298 lines · plain
1.. SPDX-License-Identifier: GPL-2.02.. include:: <isonum.txt>3 4==========================5The MSI Driver Guide HOWTO6==========================7 8:Authors: Tom L Nguyen; Martine Silbermann; Matthew Wilcox9 10:Copyright: 2003, 2008 Intel Corporation11 12About this guide13================14 15This guide describes the basics of Message Signaled Interrupts (MSIs),16the advantages of using MSI over traditional interrupt mechanisms, how17to change your driver to use MSI or MSI-X and some basic diagnostics to18try if a device doesn't support MSIs.19 20 21What are MSIs?22==============23 24A Message Signaled Interrupt is a write from the device to a special25address which causes an interrupt to be received by the CPU.26 27The MSI capability was first specified in PCI 2.2 and was later enhanced28in PCI 3.0 to allow each interrupt to be masked individually. The MSI-X29capability was also introduced with PCI 3.0. It supports more interrupts30per device than MSI and allows interrupts to be independently configured.31 32Devices may support both MSI and MSI-X, but only one can be enabled at33a time.34 35 36Why use MSIs?37=============38 39There are three reasons why using MSIs can give an advantage over40traditional pin-based interrupts.41 42Pin-based PCI interrupts are often shared amongst several devices.43To support this, the kernel must call each interrupt handler associated44with an interrupt, which leads to reduced performance for the system as45a whole. MSIs are never shared, so this problem cannot arise.46 47When a device writes data to memory, then raises a pin-based interrupt,48it is possible that the interrupt may arrive before all the data has49arrived in memory (this becomes more likely with devices behind PCI-PCI50bridges). In order to ensure that all the data has arrived in memory,51the interrupt handler must read a register on the device which raised52the interrupt. PCI transaction ordering rules require that all the data53arrive in memory before the value may be returned from the register.54Using MSIs avoids this problem as the interrupt-generating write cannot55pass the data writes, so by the time the interrupt is raised, the driver56knows that all the data has arrived in memory.57 58PCI devices can only support a single pin-based interrupt per function.59Often drivers have to query the device to find out what event has60occurred, slowing down interrupt handling for the common case. With61MSIs, a device can support more interrupts, allowing each interrupt62to be specialised to a different purpose. One possible design gives63infrequent conditions (such as errors) their own interrupt which allows64the driver to handle the normal interrupt handling path more efficiently.65Other possible designs include giving one interrupt to each packet queue66in a network card or each port in a storage controller.67 68 69How to use MSIs70===============71 72PCI devices are initialised to use pin-based interrupts. The device73driver has to set up the device to use MSI or MSI-X. Not all machines74support MSIs correctly, and for those machines, the APIs described below75will simply fail and the device will continue to use pin-based interrupts.76 77Include kernel support for MSIs78-------------------------------79 80To support MSI or MSI-X, the kernel must be built with the CONFIG_PCI_MSI81option enabled. This option is only available on some architectures,82and it may depend on some other options also being set. For example,83on x86, you must also enable X86_UP_APIC or SMP in order to see the84CONFIG_PCI_MSI option.85 86Using MSI87---------88 89Most of the hard work is done for the driver in the PCI layer. The driver90simply has to request that the PCI layer set up the MSI capability for this91device.92 93To automatically use MSI or MSI-X interrupt vectors, use the following94function::95 96 int pci_alloc_irq_vectors(struct pci_dev *dev, unsigned int min_vecs,97 unsigned int max_vecs, unsigned int flags);98 99which allocates up to max_vecs interrupt vectors for a PCI device. It100returns the number of vectors allocated or a negative error. If the device101has a requirements for a minimum number of vectors the driver can pass a102min_vecs argument set to this limit, and the PCI core will return -ENOSPC103if it can't meet the minimum number of vectors.104 105The flags argument is used to specify which type of interrupt can be used106by the device and the driver (PCI_IRQ_INTX, PCI_IRQ_MSI, PCI_IRQ_MSIX).107A convenient short-hand (PCI_IRQ_ALL_TYPES) is also available to ask for108any possible kind of interrupt. If the PCI_IRQ_AFFINITY flag is set,109pci_alloc_irq_vectors() will spread the interrupts around the available CPUs.110 111To get the Linux IRQ numbers passed to request_irq() and free_irq() and the112vectors, use the following function::113 114 int pci_irq_vector(struct pci_dev *dev, unsigned int nr);115 116Any allocated resources should be freed before removing the device using117the following function::118 119 void pci_free_irq_vectors(struct pci_dev *dev);120 121If a device supports both MSI-X and MSI capabilities, this API will use the122MSI-X facilities in preference to the MSI facilities. MSI-X supports any123number of interrupts between 1 and 2048. In contrast, MSI is restricted to124a maximum of 32 interrupts (and must be a power of two). In addition, the125MSI interrupt vectors must be allocated consecutively, so the system might126not be able to allocate as many vectors for MSI as it could for MSI-X. On127some platforms, MSI interrupts must all be targeted at the same set of CPUs128whereas MSI-X interrupts can all be targeted at different CPUs.129 130If a device supports neither MSI-X or MSI it will fall back to a single131legacy IRQ vector.132 133The typical usage of MSI or MSI-X interrupts is to allocate as many vectors134as possible, likely up to the limit supported by the device. If nvec is135larger than the number supported by the device it will automatically be136capped to the supported limit, so there is no need to query the number of137vectors supported beforehand::138 139 nvec = pci_alloc_irq_vectors(pdev, 1, nvec, PCI_IRQ_ALL_TYPES)140 if (nvec < 0)141 goto out_err;142 143If a driver is unable or unwilling to deal with a variable number of MSI144interrupts it can request a particular number of interrupts by passing that145number to pci_alloc_irq_vectors() function as both 'min_vecs' and146'max_vecs' parameters::147 148 ret = pci_alloc_irq_vectors(pdev, nvec, nvec, PCI_IRQ_ALL_TYPES);149 if (ret < 0)150 goto out_err;151 152The most notorious example of the request type described above is enabling153the single MSI mode for a device. It could be done by passing two 1s as154'min_vecs' and 'max_vecs'::155 156 ret = pci_alloc_irq_vectors(pdev, 1, 1, PCI_IRQ_ALL_TYPES);157 if (ret < 0)158 goto out_err;159 160Some devices might not support using legacy line interrupts, in which case161the driver can specify that only MSI or MSI-X is acceptable::162 163 nvec = pci_alloc_irq_vectors(pdev, 1, nvec, PCI_IRQ_MSI | PCI_IRQ_MSIX);164 if (nvec < 0)165 goto out_err;166 167Legacy APIs168-----------169 170The following old APIs to enable and disable MSI or MSI-X interrupts should171not be used in new code::172 173 pci_enable_msi() /* deprecated */174 pci_disable_msi() /* deprecated */175 pci_enable_msix_range() /* deprecated */176 pci_enable_msix_exact() /* deprecated */177 pci_disable_msix() /* deprecated */178 179Additionally there are APIs to provide the number of supported MSI or MSI-X180vectors: pci_msi_vec_count() and pci_msix_vec_count(). In general these181should be avoided in favor of letting pci_alloc_irq_vectors() cap the182number of vectors. If you have a legitimate special use case for the count183of vectors we might have to revisit that decision and add a184pci_nr_irq_vectors() helper that handles MSI and MSI-X transparently.185 186Considerations when using MSIs187------------------------------188 189Spinlocks190~~~~~~~~~191 192Most device drivers have a per-device spinlock which is taken in the193interrupt handler. With pin-based interrupts or a single MSI, it is not194necessary to disable interrupts (Linux guarantees the same interrupt will195not be re-entered). If a device uses multiple interrupts, the driver196must disable interrupts while the lock is held. If the device sends197a different interrupt, the driver will deadlock trying to recursively198acquire the spinlock. Such deadlocks can be avoided by using199spin_lock_irqsave() or spin_lock_irq() which disable local interrupts200and acquire the lock (see Documentation/kernel-hacking/locking.rst).201 202How to tell whether MSI/MSI-X is enabled on a device203----------------------------------------------------204 205Using 'lspci -v' (as root) may show some devices with "MSI", "Message206Signalled Interrupts" or "MSI-X" capabilities. Each of these capabilities207has an 'Enable' flag which is followed with either "+" (enabled)208or "-" (disabled).209 210 211MSI quirks212==========213 214Several PCI chipsets or devices are known not to support MSIs.215The PCI stack provides three ways to disable MSIs:216 2171. globally2182. on all devices behind a specific bridge2193. on a single device220 221Disabling MSIs globally222-----------------------223 224Some host chipsets simply don't support MSIs properly. If we're225lucky, the manufacturer knows this and has indicated it in the ACPI226FADT table. In this case, Linux automatically disables MSIs.227Some boards don't include this information in the table and so we have228to detect them ourselves. The complete list of these is found near the229quirk_disable_all_msi() function in drivers/pci/quirks.c.230 231If you have a board which has problems with MSIs, you can pass pci=nomsi232on the kernel command line to disable MSIs on all devices. It would be233in your best interests to report the problem to linux-pci@vger.kernel.org234including a full 'lspci -v' so we can add the quirks to the kernel.235 236Disabling MSIs below a bridge237-----------------------------238 239Some PCI bridges are not able to route MSIs between buses properly.240In this case, MSIs must be disabled on all devices behind the bridge.241 242Some bridges allow you to enable MSIs by changing some bits in their243PCI configuration space (especially the Hypertransport chipsets such244as the nVidia nForce and Serverworks HT2000). As with host chipsets,245Linux mostly knows about them and automatically enables MSIs if it can.246If you have a bridge unknown to Linux, you can enable247MSIs in configuration space using whatever method you know works, then248enable MSIs on that bridge by doing::249 250 echo 1 > /sys/bus/pci/devices/$bridge/msi_bus251 252where $bridge is the PCI address of the bridge you've enabled (eg2530000:00:0e.0).254 255To disable MSIs, echo 0 instead of 1. Changing this value should be256done with caution as it could break interrupt handling for all devices257below this bridge.258 259Again, please notify linux-pci@vger.kernel.org of any bridges that need260special handling.261 262Disabling MSIs on a single device263---------------------------------264 265Some devices are known to have faulty MSI implementations. Usually this266is handled in the individual device driver, but occasionally it's necessary267to handle this with a quirk. Some drivers have an option to disable use268of MSI. While this is a convenient workaround for the driver author,269it is not good practice, and should not be emulated.270 271Finding why MSIs are disabled on a device272-----------------------------------------273 274From the above three sections, you can see that there are many reasons275why MSIs may not be enabled for a given device. Your first step should276be to examine your dmesg carefully to determine whether MSIs are enabled277for your machine. You should also check your .config to be sure you278have enabled CONFIG_PCI_MSI.279 280Then, 'lspci -t' gives the list of bridges above a device. Reading281`/sys/bus/pci/devices/*/msi_bus` will tell you whether MSIs are enabled (1)282or disabled (0). If 0 is found in any of the msi_bus files belonging283to bridges between the PCI root and the device, MSIs are disabled.284 285It is also worth checking the device driver to see whether it supports MSIs.286For example, it may contain calls to pci_alloc_irq_vectors() with the287PCI_IRQ_MSI or PCI_IRQ_MSIX flags.288 289 290List of device drivers MSI(-X) APIs291===================================292 293The PCI/MSI subsystem has a dedicated C file for its exported device driver294APIs — `drivers/pci/msi/api.c`. The following functions are exported:295 296.. kernel-doc:: drivers/pci/msi/api.c297 :export:298