Date: Sun, 8 Jan 95 14:13:44 MST From: terry@cs.weber.edu (Terry Lambert) To: phk@ref.tfs.com (Poul-Henning Kamp) Cc: hackers@freebsd.org Subject: Dear Poul... (was: Dear Terry...) Message-ID: <9501082113.AA28360@cs.weber.edu> In-Reply-To: <199501070703.XAA07527@ref.tfs.com> from "Poul-Henning Kamp" at Jan 6, 95 11:03:10 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> Sometimes I can't help it, I get these weird ideas... > > When I read this email I felt like a Roman emperor, on my march of triumph > through Rome, and while my heart swells with pride I have this slave > wispering in my ear: "remember cesar, thou art mortal!" > (look it up!) I got the reference... personally, I feel like Claudius. > Well, I just wanted to say thanks! Sometimes we need a devils advocate ! Ana osa requiu de pasqualum... (spelling? I'm not used to Latin character sets...) 8-). > regarding PCMCIA. > > I personally think that the PCMCIA spec suffers from needless generality. > > My plan is to implement the things we actually want to have, as opposed to > things we might want to be able to do some day. That means that I will > be looking for a simple framework and a lightweight implementation. > > The single most popular cards will be modems, ethers, disks and possibly > SCSIs. I am somewhat interested in execute_in_place from a FLASH card too, > but apart from that, I don't really plan to carry it further. Oh, I wasn't suggesting any more than that. It's just that the UnixWare approach (written by Kurt Mahon) suffers from a generality in one area and a lack of generality in another. Specifically, he's written his code as a set of "enablers", the sole purpose of which is to attach absolutlely anything via PCMCIA as if it were an ISA device, and then use the ISA drivers, which don't know about anything about detach. His is too general, in that it encompasses everything, and it's too specific, in that it doesn't allow for popping a card out while it's hot (one of your design goals, you said). Actually, it's relatively trivial to handle the general case of enabling almost everything, excepting the PCMCIA cards that like to map into multiple address ranges or to multiple interrupts. The reason I suggested a bus attach is the same reason that the AHA 29xx drivers don't work: card drivers should have specific knowledge of an attach record, not specific knowledge of a bus. The same goes for cards that can be PCMCIA, ISA, or bridged on the motherboard (ie: some SoundBlaster implementations), as well as multiple SCSI and CDROM interfaces on daughter-boards on sound cards. That's actually a general problem that needs to be solved, and my prodding was to get you thinking about it so that at least it was not excluded in the design (meaning that if it is a goal, it won't have to be a total redesign later to achieve it). Regards, Terry Lambert terry@cs.weber.edu --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9501082113.AA28360>