| Message ID | 20170302232104.10136-1-andi@firstfloor.org (mailing list archive) |
|---|---|
| State | New, archived |
| Delegated to: | Bjorn Helgaas |
| Headers | show |
On Thu, 2 Mar 2017, Andi Kleen wrote: > From: Andi Kleen <ak@linux.intel.com> > > The Intel uncore driver can do a lot of PCI config accesses to read > performance counters. I had a situation on a 4S system where it > was spending 40+% of CPU time grabbing the pci_cfg_lock due to that. > > For 64bit x86 with MMCONFIG there isn't really any reason to take > a lock. The access is directly mapped to an underlying MMIO area, > which can fully operate lockless. > > Add a new flag that allows the PCI mid layer to skip the lock > and set it for the 64bit mmconfig code. > > There's a small risk that someone relies on this lock for synchronization, > but I think that's unlikely because there isn't really any useful > synchronization at this individual operation level. Any useful > synchronization would likely need to protect at least a > read-modify-write or similar. So I made it unconditional without opt-in. This part of the changelog is just crap. The reason why pci_lock exists and is taken for each single read/write config is that some ops implementations, e.g. the generic ones, must protect at this granularity level because ops->map_bus() read/writeX() needs to be 'atomic'. MMCONFIG obviously does not require this at all because it's a simple byte/word/dword read/write which is serialized by itself. So it's obvious that the serialization with pci_lock is pointless in this case. It's not that hard to figure it out and write up a proper changelog instead of handwaving about risk and whatever. Thanks, tglx
On 03/02/17 15:21, Andi Kleen wrote: > From: Andi Kleen <ak@linux.intel.com> > > The Intel uncore driver can do a lot of PCI config accesses to read > performance counters. I had a situation on a 4S system where it > was spending 40+% of CPU time grabbing the pci_cfg_lock due to that. > > For 64bit x86 with MMCONFIG there isn't really any reason to take > a lock. The access is directly mapped to an underlying MMIO area, > which can fully operate lockless. > > Add a new flag that allows the PCI mid layer to skip the lock > and set it for the 64bit mmconfig code. > > There's a small risk that someone relies on this lock for synchronization, > but I think that's unlikely because there isn't really any useful > synchronization at this individual operation level. Any useful > synchronization would likely need to protect at least a > read-modify-write or similar. So I made it unconditional without opt-in. > > Signed-off-by: Andi Kleen <ak@linux.intel.com> > --- > arch/x86/pci/mmconfig_64.c | 1 + > drivers/pci/access.c | 14 ++++++++++---- > include/linux/pci.h | 2 ++ > 3 files changed, 13 insertions(+), 4 deletions(-) > > diff --git a/arch/x86/pci/mmconfig_64.c b/arch/x86/pci/mmconfig_64.c > index bea52496aea6..8bf10f41e626 100644 > --- a/arch/x86/pci/mmconfig_64.c > +++ b/arch/x86/pci/mmconfig_64.c > @@ -121,6 +121,7 @@ int __init pci_mmcfg_arch_init(void) > } > > raw_pci_ext_ops = &pci_mmcfg; > + pci_root_ops.ll_allowed = true; > "ll_allowed" is pretty awful naming... you spend almost all the characters telling us nothing. I spend several seconds trying to figure out what "ll" stood for, and without the context of the patch I'd have had to go a massive grep. Just call it "lockless" or something. -hpa
diff --git a/arch/x86/pci/mmconfig_64.c b/arch/x86/pci/mmconfig_64.c index bea52496aea6..8bf10f41e626 100644 --- a/arch/x86/pci/mmconfig_64.c +++ b/arch/x86/pci/mmconfig_64.c @@ -121,6 +121,7 @@ int __init pci_mmcfg_arch_init(void) } raw_pci_ext_ops = &pci_mmcfg; + pci_root_ops.ll_allowed = true; return 1; } diff --git a/drivers/pci/access.c b/drivers/pci/access.c index db239547fefd..22552c6606c1 100644 --- a/drivers/pci/access.c +++ b/drivers/pci/access.c @@ -32,11 +32,14 @@ int pci_bus_read_config_##size \ int res; \ unsigned long flags; \ u32 data = 0; \ + bool ll_allowed = bus->ops->ll_allowed; \ if (PCI_##size##_BAD) return PCIBIOS_BAD_REGISTER_NUMBER; \ - raw_spin_lock_irqsave(&pci_lock, flags); \ + if (!ll_allowed) \ + raw_spin_lock_irqsave(&pci_lock, flags); \ res = bus->ops->read(bus, devfn, pos, len, &data); \ *value = (type)data; \ - raw_spin_unlock_irqrestore(&pci_lock, flags); \ + if (!ll_allowed) \ + raw_spin_unlock_irqrestore(&pci_lock, flags); \ return res; \ } @@ -46,10 +49,13 @@ int pci_bus_write_config_##size \ { \ int res; \ unsigned long flags; \ + bool ll_allowed = bus->ops->ll_allowed; \ if (PCI_##size##_BAD) return PCIBIOS_BAD_REGISTER_NUMBER; \ - raw_spin_lock_irqsave(&pci_lock, flags); \ + if (!ll_allowed) \ + raw_spin_lock_irqsave(&pci_lock, flags); \ res = bus->ops->write(bus, devfn, pos, len, value); \ - raw_spin_unlock_irqrestore(&pci_lock, flags); \ + if (!ll_allowed) \ + raw_spin_unlock_irqrestore(&pci_lock, flags); \ return res; \ } diff --git a/include/linux/pci.h b/include/linux/pci.h index e2d1a124216a..9b234cbc7ae1 100644 --- a/include/linux/pci.h +++ b/include/linux/pci.h @@ -612,6 +612,8 @@ struct pci_ops { void __iomem *(*map_bus)(struct pci_bus *bus, unsigned int devfn, int where); int (*read)(struct pci_bus *bus, unsigned int devfn, int where, int size, u32 *val); int (*write)(struct pci_bus *bus, unsigned int devfn, int where, int size, u32 val); + /* Set to true when pci_lock is not needed for read/write */ + bool ll_allowed; }; /*