tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Using wsdisplay from the Xserver

Hash: SHA1


On Jan 24, 2009, at 1:14 PM, der Mouse wrote:

No, not "support [the] hardware".  I'm not proposing to mvoe the X
server into the kernel.  Just support _access to_ the hardware.
From the perspective of INSECURE, though, the effect is the same,
since part of what you're providing access to is the video hardware's
various and sundry DMA engines

Oh, ugh.  Yes, that needs to be in the kernel.

Most of what Michael did looks fine, yes.  It's things like the
PCI_IOC_CFGREAD he mentioned

That's an interface which is already there and used by X. Why re- invent the wheel? I have neither time nor motivation to rewrite large parts of the Xserver, I'm just trying to get a little more sanity into it.

and the "probably shouldn't allow mapping PCI resources" that seem to be to be saying the X server still knows about PCI directly that prompted my reaction.

It does, and so does every single PCI/AGP graphics chip driver. Abstracting PCI-related knowledge out in a lower layer and then letting it creep back in again further up seems utterly pointless to me. There is absolutely nothing wrong with the Xserver knowing that it deals with PCI devices. I'm trying to keep it from knowing ( or caring ) too much about implementation specifics, bus layout, bridges, non-graphics-devices and whatnot - that's none of its business. Accessing a graphics chip's config space is ok in my book and indeed necessary. Reading that is, I wouldn't want X to muck with my BARs for instance, and that's one thing my additions would prevent.

And, again, mmap()ing PCI stuff through ttyE* is already restricted to ranges belonging to the respective graphics chip.

have fun

Version: GnuPG v1.4.7 (Darwin)


Home | Main Index | Thread Index | Old Index