Date: Fri, 10 Dec 1999 12:48:08 -0700 From: Warner Losh <imp@village.org> To: nate@mt.sri.com (Nate Williams) Cc: cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/usr.sbin/pccard/pccardd cardd.c Message-ID: <199912101948.MAA24279@harmony.village.org> In-Reply-To: Your message of "Fri, 10 Dec 1999 10:49:04 MST." <199912101749.KAA19687@mt.sri.com> References: <199912101749.KAA19687@mt.sri.com> <199912080747.XAA47581@freefall.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
In message <199912101749.KAA19687@mt.sri.com> Nate Williams writes: : I haven't looked at the code, but Guido and I messed around trying to : find out a good way of doing this for quite a while. There are a number : of subtle races and interactions that make this code extremely fragile, : but I certainly hope you were able to get this fixed! I doubt it. I just made some sanity assumptions. That is, if the last state was suspended and we're getting an insert event, we might want to do an explicit eject event to make sure that the state of the socket is sane before doing the insert. It added another wtiste to the code and I have grave doubts about it fixing all races once and for all. It just fixed the one that I (and lots of others) seemed to be hitting. I tried a while ago to make a tunable parameter the amount of time it waits at various stages of powerup. It worked great some of the time, but failed horribly others when I lowered it from 10s (three times) to anything less than about 7s. To say the code is extremely fragile is an excellent assessment. Warner To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199912101948.MAA24279>