI am on it ;-)
- XanClicI have already found two mistakes ;-)
- taljethIt was late...
- XanClicBut the second one was none *g*
- taljethHeh. It's not that easy. My mistakes are more subtle. ;-)
- XanClicImagined that.
- taljethThough I already knew that something was wrong, it's doesn't work this way. Just thought you'd tell me what it is. ;-)
- taljeth-'s
- taljethXanClic: Fixed and pushed
- XanClicHave you tested it?
- taljethOnly ls, but seems this was the bug
- taljethI'm wondering why there are cdi_sleep_ms calls in msd.c. Shouldn't this lowlevel stuff be handled by the UHCI driver?
- XanClicIf you send a "read" command, the device must read the data first before it can send it, I think - you may try the stuff without sleeping, if you want to, I haven't done yet ;-)
- taljethI guess you get away with no waiting at all in qemu
- taljethSo this isn't a valid test case
- taljethDoes the driver even run on real hardware now?
- XanClicIt runs under Paloxena ;-)
- XanClicHaven't tried it on týndur for a week - it takes so much time to copy the image to the hard disk
- taljethRuns? On tyndur it's more like creeping. ;-)
- XanClic*g*
- taljethHuh? cp isn't really that difficult...
- taljethAnyway, even if the sleeps are necessary, conceptually they belong to the hardware driver, not to the user (which is msd)
- taljethmsd is doing way too much lowlevel stuff, imho
- XanClicBut if any driver knew how long to sleep, it would be the MSC driver, I think
- XanClicHm, maybe you get a NAK, if the data is not already there...
- XanClicI think, that's it. I have a book about USB here and it says: "Wenn die Function nicht in der Lage ist, Daten zu liefern, wird statt der Datenphase ein NAK-Handshake an den Host gesendet. Dies informiert den Host darüber, dass die Function noch nicht bereit ist, neue Daten zu senden. Dies ist kein Fehler, sondern dient der normalen Ablaufkontrolle bei Interrupt-, Bulk- und Control-Transfer."
- XanClicThat was a quotation, hence I did not translate it ;-)
- taljethIs this msd specific or generic USB functionality?
- XanClicGeneric USB
- taljethGreat. Let's do it in main.c or uhci.c then
- taljethmsd needs to be as dumb as possible
- XanClicOK.
- taljethI guess things like write_cmd or read_status are generic as well
- XanClicI like your patch very well, don't worry ;-)
- taljethActually, it seems we don't even want to have a read_status that returns a complete struct, but more something like a bool usb_is_success()
- XanClicThat's right
- taljethCare to do some of the refactoring or do you prefer me to do it?
- XanClicI do it and you fix it, OK? *g*
- taljethGo ahead ;-)
- XanClicHm, it's not that faster without sleep and with NAK handling
- taljethI wouldn't have it expected to be faster
- taljethIt results in much cleaner code, though
- XanClicBut I hoped it would be faster, too ;-)
- taljethHm... Is there a patch that I missed? ;-)
- XanClicNo, it isn't ;-)
- taljethSo it's only complete enough to know that it's still slow, but not complete enough for sending to the ML?
- XanClicI'm entering the commit message right now ;-)
- taljethHuch... "From: Max Reitz <max@??yndur.org>"
- XanClicHm...
- XanClicI sent it via send-email ;-)
- taljethI think it just uses whatever format-patch generates
- XanClicI think it uses the stuff after "--from"
- XanClicAnd I typed "--from max@tyndur.org"
- XanClicBut I see it here, too...
- XanClicArgh.
- XanClicDead keys are NOT ftw.
- taljethLeave it out next time. send-email figures it out from the patch header by itself.
- taljethNo need for --from
- XanClicI set my mail address to "max@ŧyndur.org"... ><
- XanClicSee the "ŧ"?
- taljethNice domain. ;-)