| Message ID | 1246991791-21741-3-git-send-email-glommer@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, Jul 07, 2009 at 02:36:31PM -0400, Glauber Costa wrote: > Signed-off-by: Glauber Costa <glommer@redhat.com> > --- > hw/msix.c | 14 ++++++++++++++ > 1 files changed, 14 insertions(+), 0 deletions(-) > > diff --git a/hw/msix.c b/hw/msix.c > index b6f3948..8636f69 100644 > --- a/hw/msix.c > +++ b/hw/msix.c > @@ -63,6 +63,7 @@ > /* Flag for interrupt controller to declare MSI-X support */ > int msix_supported; > > +#ifdef USE_KVM > /* KVM specific MSIX helpers */ > static void kvm_msix_free(PCIDevice *dev) > { > @@ -157,6 +158,14 @@ static void kvm_msix_del(PCIDevice *dev, unsigned vector) > kvm_del_routing_entry(kvm_context, &dev->msix_irq_entries[vector]); > kvm_commit_irq_routes(kvm_context); > } > +#else > + > +static void kvm_msix_free(PCIDevice *dev) {} > +static void kvm_msix_update(PCIDevice *dev, int vector, > + int was_masked, int is_masked) {} > +static int kvm_msix_add(PCIDevice *dev, unsigned vector) { return -1; } > +static void kvm_msix_del(PCIDevice *dev, unsigned vector) {} > +#endif > > /* Add MSI-X capability to the config space for the device. */ > /* Given a bar and its size, add MSI-X table on top of it Thanks for fix! > @@ -337,10 +346,12 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries, > if (nentries > MSIX_MAX_ENTRIES) > return -EINVAL; > > +#ifdef KVM_CAP_IRQCHIP > if (kvm_enabled() && qemu_kvm_irqchip_in_kernel()) { > dev->msix_irq_entries = qemu_malloc(nentries * > sizeof *dev->msix_irq_entries); > } > +#endif > dev->msix_entry_used = qemu_mallocz(MSIX_MAX_ENTRIES * > sizeof *dev->msix_entry_used); > > @@ -454,10 +465,13 @@ void msix_notify(PCIDevice *dev, unsigned vector) > msix_set_pending(dev, vector); > return; > } > + > +#ifdef KVM_CAP_IRQCHIP > if (kvm_enabled() && qemu_kvm_irqchip_in_kernel()) { > kvm_set_irq(dev->msix_irq_entries[vector].gsi, 1, NULL); > return; > } > +#endif > > address = pci_get_long(table_entry + MSIX_MSG_UPPER_ADDR); > address = (address << 32) | pci_get_long(table_entry + MSIX_MSG_ADDR); I think it's time we stopped worrying about builds against old kernel headers or without them. What do we gain from it? I believe that the right thing to do is to define kvm_enabled as a macro returning 0, and let compiler optimize the code out. Avi?
On Wed, Jul 08, 2009 at 10:59:13AM +0300, Michael S. Tsirkin wrote: > On Tue, Jul 07, 2009 at 02:36:31PM -0400, Glauber Costa wrote: > > Signed-off-by: Glauber Costa <glommer@redhat.com> > > --- > > hw/msix.c | 14 ++++++++++++++ > > 1 files changed, 14 insertions(+), 0 deletions(-) > > > > diff --git a/hw/msix.c b/hw/msix.c > > index b6f3948..8636f69 100644 > > --- a/hw/msix.c > > +++ b/hw/msix.c > > @@ -63,6 +63,7 @@ > > /* Flag for interrupt controller to declare MSI-X support */ > > int msix_supported; > > > > +#ifdef USE_KVM > > /* KVM specific MSIX helpers */ > > static void kvm_msix_free(PCIDevice *dev) > > { > > @@ -157,6 +158,14 @@ static void kvm_msix_del(PCIDevice *dev, unsigned vector) > > kvm_del_routing_entry(kvm_context, &dev->msix_irq_entries[vector]); > > kvm_commit_irq_routes(kvm_context); > > } > > +#else > > + > > +static void kvm_msix_free(PCIDevice *dev) {} > > +static void kvm_msix_update(PCIDevice *dev, int vector, > > + int was_masked, int is_masked) {} > > +static int kvm_msix_add(PCIDevice *dev, unsigned vector) { return -1; } > > +static void kvm_msix_del(PCIDevice *dev, unsigned vector) {} > > +#endif > > > > /* Add MSI-X capability to the config space for the device. */ > > /* Given a bar and its size, add MSI-X table on top of it > > Thanks for fix! > > > @@ -337,10 +346,12 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries, > > if (nentries > MSIX_MAX_ENTRIES) > > return -EINVAL; > > > > +#ifdef KVM_CAP_IRQCHIP > > if (kvm_enabled() && qemu_kvm_irqchip_in_kernel()) { > > dev->msix_irq_entries = qemu_malloc(nentries * > > sizeof *dev->msix_irq_entries); > > } > > +#endif > > dev->msix_entry_used = qemu_mallocz(MSIX_MAX_ENTRIES * > > sizeof *dev->msix_entry_used); > > > > @@ -454,10 +465,13 @@ void msix_notify(PCIDevice *dev, unsigned vector) > > msix_set_pending(dev, vector); > > return; > > } > > + > > +#ifdef KVM_CAP_IRQCHIP > > if (kvm_enabled() && qemu_kvm_irqchip_in_kernel()) { > > kvm_set_irq(dev->msix_irq_entries[vector].gsi, 1, NULL); > > return; > > } > > +#endif > > > > address = pci_get_long(table_entry + MSIX_MSG_UPPER_ADDR); > > address = (address << 32) | pci_get_long(table_entry + MSIX_MSG_ADDR); > > I think it's time we stopped worrying about builds against old kernel > headers or without them. What do we gain from it? > > I believe that the right thing to do is to define kvm_enabled as a macro > returning 0, and let compiler optimize the code out. that already happens with kvm_enabled(). I have no idea why the compiler do not rip of code when we also test for qemu_kvm_irqchip_in_kernel() or other things. what we could do, is to only test for qemu_kvm_irqchip_in_kernel(), using the hidden assumption that if kvm is not enabled, irqchip tests will always return false. It is a little bit messy, though. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 07/08/2009 03:09 PM, Glauber Costa wrote: > >> I believe that the right thing to do is to define kvm_enabled as a macro >> returning 0, and let compiler optimize the code out. >> > that already happens with kvm_enabled(). > I have no idea why the compiler do not rip of code when we also test for > qemu_kvm_irqchip_in_kernel() or other things. > > what we could do, is to only test for qemu_kvm_irqchip_in_kernel(), > using the hidden assumption that if kvm is not enabled, irqchip tests > will always return false. > > It is a little bit messy, though We should make the call unconditional and do the kvm_enabled/irqchip_in_kernel checks in the kvm-specific function. This way common code is only minimally affected.
On 07/08/2009 10:59 AM, Michael S. Tsirkin wrote: > >> @@ -337,10 +346,12 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries, >> if (nentries> MSIX_MAX_ENTRIES) >> return -EINVAL; >> >> +#ifdef KVM_CAP_IRQCHIP >> if (kvm_enabled()&& qemu_kvm_irqchip_in_kernel()) { >> dev->msix_irq_entries = qemu_malloc(nentries * >> sizeof *dev->msix_irq_entries); >> } >> +#endif >> dev->msix_entry_used = qemu_mallocz(MSIX_MAX_ENTRIES * >> sizeof *dev->msix_entry_used); >> >> @@ -454,10 +465,13 @@ void msix_notify(PCIDevice *dev, unsigned vector) >> msix_set_pending(dev, vector); >> return; >> } >> + >> +#ifdef KVM_CAP_IRQCHIP >> if (kvm_enabled()&& qemu_kvm_irqchip_in_kernel()) { >> kvm_set_irq(dev->msix_irq_entries[vector].gsi, 1, NULL); >> return; >> } >> +#endif >> >> address = pci_get_long(table_entry + MSIX_MSG_UPPER_ADDR); >> address = (address<< 32) | pci_get_long(table_entry + MSIX_MSG_ADDR); >> > > I think it's time we stopped worrying about builds against old kernel > headers or without them. What do we gain from it? > qemu upstream doesn't carry its own headers, so it we want to merge, we need to work against old headers. > I believe that the right thing to do is to define kvm_enabled as a macro > returning 0, and let compiler optimize the code out. > Doesn't work with -O0 (or if it does, we can't count on it).
On Wed, Jul 08, 2009 at 03:11:45PM +0300, Avi Kivity wrote: > On 07/08/2009 10:59 AM, Michael S. Tsirkin wrote: >> >>> @@ -337,10 +346,12 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries, >>> if (nentries> MSIX_MAX_ENTRIES) >>> return -EINVAL; >>> >>> +#ifdef KVM_CAP_IRQCHIP >>> if (kvm_enabled()&& qemu_kvm_irqchip_in_kernel()) { >>> dev->msix_irq_entries = qemu_malloc(nentries * >>> sizeof *dev->msix_irq_entries); >>> } >>> +#endif >>> dev->msix_entry_used = qemu_mallocz(MSIX_MAX_ENTRIES * >>> sizeof *dev->msix_entry_used); >>> >>> @@ -454,10 +465,13 @@ void msix_notify(PCIDevice *dev, unsigned vector) >>> msix_set_pending(dev, vector); >>> return; >>> } >>> + >>> +#ifdef KVM_CAP_IRQCHIP >>> if (kvm_enabled()&& qemu_kvm_irqchip_in_kernel()) { >>> kvm_set_irq(dev->msix_irq_entries[vector].gsi, 1, NULL); >>> return; >>> } >>> +#endif >>> >>> address = pci_get_long(table_entry + MSIX_MSG_UPPER_ADDR); >>> address = (address<< 32) | pci_get_long(table_entry + MSIX_MSG_ADDR); >>> >> >> I think it's time we stopped worrying about builds against old kernel >> headers or without them. What do we gain from it? >> > > qemu upstream doesn't carry its own headers, so it we want to merge, we > need to work against old headers. Was there ever discussion on this? I think the right thing to do is to add own headers to qemu upstream. >> I believe that the right thing to do is to define kvm_enabled as a macro >> returning 0, and let compiler optimize the code out. >> > > Doesn't work with -O0 (or if it does, we can't count on it). With -O0 you get a ton of dead code anyway. Who cares?
On 07/08/2009 03:13 PM, Michael S. Tsirkin wrote: >>> I think it's time we stopped worrying about builds against old kernel >>> headers or without them. What do we gain from it? >>> >>> >> qemu upstream doesn't carry its own headers, so it we want to merge, we >> need to work against old headers. >> > > Was there ever discussion on this? I think the right thing to do is to > add own headers to qemu upstream. > Discussion yes, conclusion no. >>> I believe that the right thing to do is to define kvm_enabled as a macro >>> returning 0, and let compiler optimize the code out. >>> >>> >> Doesn't work with -O0 (or if it does, we can't count on it). >> > > With -O0 you get a ton of dead code anyway. Who cares People who debug (though -O1 works fine most of the time). It also feels unclean to rely on optimization for correctness.
On Wed, Jul 08, 2009 at 03:20:02PM +0300, Avi Kivity wrote: > On 07/08/2009 03:13 PM, Michael S. Tsirkin wrote: >>>> I think it's time we stopped worrying about builds against old kernel >>>> headers or without them. What do we gain from it? >>>> >>>> >>> qemu upstream doesn't carry its own headers, so it we want to merge, we >>> need to work against old headers. >>> >> >> Was there ever discussion on this? I think the right thing to do is to >> add own headers to qemu upstream. >> > > Discussion yes, conclusion no. > >>>> I believe that the right thing to do is to define kvm_enabled as a macro >>>> returning 0, and let compiler optimize the code out. >>>> >>>> >>> Doesn't work with -O0 (or if it does, we can't count on it). >>> >> >> With -O0 you get a ton of dead code anyway. Who cares > > People who debug (though -O1 works fine most of the time). It also > feels unclean to rely on optimization for correctness. I do not propoise to rely on optimization for correctness. The code under if (0) will get compiled with -O0 (and we need to carry our headers for this) but will never run.
On Tue, Jul 07, 2009 at 02:36:31PM -0400, Glauber Costa wrote: > Signed-off-by: Glauber Costa <glommer@redhat.com> > --- > hw/msix.c | 14 ++++++++++++++ > 1 files changed, 14 insertions(+), 0 deletions(-) > > diff --git a/hw/msix.c b/hw/msix.c > index b6f3948..8636f69 100644 > --- a/hw/msix.c > +++ b/hw/msix.c > @@ -63,6 +63,7 @@ > /* Flag for interrupt controller to declare MSI-X support */ > int msix_supported; > > +#ifdef USE_KVM > /* KVM specific MSIX helpers */ > static void kvm_msix_free(PCIDevice *dev) > { > @@ -157,6 +158,14 @@ static void kvm_msix_del(PCIDevice *dev, unsigned vector) > kvm_del_routing_entry(kvm_context, &dev->msix_irq_entries[vector]); > kvm_commit_irq_routes(kvm_context); > } > +#else > + > +static void kvm_msix_free(PCIDevice *dev) {} > +static void kvm_msix_update(PCIDevice *dev, int vector, > + int was_masked, int is_masked) {} > +static int kvm_msix_add(PCIDevice *dev, unsigned vector) { return -1; } > +static void kvm_msix_del(PCIDevice *dev, unsigned vector) {} > +#endif > > /* Add MSI-X capability to the config space for the device. */ > /* Given a bar and its size, add MSI-X table on top of it Avi, let's apply this first bit so that people can build with kvm-disabled, and think a bit about the second one? > @@ -337,10 +346,12 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries, > if (nentries > MSIX_MAX_ENTRIES) > return -EINVAL; > > +#ifdef KVM_CAP_IRQCHIP > if (kvm_enabled() && qemu_kvm_irqchip_in_kernel()) { > dev->msix_irq_entries = qemu_malloc(nentries * > sizeof *dev->msix_irq_entries); > } > +#endif > dev->msix_entry_used = qemu_mallocz(MSIX_MAX_ENTRIES * > sizeof *dev->msix_entry_used); > > @@ -454,10 +465,13 @@ void msix_notify(PCIDevice *dev, unsigned vector) > msix_set_pending(dev, vector); > return; > } > + > +#ifdef KVM_CAP_IRQCHIP > if (kvm_enabled() && qemu_kvm_irqchip_in_kernel()) { > kvm_set_irq(dev->msix_irq_entries[vector].gsi, 1, NULL); > return; > } > +#endif > > address = pci_get_long(table_entry + MSIX_MSG_UPPER_ADDR); > address = (address << 32) | pci_get_long(table_entry + MSIX_MSG_ADDR); > -- > 1.6.2.2 > > -- > To unsubscribe from this list: send the line "unsubscribe kvm" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/hw/msix.c b/hw/msix.c index b6f3948..8636f69 100644 --- a/hw/msix.c +++ b/hw/msix.c @@ -63,6 +63,7 @@ /* Flag for interrupt controller to declare MSI-X support */ int msix_supported; +#ifdef USE_KVM /* KVM specific MSIX helpers */ static void kvm_msix_free(PCIDevice *dev) { @@ -157,6 +158,14 @@ static void kvm_msix_del(PCIDevice *dev, unsigned vector) kvm_del_routing_entry(kvm_context, &dev->msix_irq_entries[vector]); kvm_commit_irq_routes(kvm_context); } +#else + +static void kvm_msix_free(PCIDevice *dev) {} +static void kvm_msix_update(PCIDevice *dev, int vector, + int was_masked, int is_masked) {} +static int kvm_msix_add(PCIDevice *dev, unsigned vector) { return -1; } +static void kvm_msix_del(PCIDevice *dev, unsigned vector) {} +#endif /* Add MSI-X capability to the config space for the device. */ /* Given a bar and its size, add MSI-X table on top of it @@ -337,10 +346,12 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries, if (nentries > MSIX_MAX_ENTRIES) return -EINVAL; +#ifdef KVM_CAP_IRQCHIP if (kvm_enabled() && qemu_kvm_irqchip_in_kernel()) { dev->msix_irq_entries = qemu_malloc(nentries * sizeof *dev->msix_irq_entries); } +#endif dev->msix_entry_used = qemu_mallocz(MSIX_MAX_ENTRIES * sizeof *dev->msix_entry_used); @@ -454,10 +465,13 @@ void msix_notify(PCIDevice *dev, unsigned vector) msix_set_pending(dev, vector); return; } + +#ifdef KVM_CAP_IRQCHIP if (kvm_enabled() && qemu_kvm_irqchip_in_kernel()) { kvm_set_irq(dev->msix_irq_entries[vector].gsi, 1, NULL); return; } +#endif address = pci_get_long(table_entry + MSIX_MSG_UPPER_ADDR); address = (address << 32) | pci_get_long(table_entry + MSIX_MSG_ADDR);
Signed-off-by: Glauber Costa <glommer@redhat.com> --- hw/msix.c | 14 ++++++++++++++ 1 files changed, 14 insertions(+), 0 deletions(-)