Date: Wed, 18 Aug 1999 13:04:31 -0700 From: Amancio Hasty <hasty@rah.star-gate.com> To: "Kenneth D. Merry" <ken@kdm.org> Cc: dillon@apollo.backplane.com, wilko@yedi.iaf.nl, current@FreeBSD.ORG Subject: Re: [re]writable cdrom drive Message-ID: <199908182004.NAA01623@rah.star-gate.com> In-Reply-To: Your message of "Wed, 18 Aug 1999 13:55:38 MDT." <199908181955.NAA21794@panzer.kdm.org>
next in thread | previous in thread | raw e-mail | index | archive | help
> Amancio Hasty wrote...
> > > There is a generic SCSI bus enumerator API, and one for PCI as well. There
> > > isn't a generic API to get trees of devices of any type in the system,
> > > though. With new-bus that might be possible, but having done two
> > > enumerator APIs already (I wrote the CAM and PCI enumerator code), I can
> > > predict that it might be very complicated to do properly. (Things get
> > > really hairy when you have to keep the device tree from changing while
> > > you're traversing it.)
> > >
> > > In any case, for the two applications you specified, cdrecord is only
> > > interested in SCSI devices, and there is already an API -- that cdrecord
> > > already supports -- to do bus/device enumeration. The cdrecord -scanbus
> > > code already uses that API.
> > >
> > > Similarly, there is already a PCI device listing interface, so fxtv should
> > > be able to easily pull out the devices it is interested in. All it has to
> > > do is search for all "meteor" or "bktr" devices to find what it wants.
> > >
> > > Ken
> >
> > Now how can I map a device name from this bus enumeration so that I can issue
> > an "open" to the device:
> >
> > pciconf -l
> > chip0@pci0:0:0: class=0x060000 card=0x00000000 chip=0x12378086 rev=0x02
> > hdr=0x00
> > isab0@pci0:1:0: class=0x060100 card=0x00000000 chip=0x70008086 rev=0x01
> > hdr=0x00
> > ide_pci0@pci0:1:1: class=0x010180 card=0x00000000 chip=0x70108086
> > rev=0x00 hdr=0x00
> > chip1@pci0:1:2: class=0x0c0300 card=0x00000000 chip=0x70208086 rev=0x01
> > hdr=0x00
> > de0@pci0:9:0: class=0x020000 card=0x11001385 chip=0x00091011 rev=0x22
> > hdr=0x00
> > bktr0@pci0:10:0: class=0x040000 card=0x13eb0070 chip=0x036e109e
> > rev=0x02 hdr=0x00
> > none0@pci0:10:1: class=0x048000 card=0x13eb0070 chip=0x0878109e
> > rev=0x02 hdr=0x00
> > ahc0@pci0:11:0: class=0x010000 card=0x00000000 chip=0x81789004 rev=0x00
> > hdr=0x00
> > ahc1@pci0:12:0: class=0x010000 card=0x00000000 chip=0x71789004 rev=0x03
> > hdr=0x00
> > vga-pci0@pci0:13:0: class=0x030000 card=0x00000000 chip=0x0519102b
> > rev=0x01 hdr=0x00
>
> Only one of the above devices actually has a character device interface in
> /dev. All you have to do is something like this:
>
> sprintf(foo, "/dev/%s%d", p->pd_name, p->pd_unit);
>
> The only way it would become complicated is if the device name in the
> kernel were different than the device name in /dev.
>
> Ken
Or if there is a system call or modification to open such that I can issue :
open("pci0:10:0", .....)
On the kernel side, I hope that the pci code can translate a bus enumeration
to an actual major , minor device.
The only problem that I see with such an approach is security and I would
rather deal with that later.
--
Amancio Hasty
hasty@rah.star-gate.com
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199908182004.NAA01623>