| Message ID | 1421312314-72330-2-git-send-email-borntraeger@de.ibm.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Delegated to: | Michael Ellerman |
| Headers | show |
On Thu, 2015-01-15 at 09:58 +0100, Christian Borntraeger wrote: > ACCESS_ONCE does not work reliably on non-scalar types. For > example gcc 4.6 and 4.7 might remove the volatile tag for such > accesses during the SRA (scalar replacement of aggregates) step > (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58145) > > Change the ppc/kvm code to replace ACCESS_ONCE with READ_ONCE. > > Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> > --- > arch/powerpc/kvm/book3s_hv_rm_xics.c | 8 ++++---- > arch/powerpc/kvm/book3s_xics.c | 16 ++++++++-------- > 2 files changed, 12 insertions(+), 12 deletions(-) > > diff --git a/arch/powerpc/kvm/book3s_hv_rm_xics.c b/arch/powerpc/kvm/book3s_hv_rm_xics.c > index 7b066f6..7c22997 100644 > --- a/arch/powerpc/kvm/book3s_hv_rm_xics.c > +++ b/arch/powerpc/kvm/book3s_hv_rm_xics.c > @@ -152,7 +152,7 @@ static void icp_rm_down_cppr(struct kvmppc_xics *xics, struct kvmppc_icp *icp, > * in virtual mode. > */ > do { > - old_state = new_state = ACCESS_ONCE(icp->state); > + old_state = new_state = READ_ONCE(icp->state); These are all icp->state. Which is a union, but it's only the size of unsigned long. So in practice there shouldn't be a bug here right? cheers
Am 16.01.2015 um 00:09 schrieb Michael Ellerman: > On Thu, 2015-01-15 at 09:58 +0100, Christian Borntraeger wrote: >> ACCESS_ONCE does not work reliably on non-scalar types. For >> example gcc 4.6 and 4.7 might remove the volatile tag for such >> accesses during the SRA (scalar replacement of aggregates) step >> (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58145) >> >> Change the ppc/kvm code to replace ACCESS_ONCE with READ_ONCE. >> >> Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> >> --- >> arch/powerpc/kvm/book3s_hv_rm_xics.c | 8 ++++---- >> arch/powerpc/kvm/book3s_xics.c | 16 ++++++++-------- >> 2 files changed, 12 insertions(+), 12 deletions(-) >> >> diff --git a/arch/powerpc/kvm/book3s_hv_rm_xics.c b/arch/powerpc/kvm/book3s_hv_rm_xics.c >> index 7b066f6..7c22997 100644 >> --- a/arch/powerpc/kvm/book3s_hv_rm_xics.c >> +++ b/arch/powerpc/kvm/book3s_hv_rm_xics.c >> @@ -152,7 +152,7 @@ static void icp_rm_down_cppr(struct kvmppc_xics *xics, struct kvmppc_icp *icp, >> * in virtual mode. >> */ >> do { >> - old_state = new_state = ACCESS_ONCE(icp->state); >> + old_state = new_state = READ_ONCE(icp->state); > > These are all icp->state. > > Which is a union, but it's only the size of unsigned long. So in practice there > shouldn't be a bug here right? This bug was that gcc lost the volatile tag when propagating aggregates to scalar types. So in theory a union could be affected. See the original problem ( http://marc.info/?i=54611D86.4040306%40de.ibm.com ) which happened on union ipte_control { unsigned long val; struct { unsigned long k : 1; unsigned long kh : 31; unsigned long kg : 32; }; }; Christian
diff --git a/arch/powerpc/kvm/book3s_hv_rm_xics.c b/arch/powerpc/kvm/book3s_hv_rm_xics.c index 7b066f6..7c22997 100644 --- a/arch/powerpc/kvm/book3s_hv_rm_xics.c +++ b/arch/powerpc/kvm/book3s_hv_rm_xics.c @@ -152,7 +152,7 @@ static void icp_rm_down_cppr(struct kvmppc_xics *xics, struct kvmppc_icp *icp, * in virtual mode. */ do { - old_state = new_state = ACCESS_ONCE(icp->state); + old_state = new_state = READ_ONCE(icp->state); /* Down_CPPR */ new_state.cppr = new_cppr; @@ -211,7 +211,7 @@ unsigned long kvmppc_rm_h_xirr(struct kvm_vcpu *vcpu) * pending priority */ do { - old_state = new_state = ACCESS_ONCE(icp->state); + old_state = new_state = READ_ONCE(icp->state); xirr = old_state.xisr | (((u32)old_state.cppr) << 24); if (!old_state.xisr) @@ -277,7 +277,7 @@ int kvmppc_rm_h_ipi(struct kvm_vcpu *vcpu, unsigned long server, * whenever the MFRR is made less favored. */ do { - old_state = new_state = ACCESS_ONCE(icp->state); + old_state = new_state = READ_ONCE(icp->state); /* Set_MFRR */ new_state.mfrr = mfrr; @@ -352,7 +352,7 @@ int kvmppc_rm_h_cppr(struct kvm_vcpu *vcpu, unsigned long cppr) icp_rm_clr_vcpu_irq(icp->vcpu); do { - old_state = new_state = ACCESS_ONCE(icp->state); + old_state = new_state = READ_ONCE(icp->state); reject = 0; new_state.cppr = cppr; diff --git a/arch/powerpc/kvm/book3s_xics.c b/arch/powerpc/kvm/book3s_xics.c index 807351f..a4a8d9f 100644 --- a/arch/powerpc/kvm/book3s_xics.c +++ b/arch/powerpc/kvm/book3s_xics.c @@ -327,7 +327,7 @@ static bool icp_try_to_deliver(struct kvmppc_icp *icp, u32 irq, u8 priority, icp->server_num); do { - old_state = new_state = ACCESS_ONCE(icp->state); + old_state = new_state = READ_ONCE(icp->state); *reject = 0; @@ -512,7 +512,7 @@ static void icp_down_cppr(struct kvmppc_xics *xics, struct kvmppc_icp *icp, * in virtual mode. */ do { - old_state = new_state = ACCESS_ONCE(icp->state); + old_state = new_state = READ_ONCE(icp->state); /* Down_CPPR */ new_state.cppr = new_cppr; @@ -567,7 +567,7 @@ static noinline unsigned long kvmppc_h_xirr(struct kvm_vcpu *vcpu) * pending priority */ do { - old_state = new_state = ACCESS_ONCE(icp->state); + old_state = new_state = READ_ONCE(icp->state); xirr = old_state.xisr | (((u32)old_state.cppr) << 24); if (!old_state.xisr) @@ -634,7 +634,7 @@ static noinline int kvmppc_h_ipi(struct kvm_vcpu *vcpu, unsigned long server, * whenever the MFRR is made less favored. */ do { - old_state = new_state = ACCESS_ONCE(icp->state); + old_state = new_state = READ_ONCE(icp->state); /* Set_MFRR */ new_state.mfrr = mfrr; @@ -679,7 +679,7 @@ static int kvmppc_h_ipoll(struct kvm_vcpu *vcpu, unsigned long server) if (!icp) return H_PARAMETER; } - state = ACCESS_ONCE(icp->state); + state = READ_ONCE(icp->state); kvmppc_set_gpr(vcpu, 4, ((u32)state.cppr << 24) | state.xisr); kvmppc_set_gpr(vcpu, 5, state.mfrr); return H_SUCCESS; @@ -721,7 +721,7 @@ static noinline void kvmppc_h_cppr(struct kvm_vcpu *vcpu, unsigned long cppr) BOOK3S_INTERRUPT_EXTERNAL_LEVEL); do { - old_state = new_state = ACCESS_ONCE(icp->state); + old_state = new_state = READ_ONCE(icp->state); reject = 0; new_state.cppr = cppr; @@ -885,7 +885,7 @@ static int xics_debug_show(struct seq_file *m, void *private) if (!icp) continue; - state.raw = ACCESS_ONCE(icp->state.raw); + state.raw = READ_ONCE(icp->state.raw); seq_printf(m, "cpu server %#lx XIRR:%#x PPRI:%#x CPPR:%#x MFRR:%#x OUT:%d NR:%d\n", icp->server_num, state.xisr, state.pending_pri, state.cppr, state.mfrr, @@ -1082,7 +1082,7 @@ int kvmppc_xics_set_icp(struct kvm_vcpu *vcpu, u64 icpval) * the ICS states before the ICP states. */ do { - old_state = ACCESS_ONCE(icp->state); + old_state = READ_ONCE(icp->state); if (new_state.mfrr <= old_state.mfrr) { resend = false;
ACCESS_ONCE does not work reliably on non-scalar types. For example gcc 4.6 and 4.7 might remove the volatile tag for such accesses during the SRA (scalar replacement of aggregates) step (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58145) Change the ppc/kvm code to replace ACCESS_ONCE with READ_ONCE. Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> --- arch/powerpc/kvm/book3s_hv_rm_xics.c | 8 ++++---- arch/powerpc/kvm/book3s_xics.c | 16 ++++++++-------- 2 files changed, 12 insertions(+), 12 deletions(-)