Messages in this thread | | | Date | Tue, 4 Aug 1998 20:35:25 -0500 (CDT) | From | Shawn Leas <> | Subject | Re: DEVFSv50 and /dev/fb? (or /dev/fb/? ???) |
| |
On Wed, 5 Aug 1998, Andrea Arcangeli wrote: > On Tue, 4 Aug 1998, Shawn Leas wrote: > > >Specifically, what is ugly about it? Is it intrinsic in nature, or > >something fixable? Is the design flawed? Then how? Back up your > > I don' t know about the Theodore idea of ugly but I agree with him. _I_ > see devfs flowed by its start. I just explained my thoughts about devfs > some month ago before the first release of devfs. devfs is a workaround. > We would not need devfs at all if the kdev_t would be a 64 bit unsigned > integer. > > Instead of reply me that with devfs the root device _can_ be mounted > readonly and we _can_ boot with a root fs with no major/minor number > support, please tell me that you need to use these features. We could also > write a videogame in the kernel and eventually play with it before mount > the root device if you really want not used features (even if this example > is not the right one, the exciting videogame probably would be played a > lot ;-).
Simply having inodes with device info (Major/minor devices) incurrs a penalty in that you have these inodes on a filesystem on disk (or maybe RAM), in any case you are busying up yet another subsystem, and so on. devfs is a shorter route to access devices, and this improves (on some large scale systems) performance when you have a zillion pseudodevices and you are accessing them a lot. Even if you give me "Dcache speeds that up", I can say, well, that's only true while the entries are in the dcache.
Think of having proc be a collection of physical directories with physical files on disk. I know, this is not an exactly similar idea, but the devfs shares some procfs-like goodies...
This is, of course, not to mention that device files only exist if they are really there... Imagine, just
if [ -r /devfs/ide/hd/blah ]; then echo "you have an ide disk" fi
if [ -d /dev/fs/ide/cd ]; then echo "you have a cdrom [or support for one]" fi
or something like that... It's VERY NICE from a system management POV. I can't wait to see what distributions do with it.
> I had to say that I never tried devfs and that it could be a very > confortable workaround but I can' t like it ;-). Probably I' ll try soon > though. I' m afraid to say this again and after the sure good work of > Richard...
Well, I cannot gaurantee the total correctness of exactly everything I claim, but I do understand that the devfs is a shortcut from userspace to devices. This is a "Good Thing" (tm), in my book.
-Shawn <=========== America Held Hostage ===========> Day 2022 for the poor and the middle class. Day 2041 for the rich and the dead. 900 days remaining in the Raw Deal. <============================================>
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html
| |