The MSI Driver Guide HOWTO Tom L Nguyen tom.l.nguyen@intel.com 10/03/2003 Revised Feb 12, 2004 by Martine Silbermann email: Martine.Silbermann@hp.com Revised Jun 25, 2004 by Tom L Nguyen Revised Jul 9, 2008 by Matthew Wilcox Copyright 2003, 2008 Intel Corporation 1. About this guide This guide describes the basics of Message Signaled Interrupts (MSIs), the advantages of using MSI over traditional interrupt mechanisms, how to change your driver to use MSI or MSI-X and some basic diagnostics to try if a device doesn't support MSIs. 2. What are MSIs? A Message Signaled Interrupt is a write from the device to a special address which causes an interrupt to be received by the CPU. The MSI capability was first specified in PCI 2.2 and was later enhanced in PCI 3.0 to allow each interrupt to be masked individually. The MSI-X capability was also introduced with PCI 3.0. It supports more interrupts per device than MSI and allows interrupts to be independently configured. Devices may support both MSI and MSI-X, but only one can be enabled at a time. 3. Why use MSIs? There are three reasons why using MSIs can give an advantage over traditional pin-based interrupts. Pin-based PCI interrupts are often shared amongst several devices. To support this, the kernel must call each interrupt handler associated with an interrupt, which leads to reduced performance for the system as a whole. MSIs are never shared, so this problem cannot arise. When a device writes data to memory, then raises a pin-based interrupt, it is possible that the interrupt may arrive before all the data has arrived in memory (this becomes more likely with devices behind PCI-PCI bridges). In order to ensure that all the data has arrived in memory, the interrupt handler must read a register on the device which raised the interrupt. PCI transaction ordering rules require that all the data arrive in memory before the value may be returned from the register. Using MSIs avoids this problem as the interrupt-generating write cannot pass the data writes, so by the time the interrupt is raised, the driver knows that all the data has arrived in memory. PCI devices can only support a single pin-based interrupt per function. Often drivers have to query the device to find out what event has occurred, slowing down interrupt handling for the common case. With MSIs, a device can support more interrupts, allowing each interrupt to be specialised to a different purpose. One possible design gives infrequent conditions (such as errors) their own interrupt which allows the driver to handle the normal interrupt handling path more efficiently. Other possible designs include giving one interrupt to each packet queue in a network card or each port in a storage controller. 4. How to use MSIs PCI devices are initialised to use pin-based interrupts. The device driver has to set up the device to use MSI or MSI-X. Not all machines support MSIs correctly, and for those machines, the APIs described below will simply fail and the device will continue to use pin-based interrupts. 4.1 Include kernel support for MSIs To support MSI or MSI-X, the kernel must be built with the CONFIG_PCI_MSI option enabled. This option is only available on some architectures, and it may depend on some other options also being set. For example, on x86, you must also enable X86_UP_APIC or SMP in order to see the CONFIG_PCI_MSI option. 4.2 Using MSI Most of the hard work is done for the driver in the PCI layer. It simply has to request that the PCI layer set up the MSI capability for this device. 4.2.1 pci_enable_msi int pci_enable_msi(struct pci_dev *dev) A successful call allocates ONE interrupt to the device, regardless of how many MSIs the device supports. The device is switched from pin-based interrupt mode to MSI mode. The dev->irq number is changed to a new number which represents the message signaled interrupt; consequently, this function should be called before the driver calls request_irq(), because an MSI is delivered via a vector that is different from the vector of a pin-based interrupt. 4.2.2 pci_enable_msi_range int pci_enable_msi_range(struct pci_dev *dev, int minvec, int maxvec) This function allows a device driver to request any number of MSI interrupts within specified range from 'minvec' to 'maxvec'. If this function returns a positive number it indicates the number of MSI interrupts that have been successfully allocated. In this case the device is switched from pin-based interrupt mode to MSI mode and updates dev->irq to be the lowest of the new interrupts assigned to it. The other interrupts assigned to the device are in the range dev->irq to dev->irq + returned value - 1. Device driver can use the returned number of successfully allocated MSI interrupts to further allocate and initialize device resources. If this function returns a negative number, it indicates an error and the driver should not attempt to request any more MSI interrupts for this device. This function should be called before the driver calls request_irq(), because MSI interrupts are delivered via vectors that are different from the vector of a pin-based interrupt. It is ideal if drivers can cope with a variable number of MSI interrupts; there are many reasons why the platform may not be able to provide the exact number that a driver asks for. There could be devices that can not operate with just any number of MSI interrupts within a range. See chapter 4.3.1.3 to get the idea how to handle such devices for MSI-X - the same logic applies to MSI. 4.2.1.1 Maximum possible number of MSI interrupts The typical usage of MSI interrupts is to allocate as many vectors as possible, likely up to the limit returned by pci_msi_vec_count() function: static int foo_driver_enable_msi(struct pci_dev *pdev, int nvec) { return pci_enable_msi_range(pdev, 1, nvec); } Note the value of 'minvec' parameter is 1. As 'minvec' is inclusive, the value of 0 would be meaningless and could result in error. Some devices have a minimal limit on number of MSI interrupts. In this case the function could look like this: static int foo_driver_enable_msi(struct pci_dev *pdev, int nvec) { return pci_enable_msi_range(pdev, FOO_DRIVER_MINIMUM_NVEC, nvec); } 4.2.1.2 Exact number of MSI interrupts If a driver is unable or unwilling to deal with a variable number of MSI interrupts it could request a particular number of interrupts by passing that number to pci_enable_msi_range() function as both 'minvec' and 'maxvec' parameters: static int foo_driver_enable_msi(struct pci_dev *pdev, int nvec) { return pci_enable_msi_range(pdev, nvec, nvec); } Note, unlike pci_enable_msi_exact() function, which could be also used to enable a particular number of MSI-X interrupts, pci_enable_msi_range() returns either a negative errno or 'nvec' (not negative errno or 0 - as pci_enable_msi_exact() does). 4.2.1.3 Single MSI mode The most notorious example of the request type described above is enabling the single MSI mode for a device. It could be done by passing two 1s as 'minvec' and 'maxvec': static int foo_driver_enable_single_msi(struct pci_dev *pdev) { return pci_enable_msi_range(pdev, 1, 1); } Note, unlike pci_enable_msi() function, which could be also used to enable the single MSI mode, pci_enable_msi_range() returns either a negative errno or 1 (not negative errno or 0 - as pci_enable_msi() does). 4.2.3 pci_enable_msi_exact int pci_enable_msi_exact(struct pci_dev *dev, int nvec) This variation on pci_enable_msi_range() call allows a device driver to request exactly 'nvec' MSIs. If this function returns a negative number, it indicates an error and the driver should not attempt to request any more MSI interrupts for this device. By contrast with pci_enable_msi_range() function, pci_enable_msi_exact() returns zero in case of success, which indicates MSI interrupts have been successfully allocated. 4.2.4 pci_disable_msi void pci_disable_msi(struct pci_dev *dev) This function should be used to undo the effect of pci_enable_msi_range(). Calling it restores dev->irq to the pin-based interrupt number and frees the previously allocated MSIs. The interrupts may subsequently be assigned to another device, so drivers should not cache the value of dev->irq. Before calling this function, a device driver must always call free_irq() on any interrupt for which it previously called request_irq(). Failure to do so results in a BUG_ON(), leaving the device with MSI enabled and thus leaking its vector. 4.2.4 pci_msi_vec_count int pci_msi_vec_count(struct pci_dev *dev) This function could be used to retrieve the number of MSI vectors the device requested (via the Multiple Message Capable register). The MSI specification only allows the returned value to be a power of two, up to a maximum of 2^5 (32). If this function returns a negative number, it indicates the device is not capable of sending MSIs. If this function returns a positive number, it indicates the maximum number of MSI interrupt vectors that could be allocated. 4.3 Using MSI-X The MSI-X capability is much more flexible than the MSI capability. It supports up to 2048 interrupts, each of which can be controlled independently. To support this flexibility, drivers must use an array of `struct msix_entry': struct msix_entry { u16 vector; /* kernel uses to write alloc vector */ u16 entry; /* driver uses to specify entry */ }; This allows for the device to use these interrupts in a sparse fashion; for example, it could use interrupts 3 and 1027 and yet allocate only a two-element array. The driver is expected to fill in the 'entry' value in each element of the array to indicate for which entries the kernel should assign interrupts; it is invalid to fill in two entries with the same number. 4.3.1 pci_enable_msix_range int pci_enable_msix_range(struct pci_dev *dev, struct msix_entry *entries, int minvec, int maxvec) Calling this function asks the PCI subsystem to allocate any number of MSI-X interrupts within specified range from 'minvec' to 'maxvec'. The 'entries' argument is a pointer to an array of msix_entry structs which should be at least 'maxvec' entries in size. On success, the device is switched into MSI-X mode and the function returns the number of MSI-X interrupts that have been successfully allocated. In this case the 'vector' member in entries numbered from 0 to the returned value - 1 is populated with the interrupt number; the driver should then call request_irq() for each 'vector' that it decides to use. The device driver is responsible for keeping track of the interrupts assigned to the MSI-X vectors so it can free them again later. Device driver can use the returned number of successfully allocated MSI-X interrupts to further allocate and initialize device resources. If this function returns a negative number, it indicates an error and the driver should not attempt to allocate any more MSI-X interrupts for this device. This function, in contrast with pci_enable_msi_range(), does not adjust dev->irq. The device will not generate interrupts for this interrupt number once MSI-X is enabled. Device drivers should normally call this function once per device during the initialization phase. It is ideal if drivers can cope with a variable number of MSI-X interrupts; there are many reasons why the platform may not be able to provide the exact number that a driver asks for. There could be devices that can not operate with just any number of MSI-X interrupts within a range. E.g., an network adapter might need let's say four vectors per each queue it provides. Therefore, a number of MSI-X interrupts allocated should be a multiple of four. In this case interface pci_enable_msix_range() can not be used alone to request MSI-X interrupts (since it can allocate any number within the range, without any notion of the multiple of four) and the device driver should master a custom logic to request the required number of MSI-X interrupts. 4.3.1.1 Maximum possible number of MSI-X interrupts The typical usage of MSI-X interrupts is to allocate as many vectors as possible, likely up to the limit returned by pci_msix_vec_count() function: static int foo_driver_enable_msix(struct foo_adapter *adapter, int nvec) { return pci_enable_msix_range(adapter->pdev, adapter->msix_entries, 1, nvec); } Note the value of 'minvec' parameter is 1. As 'minvec' is inclusive, the value of 0 would be meaningless and could result in error. Some devices have a minimal limit on number of MSI-X interrupts. In this case the function could look like this: static int foo_driver_enable_msix(struct foo_adapter *adapter, int nvec) { return pci_enable_msix_range(adapter->pdev, adapter->msix_entries, FOO_DRIVER_MINIMUM_NVEC, nvec); } 4.3.1.2 Exact number of MSI-X interrupts If a driver is unable or unwilling to deal with a variable number of MSI-X interrupts it could request a particular number of interrupts by passing that number to pci_enable_msix_range() function as both 'minvec' and 'maxvec' parameters: static int foo_driver_enable_msix(struct foo_adapter *adapter, int nvec) { return pci_enable_msix_range(adapter->pdev, adapter->msix_entries, nvec, nvec); } Note, unlike pci_enable_msix_exact() function, which could be also used to enable a particular number of MSI-X interrupts, pci_enable_msix_range() returns either a negative errno or 'nvec' (not negative errno or 0 - as pci_enable_msix_exact() does). 4.3.1.3 Specific requirements to the number of MSI-X interrupts As noted above, there could be devices that can not operate with just any number of MSI-X interrupts within a range. E.g., let's assume a device that is only capable sending the number of MSI-X interrupts which is a power of two. A routine that enables MSI-X mode for such device might look like this: /* * Assume 'minvec' and 'maxvec' are non-zero */ static int foo_driver_enable_msix(struct foo_adapter *adapter, int minvec, int maxvec) { int rc; minvec = roundup_pow_of_two(minvec); maxvec = rounddown_pow_of_two(maxvec); if (minvec > maxvec) return -ERANGE; retry: rc = pci_enable_msix_range(adapter->pdev, adapter->msix_entries, maxvec, maxvec); /* * -ENOSPC is the only error code allowed to be analized */ if (rc == -ENOSPC) { if (maxvec == 1) return -ENOSPC; maxvec /= 2; if (minvec > maxvec) return -ENOSPC; goto retry; } return rc; } Note how pci_enable_msix_range() return value is analized for a fallback - any error code other than -ENOSPC indicates a fatal error and should not be retried. 4.3.2 pci_enable_msix_exact int pci_enable_msix_exact(struct pci_dev *dev, struct msix_entry *entries, int nvec) This variation on pci_enable_msix_range() call allows a device driver to request exactly 'nvec' MSI-Xs. If this function returns a negative number, it indicates an error and the driver should not attempt to allocate any more MSI-X interrupts for this device. By contrast with pci_enable_msix_range() function, pci_enable_msix_exact() returns zero in case of success, which indicates MSI-X interrupts have been successfully allocated. Another version of a routine that enables MSI-X mode for a device with specific requirements described in chapter 4.3.1.3 might look like this: /* * Assume 'minvec' and 'maxvec' are non-zero */ static int foo_driver_enable_msix(struct foo_adapter *adapter, int minvec, int maxvec) { int rc; minvec = roundup_pow_of_two(minvec); maxvec = rounddown_pow_of_two(maxvec); if (minvec > maxvec) return -ERANGE; retry: rc = pci_enable_msix_exact(adapter->pdev, adapter->msix_entries, maxvec); /* * -ENOSPC is the only error code allowed to be analyzed */ if (rc == -ENOSPC) { if (maxvec == 1) return -ENOSPC; maxvec /= 2; if (minvec > maxvec) return -ENOSPC; goto retry; } else if (rc < 0) { return rc; } return maxvec; } 4.3.3 pci_disable_msix void pci_disable_msix(struct pci_dev *dev) This function should be used to undo the effect of pci_enable_msix_range(). It frees the previously allocated MSI-X interrupts. The interrupts may subsequently be assigned to another device, so drivers should not cache the value of the 'vector' elements over a call to pci_disable_msix(). Before calling this function, a device driver must always call free_irq() on any interrupt for which it previously called request_irq(). Failure to do so results in a BUG_ON(), leaving the device with MSI-X enabled and thus leaking its vector. 4.3.3 The MSI-X Table The MSI-X capability specifies a BAR and offset within that BAR for the MSI-X Table. This address is mapped by the PCI subsystem, and should not be accessed directly by the device driver. If the driver wishes to mask or unmask an interrupt, it should call disable_irq() / enable_irq(). 4.3.4 pci_msix_vec_count int pci_msix_vec_count(struct pci_dev *dev) This function could be used to retrieve number of entries in the device MSI-X table. If this function returns a negative number, it indicates the device is not capable of sending MSI-Xs. If this function returns a positive number, it indicates the maximum number of MSI-X interrupt vectors that could be allocated. 4.4 Handling devices implementing both MSI and MSI-X capabilities If a device implements both MSI and MSI-X capabilities, it can run in either MSI mode or MSI-X mode, but not both simultaneously. This is a requirement of the PCI spec, and it is enforced by the PCI layer. Calling pci_enable_msi_range() when MSI-X is already enabled or pci_enable_msix_range() when MSI is already enabled results in an error. If a device driver wishes to switch between MSI and MSI-X at runtime, it must first quiesce the device, then switch it back to pin-interrupt mode, before calling pci_enable_msi_range() or pci_enable_msix_range() and resuming operation. This is not expected to be a common operation but may be useful for debugging or testing during development. 4.5 Considerations when using MSIs 4.5.1 Choosing between MSI-X and MSI If your device supports both MSI-X and MSI capabilities, you should use the MSI-X facilities in preference to the MSI facilities. As mentioned above, MSI-X supports any number of interrupts between 1 and 2048. In constrast, MSI is restricted to a maximum of 32 interrupts (and must be a power of two). In addition, the MSI interrupt vectors must be allocated consecutively, so the system might not be able to allocate as many vectors for MSI as it could for MSI-X. On some platforms, MSI interrupts must all be targeted at the same set of CPUs whereas MSI-X interrupts can all be targeted at different CPUs. 4.5.2 Spinlocks Most device drivers have a per-device spinlock which is taken in the interrupt handler. With pin-based interrupts or a single MSI, it is not necessary to disable interrupts (Linux guarantees the same interrupt will not be re-entered). If a device uses multiple interrupts, the driver must disable interrupts while the lock is held. If the device sends a different interrupt, the driver will deadlock trying to recursively acquire the spinlock. Such deadlocks can be avoided by using spin_lock_irqsave() or spin_lock_irq() which disable local interrupts and acquire the lock (see Documentation/DocBook/kernel-locking). 4.6 How to tell whether MSI/MSI-X is enabled on a device Using 'lspci -v' (as root) may show some devices with "MSI", "Message Signalled Interrupts" or "MSI-X" capabilities. Each of these capabilities has an 'Enable' flag which is followed with either "+" (enabled) or "-" (disabled). 5. MSI quirks Several PCI chipsets or devices are known not to support MSIs. The PCI stack provides three ways to disable MSIs: 1. globally 2. on all devices behind a specific bridge 3. on a single device 5.1. Disabling MSIs globally Some host chipsets simply don't support MSIs properly. If we're lucky, the manufacturer knows this and has indicated it in the ACPI FADT table. In this case, Linux automatically disables MSIs. Some boards don't include this information in the table and so we have to detect them ourselves. The complete list of these is found near the quirk_disable_all_msi() function in drivers/pci/quirks.c. If you have a board which has problems with MSIs, you can pass pci=nomsi on the kernel command line to disable MSIs on all devices. It would be in your best interests to report the problem to linux-pci@vger.kernel.org including a full 'lspci -v' so we can add the quirks to the kernel. 5.2. Disabling MSIs below a bridge Some PCI bridges are not able to route MSIs between busses properly. In this case, MSIs must be disabled on all devices behind the bridge. Some bridges allow you to enable MSIs by changing some bits in their PCI configuration space (especially the Hypertransport chipsets such as the nVidia nForce and Serverworks HT2000). As with host chipsets, Linux mostly knows about them and automatically enables MSIs if it can. If you have a bridge unknown to Linux, you can enable MSIs in configuration space using whatever method you know works, then enable MSIs on that bridge by doing: echo 1 > /sys/bus/pci/devices/$bridge/msi_bus where $bridge is the PCI address of the bridge you've enabled (eg 0000:00:0e.0). To disable MSIs, echo 0 instead of 1. Changing this value should be done with caution as it could break interrupt handling for all devices below this bridge. Again, please notify linux-pci@vger.kernel.org of any bridges that need special handling. 5.3. Disabling MSIs on a single device Some devices are known to have faulty MSI implementations. Usually this is handled in the individual device driver, but occasionally it's necessary to handle this with a quirk. Some drivers have an option to disable use of MSI. While this is a convenient workaround for the driver author, it is not good practice, and should not be emulated. 5.4. Finding why MSIs are disabled on a device From the above three sections, you can see that there are many reasons why MSIs may not be enabled for a given device. Your first step should be to examine your dmesg carefully to determine whether MSIs are enabled for your machine. You should also check your .config to be sure you have enabled CONFIG_PCI_MSI. Then, 'lspci -t' gives the list of bridges above a device. Reading /sys/bus/pci/devices/*/msi_bus will tell you whether MSIs are enabled (1) or disabled (0). If 0 is found in any of the msi_bus files belonging to bridges between the PCI root and the device, MSIs are disabled. It is also worth checking the device driver to see whether it supports MSIs. For example, it may contain calls to pci_enable_msi_range() or pci_enable_msix_range().