| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10
Comments:
<0> also provides it all in a nice protected environment
<0> but, it doesnt really provide a WHIZZBANG API
<0> ;) Im a very minimalistic person.
<0> lol
<0> thast all I can really say.
<0> :P
<0> lol
<1> do you have something like address space management for user apps ?
<1> Zeii: something like vma list in linux
<2> hello
<0> hey geezusfreeek
<0> I might be a few seconds guys, configuring the packages.
<0> tehre we go
<0> :)
<2> can grub (or some other mainstream boot loader) load a kernel from an arbitrary sector on disk rather than from a filesystem? would grub be way overkill for that?
<0> Grub would be overkill, I rekon
<0> besides,I dont use grub, so I cant help you there - sorry dude.
<2> hmm, okay
<2> guess i get to make my own bootstrapping
<2> that's what i figured
<0> heheh :)
<2> been kind of looking at x86 thingies and bootstrapping tutorials for the last day or so, but i can't really find anything that helps me know what to do without giving me a boot loader that does way more than i need
<2> no good middle ground :\
<2> maybe i am just not familiar enough with the workings of the architecture
<0> ;) well trust me man
<0> if oyu spend time writing a bootloader
<0> youll get intimate with the architecture very quickly.
<2> i have caught on to that fact by now ;)
<0> heheh
<0> the basic gist of it is,
<0> bootsector must be on the first sector of the disk ;P
<0> the last word of the 512 byte file, must be AA55
<0> or 55AA
<0> I forget now.
<0> ITs Realmode so you can use interrupts and such
<2> i don't really see how i can avoid doing it& everything i find is either some sort of guide that leaves out key details or a full blown bootloader that does way to much for me to see what's going on
<2> yeah, i got that far
<2> i can display characters using bios
<2> that's all i have right now
<0> Ah.
<2> very sad state i'm in
<0> just havent set PM and stuff up?
<2> right, nor have i begun working on getting a kernel loaded from an arbitrary location on disk
<2> (or gotten any kernel code written either)
<2> (which i suppose depends on what state i leave the machine in once it's loaded)
<0> Be careful about how you load your files.
<0> you arent going from a FS are you?
<2> nope
<2> not planning to even have a file system, per se, though i guess it will have one in essence because i'll still be using the hard drive to store data
<0> aye
<2> but it's not even going to be nice to partitions or anything
<0> hehe
<0> I dont use hard drive.
<0> it isnt the kernels responsibility.
<0> atleast in my system
<2> i'm not going to use it for a looong time, though i plan to eventually
<2> only for virtual memory and persistence though, not for arbitrary data storage
<0> while it pains me to say this - ill probably use FAT as my starting filesystem
<0> since, I use fat to boot from disk.
<0> I would preferably use ELF
<0> er, EXT2
<0> but I dont understand how to boot from it
<2> if i was using a file system, i would be dying to try ZFS, but of course that has to be the most complicated one in existence
<2> and i don't think anybody has successfully booted from it yet either
<2> not even Sun programmers
<3> yeah, they haven't
<3> it's not that compicated, it's actually pretty simple in structure
<3> but finding all the members of the pool and whatnot is nontrivial
<2> i haven't looked that much into implementation& what i saw looked pretty complicated to me
<0> anyone have any advice for setting up CVS?
<2> gosh i can't wait for it in os x
<3> dont. use something better
<2> Zeii: don't
<3> geezusfreeek: the source code isn't that big really. only about 30k lines of code or so
<2> use subversion if you think like a cvs user, use darcs otherwise
<0> ???
<0> ive never used versioning stuff before
<0> so...
<2> subversion is probably more intuitive
<3> which is why we're suggesting you not use cvs
<3> yeah, i've been happy with svn for hobby stuff
<2> it's fast too
<3> and easy to set up
<2> darcs (and similars) is sloooooow
<0> brb, getting kernel compiling on old box
<2> geist: 30k isn't as much as i expected, but that's still well above average for filesystems, isn't it?
<3> no, not at all
<2> okay
<2> then i obviously just **** as life :)
<3> lots of fs implementations for complex ones (say xfs) easily cross 100k or so
<2> s/as/at/
<0> bbl people
<4> zfs was what, 50KLoC?
<3> somethign like that
<3> relatively small
<4> seems pretty full featured though, yay.
<4> ungh hate sick.
<3> monkey hate clean
<4> geist: how's life? Last time (a couple months ago) you seemed stressed w/ work, better now?
<3> yep. unemployed
<4> Oh?!
<3> as of yesterday
<4> w.t.f.?
<4> quit?
<3> yup
<4> yay!
<4> welcome to the resignation club!
<4> I was in Germany last week.
<4> it's about time to look for a new job now. :)
<4> I kind of didn't want to look before I left; but it's been a couple months now.
<4> a _nice_ couple months.
<4> what are your plans with the time off? Enjoy life for a little while?
<3> sorry, back
<3> gonna go visit parents for a few days, then start again
<3> found a job with a console game company up in the city. gonna hack ps3s
<1> I have a question
<1> should all threads in task have the same elfags ?
<1> or, each thread can has its own eflags value ?
<3> well, what is in eflags that is interesting to a thread?
<3> okay, lets make it easier. yes, each thread has it's own eflags
<3> however, in lots of systems the eflags are always the same when you context switch
<3> and thus you dont have to actually save it
<1> oh!
<1> geist: thank you !
<1> for I find some thread switch code does not change the eflags at all
<3> the only really meaningful bits in eflags are the interrupt disabled bits, but usually interrupts are disabled when context switching always
<3> so you know the state of the flags
<3> and if the context switching code is a function call, the condition bits are already trashed by the calling convention
<1> yes,
<1> you tell me clearly enough :-) So I can ignore the eflags in context switching
<1> geist: so, it seems all the usre applications' elfags are maybe same, too ?
<3> users eflags are saved on the stack when the cpu takes an interrupt or syscall
<1> I see.
<1> I mean it seems unnecessary to make elfags in each user application differs ?
<1> or can all the eflags for each task be same ?
<3> not for user space
<3> it's most definitely different and important
<1> geist: do you mean, when task in user space, the eflags is also meaningful ?
<3> you bet
<3> eflags are *very* important
<3> they just are already saved by the time a context switch happens
<1> geist: I understand you :-)
<1> thank you !
<1> I just wondering how kernel generate the first user space task :-)
<3> that's tricky
<3> but basically you write out a stack as if the task had switched to the kernel
<1> yes
<1> I just want to know, whether a sys_fork from kernel is necessary
<3> depends on how you design it. but probably not
<3> fork is dumb
<1> or, I should make a separate syscall like sys_fisrtapp :-)
Return to
#osdev
or
Go to some related
logs:
gaim not authorised
gentoo gcc 4.1 error compile open office
#lisp
ctorrent glibc detected
#linux
fglrxinfo reports wrong driver ati
#php
moduser gentoo
modconf ipw2200
#ai
|
|