MSI_MSI-X中断之源码分析

---

---

---

开发板资料：

• 开发板Firefly-rk3399资料

参考内核文件：

• Documentation\PCI\MSI-HOWTO.txt
• drivers\pci\host\pcie-rockchip.c
• drivers\nvme\host\pci.c
• drivers\irqchip\irq-gic-v3.c
• drivers\irqchip\irq-gic-v3-its.c
• drivers\irqchip\irq-gic-v3-its-pci-msi.c

一、 怎么发出MSI/MSI-X中断

PCIe设备向发出中断，它发出TLP包，往某个地址写入某个数据即可：

• 往哪个地址？GICv3 ITS的GITS_TRANSLATER寄存器，TLP包里使用的是PCI地址
• 写什么数据？0、1、2、……，要发出第1个中断时写0，要发出第2个中断时写1，……

在设备树文件rk3399.dtsi中，可以看到ITS的基地址是0xfee20000：

its: interrupt-controller@fee20000 {compatible = "arm,gic-v3-its";msi-controller;reg = <0x0 0xfee20000 0x0 0x20000>;};

在IHI0069G_gic_architecture_specification.pdf中有ITS寄存器的偏移地址：

GITS_TRANSLATER寄存器的CPU地址是：0xfee20000 + 0x010000 + 0x0040 = 0xfee30040。

对应的PCI地址也是0xfee30040(驱动程序里为例方便，故意使得CPU地址跟PCI地址相同，这2个地址属于不同地址空间)，

所以下图中PCI地址都是0xfee30040。

1.1 在RK3399上体验

在RK3399开发板，插上了NVMe SSD固态硬盘。

1.1.1 安装工具

请给RK3399刷入Ubuntu映像文件，然后在开发板上执行：

udhcpc  # 获取IP
apt update  # 更新源
apt install pciutils  # 安装lspci工具
apt install devmem2   # 安装devmem2工具

1.1.2 查看设备MSI-X信息

执行lspci -vvv，得到如下信息：

01:00.0 Non-Volatile memory controller: Silicon Motion, Inc. Device 2263 (rev 03) (prog-if 02 [NVM Express])Subsystem: Silicon Motion, Inc. Device 2263Control: I/O- Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx+Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-Latency: 0Interrupt: pin A routed to IRQ 231Region 0: Memory at fa000000 (64-bit, non-prefetchable) [size=16K]Capabilities: [40] Power Management version 3Flags: PMEClk- DSI- D1- D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-)Status: D0 NoSoftRst+ PME-Enable- DSel=0 DScale=0 PME-Capabilities: [50] MSI: Enable- Count=1/8 Maskable+ 64bit+Address: 0000000000000000  Data: 0000Masking: 00000000  Pending: 00000000Capabilities: [70] Express (v2) Endpoint, MSI 00DevCap: MaxPayload 128 bytes, PhantFunc 0, Latency L0s unlimited, L1 unlimitedExtTag- AttnBtn- AttnInd- PwrInd- RBE+ FLReset+ SlotPowerLimit 0.000WDevCtl: Report errors: Correctable- Non-Fatal- Fatal- Unsupported-RlxdOrd+ ExtTag- PhantFunc- AuxPwr- NoSnoop- FLReset-MaxPayload 128 bytes, MaxReadReq 512 bytesDevSta: CorrErr- UncorrErr- FatalErr- UnsuppReq- AuxPwr+ TransPend-LnkCap: Port #0, Speed 8GT/s, Width x4, ASPM L1, Exit Latency L0s <1us, L1 <8usClockPM+ Surprise- LLActRep- BwNot- ASPMOptComp+LnkCtl: ASPM Disabled; RCB 64 bytes Disabled- CommClk-ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-LnkSta: Speed 2.5GT/s, Width x4, TrErr- Train- SlotClk+ DLActive- BWMgmt- ABWMgmt-DevCap2: Completion Timeout: Range ABCD, TimeoutDis+, LTR+, OBFF Not SupportedDevCtl2: Completion Timeout: 50us to 50ms, TimeoutDis-, LTR-, OBFF DisabledLnkCtl2: Target Link Speed: 8GT/s, EnterCompliance- SpeedDis-Transmit Margin: Normal Operating Range, EnterModifiedCompliance- ComplianceSOS-Compliance De-emphasis: -6dBLnkSta2: Current De-emphasis Level: -3.5dB, EqualizationComplete-, EqualizationPhase1-EqualizationPhase2-, EqualizationPhase3-, LinkEqualizationRequest-Capabilities: [b0] MSI-X: Enable+ Count=16 Masked-Vector table: BAR=0 offset=00002000PBA: BAR=0 offset=00002100Capabilities: [100 v2] Advanced Error ReportingUESta:  DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt- UnxCmplt- RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-UEMsk:  DLP- SDES- TLP- FCP- CmpltTO- CmpltAbrt- UnxCmplt- RxOF- MalfTLP- ECRC- UnsupReq- ACSViol-UESvrt: DLP+ SDES+ TLP- FCP+ CmpltTO- CmpltAbrt- UnxCmplt- RxOF+ MalfTLP+ ECRC- UnsupReq- ACSViol-CESta:  RxErr- BadTLP- BadDLLP- Rollover- Timeout- NonFatalErr-CEMsk:  RxErr- BadTLP- BadDLLP- Rollover- Timeout- NonFatalErr+AERCap: First Error Pointer: 00, GenCap+ CGenEn- ChkCap+ ChkEn-Capabilities: [158 v1] #19Capabilities: [178 v1] Latency Tolerance ReportingMax snoop latency: 0nsMax no snoop latency: 0nsCapabilities: [180 v1] L1 PM SubstatesL1SubCap: PCI-PM_L1.2+ PCI-PM_L1.1+ ASPM_L1.2+ ASPM_L1.1+ L1_PM_Substates+PortCommonModeRestoreTime=10us PortTPowerOnTime=10usL1SubCtl1: PCI-PM_L1.2- PCI-PM_L1.1- ASPM_L1.2- ASPM_L1.1-T_CommonMode=0us LTR1.2_Threshold=0nsL1SubCtl2: T_PwrOn=10usKernel driver in use: nvme

从上述信息可以看到：

        Region 0: Memory at fa000000 (64-bit, non-prefetchable) [size=16K]Capabilities: [b0] MSI-X: Enable+ Count=16 Masked-Vector table: BAR=0 offset=00002000PBA: BAR=0 offset=00002100

这表示：

• MSI-X: Enable+：使用MSI-X功能
• Vector table: BAR=0 offset=00002000：MSI的向量在BAR 0偏移地址0x00002000处
• Region 0: Memory at fa000000：
  • BAR 0的PCI地址是0xfa000000，
  • 驱动程序里为了方便令CPU地址等于PCI地址，所以BAR的CPU地址也是0xfa000000。

我们可以去读取 0xfa000000 + 0x00002000开始的向量表，验证里：

• msg addr为0xfee30040
• msg data为0、1、……
  

1.1.3 验证MSI-X信息

二、 怎么使用MSI/MSI-X

参考内核文档：Documentation\PCI\MSI-HOWTO.txt、drivers\nvme\host\pci.c

主要函数是这2个:

int pci_enable_msix_range(struct pci_dev *dev, struct msix_entry *entries,int minvec, int maxvec);
int pci_enable_msi_range(struct pci_dev *dev, int minvec, int maxvec);

示例代码如下：

	// 分配 msix_entry 数组，每一数组项用来保存一个中断的信息dev->entry = kzalloc_node(num_possible_cpus() * sizeof(*dev->entry),GFP_KERNEL, node);// 先尝试使用MSI-Xvecs = pci_enable_msix_range(pdev, dev->entry, 1, nr_io_queues);if (vecs < 0) {// 再尝试使用MSIvecs = pci_enable_msi_range(pdev, 1, min(nr_io_queues, 32));if (vecs < 0) {vecs = 1;} else {for (i = 0; i < vecs; i++)dev->entry[i].vector = i + pdev->irq;}}// request_irq: 中断号都保存在dev->entry[i].vector里for (i = 0; i < vecs; i++)request_irq(dev->entry[i].vector, ...);

注意，在pci_enable_msix_range或者pci_enable_msi_range函数中：

• minvec从1开始
• 对于pci_enable_msix_range，中断号保存在entries[i].vector里
• 对于pci_enable_msi_range，第1个中断号保存在pdev->irq里

三、 MSI/MSI-X中断源码分析

3.1 IRQ Domain创建流程

从PCI设备触发，涉及三个IRQ Domain：

• drivers\irqchip\irq-gic-v3-its-pci-msi.c
• drivers\irqchip\irq-gic-v3-its.c
• drivers\irqchip\irq-gic-v3.c

3.1.1 GIC

设备树：

        gic: interrupt-controller@fee00000 {compatible = "arm,gic-v3";#interrupt-cells = <4>;#address-cells = <2>;#size-cells = <2>;ranges;interrupt-controller;reg = <0x0 0xfee00000 0 0x10000>, /* GICD */<0x0 0xfef00000 0 0xc0000>, /* GICR */<0x0 0xfff00000 0 0x10000>, /* GICC */<0x0 0xfff10000 0 0x10000>, /* GICH */<0x0 0xfff20000 0 0x10000>; /* GICV */interrupts = <GIC_PPI 9 IRQ_TYPE_LEVEL_HIGH 0>;its: interrupt-controller@fee20000 {compatible = "arm,gic-v3-its";msi-controller;reg = <0x0 0xfee20000 0x0 0x20000>;};

驱动代码：`drivers\irqchip\irq-gic-v3.c

3.1.2 ITS

设备树：

its: interrupt-controller@fee20000 {compatible = "arm,gic-v3-its";msi-controller;reg = <0x0 0xfee20000 0x0 0x20000>;};

驱动代码：`drivers\irqchip\irq-gic-v3-its.c

3.1.3 PCI MSI

对应的设备节点跟ITS驱动使用的一样的：

its: interrupt-controller@fee20000 {compatible = "arm,gic-v3-its";msi-controller;reg = <0x0 0xfee20000 0x0 0x20000>;};

驱动代码：drivers\irqchip\irq-gic-v3-its-pci-msi.c，它只是在ITS下面再增加了一个处理层：

3.1.4 PCIe控制器

设备树：

        pcie0: pcie@f8000000 {compatible = "rockchip,rk3399-pcie";#address-cells = <3>;#size-cells = <2>;aspm-no-l0s;clocks = <&cru ACLK_PCIE>, <&cru ACLK_PERF_PCIE>,<&cru PCLK_PCIE>, <&cru SCLK_PCIE_PM>;clock-names = "aclk", "aclk-perf","hclk", "pm";bus-range = <0x0 0x1f>;max-link-speed = <1>;linux,pci-domain = <0>;msi-map = <0x0 &its 0x0 0x1000>;

里面的msi-map = <0x0 &its 0x0 0x1000>;是用来把PCIe设备映射到MSI控制器，它的格式为：

msi-map = <rid-base &msi-controller msi-base length>;

• rid-base：第1个Request ID，就是使用<bus, dev, function>组成的一个数字
• msi-controller：这个PCIe设备映射到哪个MSI控制器？
• msi-base：第1个PCIe设备映射到MSI控制器哪个中断？
• length：能映射多少个设备

3.2 分配中断

代码：drivers\nvme\host\pci.c

nvme_probe > nvme_probe_work > nvme_setup_io_queues pci_enable_msix_rangepci_enable_msix(dev, entries, nvec);msix_capability_init(dev, entries, nvec);pci_enable_msi_rangemsi_capability_init(dev, nvec);msix_capability_init/msi_capability_initpci_msi_setup_msi_irqspci_msi_domain_alloc_irqsmsi_domain_alloc_irqsret = ops->msi_prepare(domain, dev, nvec, &arg); // its_pci_msi_prepareits_pci_msi_prepare // irq-gic-v3-its-pci-msi.c// rid = (bus << 8) | (dev << 4) | functioninfo->scratchpad[0].ul = pci_msi_domain_get_msi_rid(domain, pdev);	return msi_info->ops->msi_prepare(...) // 上一层irq-gic-v3-its.cits_msi_preparedev_id = info->scratchpad[0].ul;  // ridits_dev = its_create_device(its, dev_id, nvec);// 从ITS全局的位图里找到空闲位 chunk// 一个chunk表示32个中断// its的hwirq = (chunk << 5) + 8192// 这也是GIC的hwirqlpi_map = its_lpi_alloc_chunks(nvecs, &lpi_base, &nr_lpis);// 等于(chunk << 5) + 8192 dev->event_map.lpi_base = lpi_base;__irq_domain_alloc_irqsirq_domain_alloc_irqs_recursiveret = domain->ops->alloc(domain, irq_base, nr_irqs, arg);its_irq_domain_allocerr = its_alloc_device_irq(its_dev, &hwirq);*hwirq = dev->event_map.lpi_base + idx;irq_domain_set_hwirq_and_chipirq_data->hwirq = hwirq;irq_data->chip = chip ? chip : &no_irq_chip;irq_domain_activate_irq(irq_data);domain->ops->activate(domain, irq_data);msi_domain_activateirq_chip_compose_msi_msg(irq_data, &msg)	// 构造msg，里面含有MSI或msi-x的addr/valits_irq_compose_msi_msgaddr = its->phys_base + GITS_TRANSLATER;msg->address_lo		= addr & ((1UL << 32) - 1);msg->address_hi		= addr >> 32;// its_get_event_id:// d->hwirq - its_dev->event_map.lpi_base;msg->data		= its_get_event_id(d);                   // 设置msi-x的entry地址	irq_chip_write_msi_msg(irq_data, &msg);data->chip->irq_write_msi_msg(data, msg);pci_msi_domain_write_msg__pci_write_msi_msg(desc, msg);__pci_write_msi_msg(desc, msg);// 对于MSI-Xwritel(msg->address_lo, base + PCI_MSIX_ENTRY_LOWER_ADDR);writel(msg->address_hi, base + PCI_MSIX_ENTRY_UPPER_ADDR);writel(msg->data, base + PCI_MSIX_ENTRY_DATA);// 对于MSIpci_write_config_word(dev, pos + PCI_MSI_FLAGS, msgctl);pci_write_config_dword(dev, pos + PCI_MSI_ADDRESS_LO,msg->address_lo);// 为PCI设备确定hwirq
its_domain_ops.alloc
its_irq_domain_allocits_alloc_device_irq*hwirq = dev->event_map.lpi_base + idx;

---

---

致谢

以上笔记源自韦东山老师的视频课程，感谢韦老师,韦老师是嵌入式培训界一股清流，为嵌入式linux开发点起的星星之火，也愿韦老师桃李满园。聚是一团火，散是满天星！

在这样一个速食的时代，坚持做自己，慢下来，潜心琢磨，心怀敬畏，领悟知识，才能向下扎到根，向上捅破天，背着世界往前行！
仅此向嵌入行业里的每一个认真做技术的从业者致敬！

---

---