Re: [Qemu-devel] [PATCHv3 3/9] pseries: Clean up hash page table allocation error handling

From: David Gibson <david@gibson.dropbear.id.au>
To: Markus Armbruster <armbru@redhat.com>
Cc: lvivier@redhat.com, Thomas Huth <thuth@redhat.com>,
	aik@ozlabs.ru, mdroth@linux.vnet.ibm.com, qemu-devel@nongnu.org,
	qemu-ppc@nongnu.org
Subject: Re: [Qemu-devel] [PATCHv3 3/9] pseries: Clean up hash page table allocation error handling
Date: Tue, 19 Jan 2016 12:12:10 +1100	[thread overview]
Message-ID: <20160119011210.GW9301@voom.fritz.box> (raw)
In-Reply-To: <874mebqjcb.fsf@blackfin.pond.sub.org>

[-- Attachment #1: Type: text/plain, Size: 3954 bytes --]

On Mon, Jan 18, 2016 at 11:21:08AM +0100, Markus Armbruster wrote:
> Thomas Huth <thuth@redhat.com> writes:
> 
> > On 18.01.2016 05:24, David Gibson wrote:
> >> The spapr_alloc_htab() and spapr_reset_htab() functions currently handle
> >> all errors with error_setg(&error_abort, ...).
> >> 
> >> But really, the callers are really better placed to decide on the error
> >> handling.  So, instead make the functions use the error propagation
> >> infrastructure.
> >> 
> >> In the callers we change to &error_fatal instead of &error_abort, since
> >> this can be triggered by a bad configuration or kernel error rather than
> >> indicating a programming error in qemu.
> >> 
> >> While we're at it improve the messages themselves a bit, and clean up the
> >> indentation a little.
> >> 
> >> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> >> ---
> >>  hw/ppc/spapr.c | 24 ++++++++++++++++--------
> >>  1 file changed, 16 insertions(+), 8 deletions(-)
> >> 
> >> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
> >> index b7fd09a..d28e349 100644
> >> --- a/hw/ppc/spapr.c
> >> +++ b/hw/ppc/spapr.c
> >> @@ -1016,7 +1016,7 @@ static void emulate_spapr_hypercall(PowerPCCPU *cpu)
> >>  #define CLEAN_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) &= tswap64(~HPTE64_V_HPTE_DIRTY))
> >>  #define DIRTY_HPTE(_hpte)  ((*(uint64_t *)(_hpte)) |= tswap64(HPTE64_V_HPTE_DIRTY))
> >>  
> >> -static void spapr_alloc_htab(sPAPRMachineState *spapr)
> >> +static void spapr_alloc_htab(sPAPRMachineState *spapr, Error **errp)
> >>  {
> >>      long shift;
> >>      int index;
> >> @@ -1031,7 +1031,8 @@ static void spapr_alloc_htab(sPAPRMachineState *spapr)
> >>           * For HV KVM, host kernel will return -ENOMEM when requested
> >>           * HTAB size can't be allocated.
> >>           */
> >> -        error_setg(&error_abort, "Failed to allocate HTAB of requested size, try with smaller maxmem");
> >> +        error_setg_errno(errp, -shift,
> >> +                         "Error allocating KVM hash page table, try smaller maxmem");
> >>      } else if (shift > 0) {
> >>          /*
> >>           * Kernel handles htab, we don't need to allocate one
> >> @@ -1040,7 +1041,10 @@ static void spapr_alloc_htab(sPAPRMachineState *spapr)
> >>           * but we don't allow booting of such guests.
> >>           */
> >>          if (shift != spapr->htab_shift) {
> >> -            error_setg(&error_abort, "Failed to allocate HTAB of requested size, try with smaller maxmem");
> >> +            error_setg(errp,
> >> +                "Small allocation for KVM hash page table (%ld < %"
> >> +                PRIu32 "), try smaller maxmem",
> >> +                shift, spapr->htab_shift);
> >
> > Maybe you should add an "return" statement here - theoretically you do
> > not want to continue with "kvmppc_kern_htab = true" in case of errors.
> > (practically this does not happen because errp = error_fatal, but in
> > case the caller gets changed, this might introduce subtle errors otherwise)
> 
> Good point.
> 
> With abort() / exit(), we don't have to worry about recovery.  In
> particular, we don't have to revert half-done changes.
> 
> Conversions away from abort() / exit() need to consider error recovery.
> We have to make sure the function leaves things in a sane state on
> error.  This normally means taking an early return, and often means
> reverting some state changes.

That's true, but Thomas is mistaken about what error recovery is
needed here.

However, I'm going to drop this patch from the series anyway - I've
realised I need to rework the htab allocation substantially for other
reasons, so it would be better to not have that conflict with this
series.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]