Re: [PATCH LINUX] xen: event channel arrays are xen_ulong_t and not unsigned long

[Ian Campbell <Ian.Campbell@citrix.com>]

On Tue, 2013-02-19 at 17:11 +0000, Stefano Stabellini wrote:
> On Tue, 19 Feb 2013, Ian Campbell wrote:
> > On ARM we want these to be the same size on 32- and 64-bit.
> > 
> > This is an ABI change on ARM. X86 does not change.
> > 
> > Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
> > Cc: Jan Beulich <JBeulich@suse.com>
> > Cc: Keir (Xen.org) <keir@xen.org>
> > Cc: Tim Deegan <tim@xen.org>
> > Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
> > Cc: linux-arm-kernel@lists.infradead.org
> > Cc: xen-devel@lists.xen.org
> > ---
> > Changes since V1
> >   use find_first_set not __ffs
> >   fix some more unsigned long -> xen_ulong_t
> >   use more generic xchg_xen_ulong instead of ...read_evtchn...
> > ---
> >  arch/arm/include/asm/xen/events.h |   22 ++++++++
> >  arch/x86/include/asm/xen/events.h |    3 +
> >  drivers/xen/events.c              |  103 ++++++++++++++++++++-----------------
> >  include/xen/interface/xen.h       |    8 ++--
> >  4 files changed, 84 insertions(+), 52 deletions(-)
> 
> You might have to rebase this patch: it doesn't apply on Linux 3.8.

Will do.

> 
> 
> > diff --git a/arch/arm/include/asm/xen/events.h b/arch/arm/include/asm/xen/events.h
> > index 94b4e90..9bb5f50 100644
> > --- a/arch/arm/include/asm/xen/events.h
> > +++ b/arch/arm/include/asm/xen/events.h
> > @@ -15,4 +15,26 @@ static inline int xen_irqs_disabled(struct pt_regs *regs)
> >         return raw_irqs_disabled_flags(regs->ARM_cpsr);
> >  }
> > 
> > +/*
> > + * We cannot use xchg because it does not support 8-byte
> > + * values. However it is safe to use {ldr,dtd}exd directly because all
> > + * platforms which Xen can run on support those instructions.
> > + */
> > +static inline xen_ulong_t xchg_xen_ulong(xen_ulong_t *ptr, xen_ulong_t val)
> > +{
> > +       xen_ulong_t oldval;
> > +       unsigned int tmp;
> > +
> > +       wmb();
> > +       asm volatile("@ read_evtchn_pending_sel\n"
>                            ^ do we need this?

It means that the .s files (if you create them) and such are a bit more
readable, it's not that uncommon in the ARM inline assembly.

> > +               "1:     ldrexd  %0, %H0, [%3]\n"
> > +               "       strexd  %1, %2, %H2, [%3]\n"
> > +               "       teq     %1, #0\n"
> > +               "       bne     1b"
> > +               : "=&r" (oldval), "=&r" (tmp)
> > +               : "r" (val), "r" (ptr)
> > +               : "memory", "cc");
> > +       return oldval;
> > +}
> > +
> >  #endif /* _ASM_ARM_XEN_EVENTS_H */
> 
> [...]
> 
> > @@ -1295,18 +1306,14 @@ static void __xen_evtchn_do_upcall(void)
> >         unsigned count;
> > 
> >         do {
> > -               unsigned long pending_words;
> > +               xen_ulong_t pending_words;
> > 
> >                 vcpu_info->evtchn_upcall_pending = 0;
> > 
> >                 if (__this_cpu_inc_return(xed_nesting_count) - 1)
> >                         goto out;
> > 
> > -#ifndef CONFIG_X86 /* No need for a barrier -- XCHG is a barrier on x86. */
> > -               /* Clear master flag /before/ clearing selector flag. */
> > -               wmb();
> > -#endif
> 
> Even though I understand that moving wmb() into xchg_xen_ulong gets rid
> of an ugly ifndef, I am not sure whether it is a good thing from the
> code readability point of view. I'll let Konrad decide on this one.

I actually see this the other way -- this barrier is logically part of
the exchange operation, implicitly on x86 and explicitly on ARM. I will
add a comment here though, which I think is best of both.

> 
> 
> > -               pending_words = xchg(&vcpu_info->evtchn_pending_sel, 0);
> > +               pending_words = xchg_xen_ulong(&vcpu_info->evtchn_pending_sel, 0);
> > 
> >                 start_word_idx = __this_cpu_read(current_word_idx);
> >                 start_bit_idx = __this_cpu_read(current_bit_idx);
> > @@ -1314,8 +1321,8 @@ static void __xen_evtchn_do_upcall(void)
> >                 word_idx = start_word_idx;
> > 
> >                 for (i = 0; pending_words != 0; i++) {
> > -                       unsigned long pending_bits;
> > -                       unsigned long words;
> > +                       xen_ulong_t pending_bits;
> > +                       xen_ulong_t words;
> > 
> >                         words = MASK_LSBS(pending_words, word_idx);
> > 
> > @@ -1327,7 +1334,7 @@ static void __xen_evtchn_do_upcall(void)
> >                                 bit_idx = 0;
> >                                 continue;
> >                         }
> > -                       word_idx = __ffs(words);
> > +                       word_idx = find_first_bit(BM(&words), sizeof(words));
> > 
> >                         pending_bits = active_evtchns(cpu, s, word_idx);
> >                         bit_idx = 0; /* usually scan entire word from start */
> 
> is that because find_first_bit can actually cope with a bit number >=
> 32 and __ffs can't?

Yes.

> In that case it is worth adding a comment somewhere in this file,
> reminding people to only use bit operations that can handle size >
> unsigned long

Will do.