[lkml]   [2000]   [Oct]   [30]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: 2.2.18Pre Lan Performance Rocks!
> > 
> > I just guess the end result will be as crash prone as Netware when you install any third
> > party software ;)
> >
> > lazy mm is probably a better path, as long as you stay in kernel threads and a single user mm
> > it'll never switch VMs.
> It's not that bad. Some code has been running for years and will be OK.
> Also, the number of Ooops people get anyway on Linux today is about the
> same state as Abends were in NetWare in their frequency in Live customers
> accounts. I've also seen miscreant user apps in Linux wreck havoc
> from user space as deadly as an Oops.
> It's not the activity in the lan path that's the problem, it's the switching
> period that's the problem. The fix goes in the loader -- trusted apps get
> loaded in the kernel, non-trusted apps get loaded in a ring 3 address space.
> The WTD thing I described stops them from context switching until the
> system goes ile, since LAN I/O always gets moved to the front of the
> work queue, if you remember my post describing it. NetWare does this
> trick by holding off CR3 reloads to apps until the server goes idle
> relative to incoming or outging I/O. Linus just lets both happen
> all over the place. I'll show you end of December with the port. Ring 3
> apps will still work - you'll just be able to load some of them at ring 0,
> and run like bat out of hell.
> You are right though, Linux is like a tank going 25 miles an hour crushing
> everything in it's path, while NetWare is an aluminum frame speed racer
> that weighs 20 lbs., gets 1000 miles to the gallon, runs at MACH VI, and
> will explode in flames if it hits a tack in the road...
> :-)
> > > stack. Since you look at MANOS code, you'll note that in
> > > CONTEXT.386, I do and add esp, 3 * 4 instead of poppping the segment
> > > registers off the stack if they are in the kernel address space.
> > >
> > > Linux should do the same, if possible as an optimization.
> >
> > Interesting. You could do easily the same in Linux by changing (in 2.2) the SAVE_ALL macro
> > in arch/i386/kernel/irq.h and doing the same in ret_from_intr's RESTORE_ALL macro (after
> > changing it to a RESTORE_ALL_INT or you'll break system calls)
> >
> > Actually I don't even know why irq.h's SAVE_ALL even loads __KERNEL_CS, it should be already
> > set by the interrupt gate. __KERNEL_DS probably needs to be still loaded, but only when
> > you came from user space.
> Sounds like a fun patch. I'll get to work on it. Might want to ping Linus,
> and let him put this one in, since it's his code, and would be a very easy
> fix. The more lipo-suction we could do, the better.
> :-)
> Jeff
> >
> > -Andi

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

 \ /
  Last update: 2005-03-22 12:45    [W:0.057 / U:2.720 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site