Re: Using __kuser_cmpxchg

[Andrew Haley <aph at redhat dot com>]

Matthieu CASTET writes:
 > Andrew Haley wrote:
 > > Nicolas Pitre writes:
 > >  > However, on a non SMP machine, the only thing that can break user
 > >  > space atomicity is a processor exception.  When an exception occurs
 > >  > because of an IRQ or a page fault, the kernel may well decide to
 > >  > schedule another user process or thread unexpectedly.
 > >  > 
 > >  > So we cheat a bit by simply performing the load+compare+store with 
 > >  > plain standard instructions as follows:
 > >  > 
 > >  >         teq     ip, ip                  @ set Z flag
 > >  >         ldr     ip, [r2]                @ load current val
 > >  >         add     r3, r2, #1              @ prepare store ptr
 > >  >         teqeq   ip, r0                  @ compare with oldval if still allowed
 > >  >         streq   r1, [r3, #-1]!          @ store newval if still allowed
 > >  >         subs    r0, r2, r3              @ if r2 == r3 the str occured
 > >  > 
 > >  > But there is a twist such that if ever an exception occurs during that 
 > >  > sequence, the Z flag is cleared by the exception handler before that 
 > >  > sequence is resumed and the store won't occur.
 > > 
 > > Ok, please stop there.  I don't understand this at all: if an
 > > exception occurs surely we'll save and restore the flags.  How is it
 > > possible for an exception to corrupt the user space Z flag?
 > > 
 > Look in arch/arm/kernel/entry-armv.S the usr_entry macro.
 > 
 > There are special code in the exception handler to detect if we come 
 > from __kuser_cmpxchg code and clear the Z flags.

Ah, gotcha.

	@ make sure our user space atomic helper is aborted
	cmp	r2, #TASK_SIZE
	bichs	r3, r3, #PSR_Z_BIT

Right, now I see.  The comment in the "Theory of operation" explained
that all along, but I didn't actually *get it*.  I am so used to the
idea that interrupts always preserve user space context -- no matter
what -- that I couldn't see what was before my eyes.  :-)

So, as Nicolas said, it's a matter of restarting the helper if an
exception occurs.

Thanks,
Andrew.

-- 
Red Hat UK Ltd, Amberley Place, 107-111 Peascod Street, Windsor, Berkshire, SL4 1TE, UK
Registered in England and Wales No. 3798903