Coalesced Interrupt Delivery with posted MSI
Subsystems (6)
v2 cover Coalesced Interrupt Delivery with posted MSI
Hi Thomas and all,
This patch set is aimed to improve IRQ throughput on Intel Xeon by making use of posted interrupts.
There is a session at LPC2023 IOMMU/VFIO/PCI MC where I have presented this topic.
https://lpc.events/event/17/sessions/172/#20231115
Background
==========
On modern x86 server SoCs, interrupt remapping (IR) is required and turned on by default to support X2APIC. Two interrupt remapping modes can be supported by IOMMU/VT-d:
- Remappable (host)
- Posted (guest only so far)
With remappable mode, the device MSI to CPU process is a HW flow without system software touch points, it roughly goes as follows:
1. Devices issue interrupt requests with writes to 0xFEEx_xxxx
2. The system agent accepts and remaps/translates the IRQ
3. Upon receiving the translation response, the system agent notifies the
destination CPU with the translated MSI
4. CPU's local APIC accepts interrupts into its IRR/ISR registers
5. Interrupt delivered through IDT (MSI vector)
The above process can be inefficient under high IRQ rates. The notifications in step #3 are often unnecessary when the destination CPU is already overwhelmed with handling bursts of IRQs. On some architectures, such as Intel Xeon, step #3 is also expensive and requires strong ordering w.r.t DMA. As a result, slower IRQ rates can become a limiting factor for DMA I/O performance.
For example, on Intel Xeon Sapphire Rapids SoC, as more NVMe disks are attached to the same socket, FIO (libaio engine) 4K block random read performance per-disk drops quickly.
# of disks 2 4 8 ------------------------------------- IOPS(million) 1.991 1.136 0.834 (NVMe Gen 5 Samsung PM174x)
With posted mode enabled in interrupt remapping, the interrupt flow is divided into two parts: posting (storing pending IRQ vector information in memory) and CPU notification.
The above remappable IRQ flow becomes the following (1 and 2 unchanged):
3. Notifies the destination CPU with a notification vector
- IOMMU suppresses CPU notification
- IOMMU atomic swap/store IRQ status to memory-resident posted interrupt descriptor (PID)
4. CPU's local APIC accepts the notification interrupt into its IRR/ISR
registers
5. Interrupt delivered through IDT (notification vector handler)
System SW allows new notifications by clearing outstanding notification (ON) bit in PID.
(The above flow is not in Linux today since we only use posted mode for VM)
Note that the system software can now suppress CPU notifications at runtime as needed. This allows the system software to coalesce the expensive CPU notifications and in turn, improve IRQ throughput and DMA performance.
Consider the following scenario when MSIs arrive at a CPU in high-frequency bursts:
Time ----------------------------------------------------------------------->
^ ^ ^ ^ ^ ^ ^ ^ ^ MSIs A B C D E F G H I
RI N N' N' N N' N' N' N N
PI N N N N
RI: remappable interrupt; PI: posted interrupt;
N: interrupt notification, N': superfluous interrupt notification
With remappable interrupt (row titled RI), every MSI generates a notification event to the CPU.
With posted interrupts enabled in this patch set (row titled PI), CPU notifications are coalesced during IRQ bursts. N's are eliminated in the flow above. We refer to this mechanism Coalesced Interrupt Delivery (CID).
Post interrupts have existed for a long time, they have been used for virtualization where MSIs from directly assigned devices can be delivered to the guest kernel without VMM intervention. On x86 Intel platforms, posted interrupts can be used on the host as well. Only host physical address of Posted interrupt descriptor (PID) is used.
This patch set enables a new usage of posted interrupts on existing (and new hardware) for host kernel device MSIs. It is referred to as Posted MSIs throughout this patch set.
Performance (with this patch set):
==================================
Test #1. NVMe FIO
FIO libaio (million IOPS/sec/disk) Gen 5 NVMe Samsung PM174x disks on a single socket, Intel Xeon Sapphire Rapids. Random read with 4k block size. NVMe IRQ affinity is managed by the kernel with one vector per CPU.
#disks Before After %Gain
---------------------------------------------
8 0.834 1.943 132%
4 1.136 2.023 78%
Other observations:
- Increased block sizes shows diminishing benefits, e.g. with 4 NVME disks on one x16 PCIe slot, the combined IOPS looks like:
Block Size Baseline PostedMSI
-------------------------------------
4K 6475 8778
8K 5727 5896
16k 2864 2900
32k 1546 1520
128k 397 398- Submission/Completion latency (usec) also improved at 4K block size only FIO report SLAT
---------------------------------------
Block Size Baseline postedMSI
4k 2177 2282
8k 4416 3967
16k 2950 3053
32k 3453 3505
128k 5911 5801
FIO report CLAT
---------------------------------------
Block Size Baseline postedMSI
4k 313 230
8k 352 343
16k 711 702
32k 1320 1343
128k 5146 5137
Test #2. Intel Data Streaming Accelerator
Two dedicated workqueues from two PCI root complex integrated endpoint (RCIEP) devices, pin IRQ affinity of the two interrupts to a single CPU.
Before After %Gain -------------------------------------
DSA memfill (mil IRQs/sec) 5.157 8.987 74%
DMA throughput has similar improvements.
At lower IRQ rate (< 1 million/second), no performance benefits nor regression observed so far.
No harm tests also performed to ensure no performance regression on workloads that do not have high interrupt rate. These tests include:
- kernel compile time
- file copy
- FIO NVME random writes
Implementation choices:
======================
- Transparent to the device drivers
- System-wide option instead of per-device or per-IRQ opt-in, i.e. once enabled all device MSIs are posted. The benefit is that we only need to change IR irq_chip and domain layer. No change to PCI MSI.
Exceptions are: IOAPIC, HPET, and VT-d's own IRQs
- Limit the number of polling/demuxing loops per CPU notification event
- Only change Intel-IR in IRQ domain hierarchy VECTOR->INTEL-IR->PCI-MSI,
- X86 Intel only so far, can be extended to other architectures with posted interrupt support (ARM and AMD), RFC.
- Bare metal only, no posted interrupt capable virtual IOMMU.
Changes and implications (moving from remappable to posted mode)
===============================
1. All MSI vectors are multiplexed into a single notification vector for each CPU MSI vectors are then de-multiplexed by SW, no IDT delivery for MSIs
2. Losing the following features compared to the remappable mode (AFAIK, none of the below matters for device MSIs)
- Control of delivery mode, e.g. NMI for MSIs
- No logical destinations, posted interrupt destination is x2APIC physical APIC ID
- No per vector stack, since all MSI vectors are multiplexed into one
Runtime changes
===============
The IRQ runtime behavior has changed with this patch, here is a pseudo trace comparison for 3 MSIs of different vectors arriving in a burst on the same CPU. A system vector interrupt (e.g. timer) arrives randomly.
BEFORE:
interrupt(MSI)
irq_enter()
handler() /* EOI */
irq_exit()
process_softirq()interrupt(timer)
interrupt(MSI)
irq_enter()
handler() /* EOI */
irq_exit()
process_softirq()interrupt(MSI)
irq_enter()
handler() /* EOI */
irq_exit()
process_softirq()AFTER:
interrupt /* Posted MSI notification vector */
irq_enter()
atomic_xchg(PIR)
handler()
handler()
handler()
pi_clear_on()
apic_eoi()
irq_exit()interrupt(timer)
process_softirq()
With posted MSI (as pointed out by Thomas Gleixner), both high-priority interrupts (system interrupt vectors) and softIRQs are blocked during MSI vector demux loop. Some can be timing sensitive.
Here are the options I have attempted or still working on:
1. Use self-IPI to invoke MSI vector handler but that took away the majority of the performance benefits.
2. Limit the # of demuxing loops, this is implemented in this patch. Note that today, we already allow one low priority MSI to block system interrupts. System vector can preempt MSI vectors without waiting for EOI but we have IRQ disabled in the ISR.
Performance data (on DSA with MEMFILL) also shows that coalescing more than 3 loops yields diminishing benefits. Therefore, the max loops for coalescing is set to 3 in this patch.
MaxLoop IRQ/sec bandwidth Mbps
-------------------------------------------------------------------------
2 6157107 25219 3 6226611 25504 4 6557081 26857 5 6629683 27155 6 6662425 27289
3. limit the time that system interrupts can be blocked (WIP).
In addition, posted MSI uses atomic xchg from both CPU and IOMMU. Compared to remappable mode, there may be additional cache line ownership contention over PID. However, we have not observed performance regression at lower IRQ rates. At high interrupt rate, posted mode always wins.
Testing:
========
The following tests have been performed and continue to be evaluated.
- IRQ affinity change, migration
- CPU offlining
- Multi vector coalescing
- Low IRQ rate, general no-harm test
- VM device assignment via VFIO
- General no harm test, performance regressions have not been observed for low IRQ rate workload.
With the patch, a new entry in /proc/interrupts is added. cat /proc/interrupts | grep PMN
PMN: 13868907 Posted MSI notification event
No change to the device MSI accounting.
A new INTEL-IR-POST irq_chip is visible at IRQ debugfs, e.g.
domain: IR-PCI-MSIX-0000:6f:01.0-12
hwirq: 0x8
chip: IR-PCI-MSIX-0000:6f:01.0
flags: 0x430
IRQCHIP_SKIP_SET_WAKE IRQCHIP_ONESHOT_SAFE
parent:
domain: INTEL-IR-12-13
hwirq: 0x90000
chip: INTEL-IR-POST /* For posted MSIs */
flags: 0x0
parent:
domain: VECTOR
hwirq: 0x65
chip: APICAcknowledgment
==============
- Rajesh Sankaran and Ashok Raj for the original idea
- Thomas Gleixner for reviewing and guiding the upstream direction of PoC patches. Help correct my many misunderstandings of the IRQ subsystem.
- Jie J Yan(Jeff), Sebastien Lemarie, and Dan Liang for performance evaluation with NVMe and network workload
- Bernice Zhang and Scott Morris for functional validation
- Michael Prinke helped me understand how VT-d HW works
- Sanjay Kumar for providing the DSA IRQ test suite
Changelogs:
V2:
- Code change logs are in individual patches.
- Use "Originally-by" and "Suggested-by" tags to clarify credits/responsibilities.
- More performance evaluation done on FIO
4K rand read test. Four Samsung PM174x NVMe drives on a single x16 PCIe gen5 lane. Fixed CPU frequency at 2.7GHz (p1, highest non-turbo).
IOPS* CPU% sys% user% Ints/sec IOPS/CPU LAT**
AIO (before) 6231 55.5 39.7 15.8 5714721 112.2702703 328
AIO (after) 8936 71.5 51.5 20 7397543 124.979021 229
IOURING(before) 6880 43.7 30.3 13.4 6512402 157.4370709 149
IOURING(after) 8688 58.3 41.3 17 7625158 149.0222985 118
IOURING POLLEDQ 13100 100 85.1 14.9 8000 131 156
* x1000 4 drives combined
** 95% usec.
This patchset improves IOPS, IRQ throughput, and reduces latency for non-polled queues.
V1 (since RFC)
- Removed mentioning of wishful features, IRQ preemption, separate and full MSI vector space
- Refined MSI handler de-multiplexing loop based on suggestions from Peter and Thomas. Reduced xchg() usage and code duplication
- Assign the new posted IR irq_chip only to device MSI/x, avoid changing IO-APIC code
- Extract and use common code for preventing lost interrupt during affinity change
- Added more test results to the cover letter
Thanks,
Jacob
Jacob Pan (13):
x86/irq: Move posted interrupt descriptor out of vmx code x86/irq: Unionize PID.PIR for 64bit access w/o casting x86/irq: Remove bitfields in posted interrupt descriptor x86/irq: Add a Kconfig option for posted MSI
x86/irq: Reserve a per CPU IDT vector for posted MSIs x86/irq: Set up per host CPU posted interrupt descriptors x86/irq: Factor out calling ISR from common_interrupt x86/irq: Install posted MSI notification handler x86/irq: Factor out common code for checking pending interrupts x86/irq: Extend checks for pending vectors to posted interrupts iommu/vt-d: Make posted MSI an opt-in cmdline option iommu/vt-d: Add an irq_chip for posted MSIs iommu/vt-d: Enable posted mode for device MSIs
.../admin-guide/kernel-parameters.txt | 1 + arch/x86/Kconfig | 11 ++ arch/x86/include/asm/apic.h | 12 ++ arch/x86/include/asm/hardirq.h | 6 + arch/x86/include/asm/idtentry.h | 3 + arch/x86/include/asm/irq_remapping.h | 11 ++ arch/x86/include/asm/irq_vectors.h | 9 +- arch/x86/include/asm/posted_intr.h | 108 ++++++++++++ arch/x86/kernel/apic/vector.c | 5 +- arch/x86/kernel/cpu/common.c | 3 + arch/x86/kernel/idt.c | 3 + arch/x86/kernel/irq.c | 164 ++++++++++++++++-- arch/x86/kvm/vmx/posted_intr.c | 4 +- arch/x86/kvm/vmx/posted_intr.h | 93 +--------- arch/x86/kvm/vmx/vmx.c | 3 +- arch/x86/kvm/vmx/vmx.h | 2 +- drivers/iommu/intel/irq_remapping.c | 115 +++++++++++- drivers/iommu/irq_remapping.c | 13 +- tools/arch/x86/include/asm/irq_vectors.h | 9 +- 19 files changed, 457 insertions(+), 118 deletions(-) create mode 100644 arch/x86/include/asm/posted_intr.h
Signature · 1 line
2.25.1
PATCH v2 1/13 x86/irq: Move posted interrupt descriptor out of vmx code
To prepare native usage of posted interrupt, move PID declaration out of VMX code such that they can be shared.
--- arch/x86/include/asm/posted_intr.h | 88 ++++++++++++++++++++++++++++ arch/x86/kvm/vmx/posted_intr.h | 93 +----------------------------- arch/x86/kvm/vmx/vmx.c | 1 + arch/x86/kvm/vmx/vmx.h | 2 +- 4 files changed, 91 insertions(+), 93 deletions(-) create mode 100644 arch/x86/include/asm/posted_intr.h
arch/x86/include/asm/posted_intr.h new file +88 -0
@@ -0,0 +1,88 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#ifndef _X86_POSTED_INTR_H +#define _X86_POSTED_INTR_H + +#define POSTED_INTR_ON 0 +#define POSTED_INTR_SN 1 + +#define PID_TABLE_ENTRY_VALID 1 + +/* Posted-Interrupt Descriptor */ +struct pi_desc { +u32 pir[8 ];/* Posted interrupt requested */ +union { +struct { +/* bit 256 - Outstanding Notification */ +u16 on :1 , +/* bit 257 - Suppress Notification */ + sn :1 , +/* bit 271:258 - Reserved */ + rsvd_1 :14 ; +/* bit 279:272 - Notification Vector */ +u8 nv; +/* bit 287:280 - Reserved */ +u8 rsvd_2; +/* bit 319:288 - Notification Destination */ +u32 ndst; + }; +u64 control; + }; +u32 rsvd[6 ]; +}__aligned(64) ; + +static inline bool pi_test_and_set_on (struct pi_desc *pi_desc) +{ +return test_and_set_bit (POSTED_INTR_ON , (unsigned long *)&pi_desc->control); +} + +static inline bool pi_test_and_clear_on (struct pi_desc *pi_desc) +{ +return test_and_clear_bit (POSTED_INTR_ON , (unsigned long *)&pi_desc->control); +} + +static inline bool pi_test_and_clear_sn (struct pi_desc *pi_desc) +{ +return test_and_clear_bit (POSTED_INTR_SN , (unsigned long *)&pi_desc->control); +} + +static inline bool pi_test_and_set_pir (int vector,struct pi_desc *pi_desc) +{ +return test_and_set_bit (vector, (unsigned long *)pi_desc->pir); +} + +static inline bool pi_is_pir_empty (struct pi_desc *pi_desc) +{ +return bitmap_empty ((unsigned long *)pi_desc->pir,NR_VECTORS ); +} + +static inline void pi_set_sn (struct pi_desc *pi_desc) +{ +set_bit (POSTED_INTR_SN , (unsigned long *)&pi_desc->control); +} + +static inline void pi_set_on (struct pi_desc *pi_desc) +{ +set_bit (POSTED_INTR_ON , (unsigned long *)&pi_desc->control); +} + +static inline void pi_clear_on (struct pi_desc *pi_desc) +{ +clear_bit (POSTED_INTR_ON , (unsigned long *)&pi_desc->control); +} + +static inline void pi_clear_sn (struct pi_desc *pi_desc) +{ +clear_bit (POSTED_INTR_SN , (unsigned long *)&pi_desc->control); +} + +static inline bool pi_test_on (struct pi_desc *pi_desc) +{ +return test_bit (POSTED_INTR_ON , (unsigned long *)&pi_desc->control); +} + +static inline bool pi_test_sn (struct pi_desc *pi_desc) +{ +return test_bit (POSTED_INTR_SN , (unsigned long *)&pi_desc->control); +} + +#endif /* _X86_POSTED_INTR_H */
arch/x86/kvm/vmx/posted_intr.h +1 -92
99 lines not on this page · raw message
arch/x86/kvm/vmx/vmx.c +1 -0
@@ -70,6 +70,7 @@#include "x86.h" #include "smm.h" #include "vmx_onhyperv.h" +#include "posted_intr.h" MODULE_AUTHOR ("Qumranet" );MODULE_LICENSE ("GPL" );
arch/x86/kvm/vmx/vmx.h +1 -1
@@ -7,10 +7,10 @@#include <asm/kvm.h> #include <asm/intel_pt.h> #include <asm/perf_event.h> +#include <asm/posted_intr.h> #include "capabilities.h" #include "../kvm_cache_regs.h" -#include "posted_intr.h" #include "vmcs.h" #include "vmx_ops.h" #include "../cpuid.h"
Signature · 1 line
2.25.1
Re: [PATCH v2 01/13] x86/irq: Move posted interrupt descriptor out of vmx code
"KVM" in the scope would be nice.
Quoted 11 lines · Jacob Pan
To prepare native usage of posted interrupt, move PID declaration out of VMX code such that they can be shared.
Signed-off-by Jacob Pan <[email protected]>--- arch/x86/include/asm/posted_intr.h | 88 ++++++++++++++++++++++++++++ arch/x86/kvm/vmx/posted_intr.h | 93 +----------------------------- arch/x86/kvm/vmx/vmx.c | 1 + arch/x86/kvm/vmx/vmx.h | 2 +- 4 files changed, 91 insertions(+), 93 deletions(-) create mode 100644 arch/x86/include/asm/posted_intr.h
Re: [PATCH v2 01/13] x86/irq: Move posted interrupt descriptor out of vmx code
Hi Sean,
Quoted 1 line · Sean Christopherson
"KVM" in the scope would be nice.
will change to "KVM: VMX:"
Quoted 14 lines · Jacob Pan
Quoted 11 lines
To prepare native usage of posted interrupt, move PID declaration out of VMX code such that they can be shared.
Signed-off-by Jacob Pan <[email protected]>--- arch/x86/include/asm/posted_intr.h | 88 ++++++++++++++++++++++++++++ arch/x86/kvm/vmx/posted_intr.h | 93 +----------------------------- arch/x86/kvm/vmx/vmx.c | 1 + arch/x86/kvm/vmx/vmx.h | 2 +- 4 files changed, 91 insertions(+), 93 deletions(-) create mode 100644 arch/x86/include/asm/posted_intr.h
Quoted 1 line
Thanks,
Jacob
PATCH v2 2/13 x86/irq: Unionize PID.PIR for 64bit access w/o casting
In reply to Jacob PanMake PIR field into u64 such that atomic xchg64 can be used without ugly casting.
--- arch/x86/include/asm/posted_intr.h | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
arch/x86/include/asm/posted_intr.h +4 -1
@@ -9,7 +9,10 @@/* Posted-Interrupt Descriptor */ struct pi_desc { - u32 pir[8]; /* Posted interrupt requested */ + union { + u32 pir[8]; /* Posted interrupt requested */ + u64 pir64[4]; + }; union { struct { /* bit 256 - Outstanding Notification */
Signature · 1 line
2.25.1
PATCH v2 3/13 x86/irq: Remove bitfields in posted interrupt descriptor
In reply to Jacob PanMixture of bitfields and types is weird and really not intuitive, remove bitfields and use typed data exclusively.
---
v2:
- Replace bitfields, no more mix.
--- arch/x86/include/asm/posted_intr.h | 10 +--------- arch/x86/kvm/vmx/posted_intr.c | 4 ++-- arch/x86/kvm/vmx/vmx.c | 2 +- 3 files changed, 4 insertions(+), 12 deletions(-)
arch/x86/include/asm/posted_intr.h +1 -9
@@ -15,17 +15,9 @@ struct pi_desc { };union {struct { -/* bit 256 - Outstanding Notification */ -u16 on :1 , -/* bit 257 - Suppress Notification */ - sn :1 , -/* bit 271:258 - Reserved */ - rsvd_1 :14 ; -/* bit 279:272 - Notification Vector */ +u16 notif_ctrl;/* Suppress and outstanding bits */ u8 nv; -/* bit 287:280 - Reserved */ u8 rsvd_2; -/* bit 319:288 - Notification Destination */ u32 ndst; };u64 control;
arch/x86/kvm/vmx/posted_intr.c +2 -2
@@ -107,7 +107,7 @@ void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int cpu)* handle task migration (@cpu != vcpu->cpu). */ new.ndst = dest; - new.sn =0 ; + new.notif_ctrl &= ~POSTED_INTR_SN ;/* * Restore the notification vector; in the blocking case, the
@@ -157,7 +157,7 @@ static void pi_enable_wakeup_handler(struct kvm_vcpu *vcpu) &per_cpu(wakeup_vcpus_on_cpu, vcpu->cpu));raw_spin_unlock (&per_cpu (wakeup_vcpus_on_cpu_lock, vcpu->cpu)); -WARN (pi_desc->sn,"PI descriptor SN field set before blocking" ); +WARN (pi_desc->notif_ctrl &POSTED_INTR_SN ,"PI descriptor SN field set before blocking" ); old.control =READ_ONCE (pi_desc->control); do {
arch/x86/kvm/vmx/vmx.c +1 -1
@@ -4843,7 +4843,7 @@ static void __vmx_vcpu_reset(struct kvm_vcpu *vcpu)* or POSTED_INTR_WAKEUP_VECTOR. */ vmx->pi_desc.nv =POSTED_INTR_VECTOR ; - vmx->pi_desc.sn =1 ; + vmx->pi_desc.notif_ctrl |=POSTED_INTR_SN ; }static void vmx_vcpu_reset (struct kvm_vcpu *vcpu,bool init_event)
Signature · 1 line
2.25.1
Sean Christopherson
maintainerRe: [PATCH v2 03/13] x86/irq: Remove bitfields in posted interrupt descriptor
"KVM" here would be nice too.
Quoted 45 lines · Jacob Pan
Mixture of bitfields and types is weird and really not intuitive, remove bitfields and use typed data exclusively.
Link https://lore.kernel.org/all/20240404101735.402feec8@jacob-builder/T/#mf66e34a82a48f4d8e2926b5581eff59a122de53a open hereSuggested-by Sean Christopherson <[email protected]>Suggested-by Thomas Gleixner <[email protected]>Signed-off-by Jacob Pan <[email protected]>---
v2:
- Replace bitfields, no more mix.--- arch/x86/include/asm/posted_intr.h | 10 +--------- arch/x86/kvm/vmx/posted_intr.c | 4 ++-- arch/x86/kvm/vmx/vmx.c | 2 +- 3 files changed, 4 insertions(+), 12 deletions(-)arch/x86/include/asm/posted_intr.h +1
-9@@ -15,17 +15,9 @@ struct pi_desc { };union {struct { -/* bit 256 - Outstanding Notification */ -u16 on :1 , -/* bit 257 - Suppress Notification */ - sn :1 , -/* bit 271:258 - Reserved */ - rsvd_1 :14 ; -/* bit 279:272 - Notification Vector */ +u16 notif_ctrl;/* Suppress and outstanding bits */ u8 nv; -/* bit 287:280 - Reserved */ u8 rsvd_2; -/* bit 319:288 - Notification Destination */ u32 ndst; };u64 control;arch/x86/kvm/vmx/posted_intr.c +0
-0
arch/x86/kvm/vmx/posted_intr.c void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int cpu) · line 110
Quoted 5 lines
arch/x86/kvm/vmx/posted_intr.c +1
-1@@ -107,7 +107,7 @@ void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int cpu)* handle task migration (@cpu != vcpu->cpu). */ new.ndst = dest; - new.sn =0 ; + new.notif_ctrl &= ~POSTED_INTR_SN ;
At the risk of creating confusing, would it make sense to add double-underscore, non-atomic versions of the set/clear helpers for ON and SN?
I can't tell if that's a net positive versus open coding clear() and set() here and below.
arch/x86/kvm/vmx/posted_intr.c static void pi_enable_wakeup_handler(struct kvm_vcpu *vcpu) · line 160
Quoted 8 lines
arch/x86/kvm/vmx/posted_intr.c +1
-1/* * Restore the notification vector; in the blocking case, the @@ -157,7 +157,7 @@ static void pi_enable_wakeup_handler(struct kvm_vcpu *vcpu) &per_cpu(wakeup_vcpus_on_cpu, vcpu->cpu));raw_spin_unlock (&per_cpu (wakeup_vcpus_on_cpu_lock, vcpu->cpu)); -WARN (pi_desc->sn,"PI descriptor SN field set before blocking" ); +WARN (pi_desc->notif_ctrl &POSTED_INTR_SN ,"PI descriptor SN field set before blocking" );
This can use pi_test_sn(), as test_bit() isn't atomic, i.e. doesn't incur a LOCK.
Quoted 13 lines
arch/x86/kvm/vmx/posted_intr.c +0
-0old.control =READ_ONCE (pi_desc->control); do {arch/x86/kvm/vmx/vmx.c +1
-1@@ -4843,7 +4843,7 @@ static void __vmx_vcpu_reset(struct kvm_vcpu *vcpu)* or POSTED_INTR_WAKEUP_VECTOR. */ vmx->pi_desc.nv =POSTED_INTR_VECTOR ; - vmx->pi_desc.sn =1 ; + vmx->pi_desc.notif_ctrl |=POSTED_INTR_SN ;
Re: [PATCH v2 03/13] x86/irq: Remove bitfields in posted interrupt descriptor
Hi Sean,
Quoted 61 lines · Sean Christopherson
"KVM" here would be nice too.
Quoted 47 lines · Jacob Pan
Mixture of bitfields and types is weird and really not intuitive, remove bitfields and use typed data exclusively.
Thomas Gleixner <[email protected]> Signed-off-by: Jacob Pan <[email protected]>
---
v2:
- Replace bitfields, no more mix.--- arch/x86/include/asm/posted_intr.h | 10 +--------- arch/x86/kvm/vmx/posted_intr.c | 4 ++-- arch/x86/kvm/vmx/vmx.c | 2 +- 3 files changed, 4 insertions(+), 12 deletions(-)(quoted diff) +0
-0
b/arch/x86/include/asm/posted_intr.hindexacf237b2882e..c682c41d4e44100644 ---a/arch/x86/include/asm/posted_intr.h
+++b/arch/x86/include/asm/posted_intr.harch/x86/include/asm/posted_intr.h +1
-9@@ -15,17 +15,9 @@ struct pi_desc {}; union { struct { -/* bit 256 - Outstanding Notification */ - u16 on : 1, -/* bit 257 - Suppress Notification */ - sn : 1, -/* bit 271:258 - Reserved */ - rsvd_1 : 14; -/* bit 279:272 - Notification Vector */ +u16 notif_ctrl;/* Suppress and outstanding bits */ u8 nv; -/* bit 287:280 - Reserved */ u8 rsvd_2; -/* bit 319:288 - Notification Destination */ u32 ndst; };arch/x86/include/asm/posted_intr.h +0
-0u64 control;(quoted diff) +0
-0
b/arch/x86/kvm/vmx/posted_intr.cindexaf662312fd07..592dbb765675100644arch/x86/kvm/vmx/posted_intr.c void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int · line 111
Quoted 6 lines
arch/x86/kvm/vmx/posted_intr.c +1
-1@@ -107,7 +107,7 @@ void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, intcpu) * handle task migration (@cpu != vcpu->cpu). */ new.ndst = dest; - new.sn =0 ; + new.notif_ctrl &= ~POSTED_INTR_SN ;At the risk of creating confusing, would it make sense to add double-underscore, non-atomic versions of the set/clear helpers for ON and SN?
Quoted 2 lines
I can't tell if that's a net positive versus open coding
clear()andset()here and below.
IMHO, we can add non-atomic helpers when we have more than one user for each operation.
I do have a stupid bug here, it should be:
- new.notif_ctrl &= ~POSTED_INTR_SN;
+ new.notif_ctrl &= ~BIT(POSTED_INTR_SN);
Same as below.
Thanks to Oliver(LKP kvm self test). I didn't catch that in my VFIO device assignment test.
Quoted 14 lines
Quoted 11 lines
arch/x86/kvm/vmx/posted_intr.c +0
-0/* * Restore the notification vector; in the blocking case, the @@ -157,7 +157,7 @@ static void
pi_enable_wakeup_handler(struct kvm_vcpu *vcpu)&per_cpu(wakeup_vcpus_on_cpu,vcpu->cpu));raw_spin_unlock(&per_cpu(wakeup_vcpus_on_cpu_lock, vcpu->cpu));arch/x86/kvm/vmx/posted_intr.c +0
-1- WARN(pi_desc->sn, "PI descriptor SN field set beforeblocking");
arch/x86/kvm/vmx/posted_intr.c +1
-0+ WARN(pi_desc->notif_ctrl & POSTED_INTR_SN, "PI descriptor SNfield set before blocking");
This can use
pi_test_sn(), astest_bit()isn't atomic, i.e. doesn't incur a LOCK.
make sense. will do.
Quoted 14 lines
Quoted 8 lines
arch/x86/kvm/vmx/posted_intr.c +0
-0old.control =READ_ONCE (pi_desc->control); do {arch/x86/kvm/vmx/vmx.c +0
-0arch/x86/kvm/vmx/vmx.c static void __vmx_vcpu_reset(struct kvm_vcpu · line 4847
Quoted 6 lines
arch/x86/kvm/vmx/vmx.c +1
-1@@ -4843,7 +4843,7 @@ static void __vmx_vcpu_reset(struct kvm_vcpu*vcpu) * or POSTED_INTR_WAKEUP_VECTOR. */ vmx->pi_desc.nv =POSTED_INTR_VECTOR ; - vmx->pi_desc.sn =1 ; + vmx->pi_desc.notif_ctrl |=POSTED_INTR_SN ;
Thanks,
Jacob
Re: [PATCH v2 03/13] x86/irq: Remove bitfields in posted interrupt descriptor
Quoted 24 lines · Jacob Pan
Quoted 16 lines · Sean Christopherson
arch/x86/kvm/vmx/posted_intr.c void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int · line 111
Quoted 11 lines
(quoted diff) +0
-0
b/arch/x86/kvm/vmx/posted_intr.cindexaf662312fd07..592dbb765675100644arch/x86/kvm/vmx/posted_intr.c +1
-1@@ -107,7 +107,7 @@ void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, intcpu) * handle task migration (@cpu != vcpu->cpu). */ new.ndst = dest; - new.sn =0 ; + new.notif_ctrl &= ~POSTED_INTR_SN ;At the risk of creating confusing, would it make sense to add double-underscore, non-atomic versions of the set/clear helpers for ON and SN?
Quoted 2 lines
I can't tell if that's a net positive versus open coding
clear()andset()here and below.IMHO, we can add non-atomic helpers when we have more than one user for each operation.
I do have a stupid bug here, it should be:
Review
Quoted 2 lines
(quoted diff) +1
-1- new.notif_ctrl &= ~POSTED_INTR_SN; + new.notif_ctrl &= ~BIT(POSTED_INTR_SN);
That's a perfect reason to use a proper helper.
Re: [PATCH v2 03/13] x86/irq: Remove bitfields in posted interrupt descriptor
Hi Thomas,
Quoted 21 lines · Thomas Gleixner
Quoted 20 lines · Jacob Pan
Quoted 18 lines · Sean Christopherson
arch/x86/kvm/vmx/posted_intr.c void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int · line 111
Quoted 11 lines
(quoted diff) +0
-0
b/arch/x86/kvm/vmx/posted_intr.cindexaf662312fd07..592dbb765675100644 ---a/arch/x86/kvm/vmx/posted_intr.c
+++b/arch/x86/kvm/vmx/posted_intr.carch/x86/kvm/vmx/posted_intr.c +1
-1@@ -107,7 +107,7 @@ void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, intcpu) * handle task migration (@cpu != vcpu->cpu). */ new.ndst = dest; - new.sn =0 ; + new.notif_ctrl &= ~POSTED_INTR_SN ;At the risk of creating confusing, would it make sense to add double-underscore, non-atomic versions of the set/clear helpers for ON and SN?
I can't tell if that's a net positive versus open coding
clear()andset()here and below.
Quoted 8 lines
Review
Quoted 6 lines
IMHO, we can add non-atomic helpers when we have more than one user for each operation.
I do have a stupid bug here, it should be:
(quoted diff) +1
-1- new.notif_ctrl &= ~POSTED_INTR_SN; + new.notif_ctrl &= ~BIT(POSTED_INTR_SN);That's a perfect reason to use a proper helper.
I just proved I was wrong:) will do.
Thanks,
Jacob
PATCH v2 4/13 x86/irq: Add a Kconfig option for posted MSI
In reply to Jacob PanThis option will be used to support delivering MSIs as posted interrupts. Interrupt remapping is required.
---
v2: Remove x2apic dependency
--- arch/x86/Kconfig | 11 +++++++++++ 1 file changed, 11 insertions(+)
arch/x86/Kconfig +11 -0
@@ -464,6 +464,17 @@ config X86_X2APIC If you don't know what to do here, say N. +config X86_POSTED_MSI + bool "Enable MSI and MSI-x delivery by posted interrupts" + depends on X86_64 && IRQ_REMAP + help + This enables MSIs that are under interrupt remapping to be delivered as + posted interrupts to the host kernel. Interrupt throughput can + potentially be improved by coalescing CPU notifications during high + frequency bursts. + + If you don't know what to do here, say N. + config X86_MPPARSE bool "Enable MPS table" if ACPI default y
Signature · 1 line
2.25.1
PATCH v2 5/13 x86/irq: Reserve a per CPU IDT vector for posted MSIs
In reply to Jacob PanWhen posted MSI is enabled, all device MSIs are multiplexed into a single notification vector. MSI handlers will be de-multiplexed at run-time by system software without IDT delivery.
---
v2:
- Add missing CONFIG_ in #ifdef
- Extend changes to x86 tools
--- arch/x86/include/asm/irq_vectors.h | 9 ++++++++- tools/arch/x86/include/asm/irq_vectors.h | 9 ++++++++- 2 files changed, 16 insertions(+), 2 deletions(-)
arch/x86/include/asm/irq_vectors.h +8 -1
@@ -97,9 +97,16 @@#define LOCAL_TIMER_VECTOR 0xec +/* +* Posted interrupt notification vector for all device MSIs delivered to +* the host kernel. +*/ +#define POSTED_MSI_NOTIFICATION_VECTOR 0xeb#define NR_VECTORS 256 -#ifdef CONFIG_X86_LOCAL_APIC +#ifdef CONFIG_X86_POSTED_MSI +#define FIRST_SYSTEM_VECTOR POSTED_MSI_NOTIFICATION_VECTOR +#elif defined(CONFIG_X86_LOCAL_APIC )#define FIRST_SYSTEM_VECTOR LOCAL_TIMER_VECTOR#else #define FIRST_SYSTEM_VECTOR NR_VECTORS
tools/arch/x86/include/asm/irq_vectors.h +8 -1
@@ -99,9 +99,16 @@#define LOCAL_TIMER_VECTOR 0xec +/* +* Posted interrupt notification vector for all device MSIs delivered to +* the host kernel. +*/ +#define POSTED_MSI_NOTIFICATION_VECTOR 0xeb#define NR_VECTORS 256 -#ifdef CONFIG_X86_LOCAL_APIC +#ifdef CONFIG_X86_POSTED_MSI +#define FIRST_SYSTEM_VECTOR POSTED_MSI_NOTIFICATION_VECTOR +#elif defined(CONFIG_X86_LOCAL_APIC )#define FIRST_SYSTEM_VECTOR LOCAL_TIMER_VECTOR#else #define FIRST_SYSTEM_VECTOR NR_VECTORS
Signature · 1 line
2.25.1
Re: [PATCH v2 05/13] x86/irq: Reserve a per CPU IDT vector for posted MSIs
Quoted 12 lines · Jacob Pan
arch/x86/include/asm/irq_vectors.h +4
-0@@ -97,9 +97,16 @@#define LOCAL_TIMER_VECTOR 0xec +/* +* Posted interrupt notification vector for all device MSIs delivered to +* the host kernel. +*/
arch/x86/include/asm/irq_vectors.h line 112
Quoted 10 lines
arch/x86/include/asm/irq_vectors.h +4
-1+#define POSTED_MSI_NOTIFICATION_VECTOR 0xeb#define NR_VECTORS 256 -#ifdef CONFIG_X86_LOCAL_APIC +#ifdef CONFIG_X86_POSTED_MSI +#define FIRST_SYSTEM_VECTOR POSTED_MSI_NOTIFICATION_VECTOR +#elif defined(CONFIG_X86_LOCAL_APIC )#define FIRST_SYSTEM_VECTOR LOCAL_TIMER_VECTOR#else #define FIRST_SYSTEM_VECTOR NR_VECTORS
This is horrible and we had attempts before to make the system vector space dense. They all did not work and making an exception for this is not what we want.
If we really care then we do it proper for _all_ of them. Something like the uncompiled below. There is certainly a smarter way to do the build thing, but my kbuild foo is rusty.
Thanks,
tglx
---
arch/x86/Makefile +1 -0
@@ -245,6 +245,7 @@ archscripts: scripts_basic archheaders: $(Q)$(MAKE) $(build)=arch/x86/entry/syscalls all + $(Q)$(MAKE) $(build)=arch/x86/kernel/irqvectors all ### # Kernel objects
arch/x86/include/asm/irq_vectors.h +31 -44
90 lines not on this page · raw message
arch/x86/kernel/irqvectors/Makefile new file +11 -0
@@ -0,0 +1,11 @@ +# SPDX-License-Identifier: GPL-2.0 + +irqvectors-file := arch/$(SRCARCH)/include/generated/asm/irqvectors.h +targets += arch/$(SRCARCH)/kernel/irqvectors/irqvectors.s + +$(irqvectors-file): arch/$(SRCARCH)/kernel/irqvectors/irqvectors.s FORCE + $(call filechk,offsets,__ASM_IRQVECTORS_H__) + +PHONY += all +all: $(irqvectors-file) + @:
arch/x86/kernel/irqvectors/irqvectors.c new file +42 -0
42 lines not on this page · raw message
Re: [PATCH v2 05/13] x86/irq: Reserve a per CPU IDT vector for posted MSIs
Hi Thomas,
Quoted 29 lines · Thomas Gleixner
Quoted 19 lines · Jacob Pan
(quoted diff) +0
-0
b/arch/x86/include/asm/irq_vectors.hindexd18bfb238f66..1ee00be8218d100644 ---a/arch/x86/include/asm/irq_vectors.h
+++b/arch/x86/include/asm/irq_vectors.h(quoted diff) +2
-0@@ -97,9 +97,16 @@ #define LOCAL_TIMER_VECTOR 0xec +/* + * Posted interrupt notification vector for all device MSIs deliveredto
+ * the host kernel.
+ */(quoted diff) +3
-1+#define POSTED_MSI_NOTIFICATION_VECTOR 0xeb #define NR_VECTORS 256 -#ifdef CONFIG_X86_LOCAL_APIC +#ifdef CONFIG_X86_POSTED_MSI +#define FIRST_SYSTEM_VECTORQuoted 4 lines
POSTED_MSI_NOTIFICATION_VECTOR+#elifdefined(CONFIG_X86_LOCAL_APIC)
#defineFIRST_SYSTEM_VECTORLOCAL_TIMER_VECTOR
#else
#defineFIRST_SYSTEM_VECTORNR_VECTORSThis is horrible and we had attempts before to make the system vector space dense. They all did not work and making an exception for this is not what we want.
Quoted 3 lines
If we really care then we do it proper for _all_ of them. Something like the uncompiled below. There is certainly a smarter way to do the build thing, but my kbuild foo is rusty.
I too had the concern of the wasting system vectors, but did not know how to fix it. But now your code below works well. Tested without KVM in .config to show the gaps:
In VECTOR IRQ domain.
BEFORE:
System: 46: 0-31,50,235-236,244,246-255
AFTER:
System: 46: 0-31,50,241-242,245-255
The only gap is MANAGED_IRQ_SHUTDOWN_VECTOR(243), which is expected on a running system.
Verified in irqvectors.s: .ascii "->MANAGED_IRQ_SHUTDOWN_VECTOR $243
POSTED MSI/first system vector moved up from 235 to 241 for this case.
Will try to let tools/arch/x86/include/asm/irq_vectors.h also use it instead of manually copy over each time. Any suggestions greatly appreciated.
Quoted 156 lines
---
arch/x86/Makefile +1
-0@@ -245,6 +245,7 @@ archscripts: scripts_basic archheaders: $(Q)$(MAKE) $(build)=arch/x86/entry/syscalls all + $(Q)$(MAKE) $(build)=arch/x86/kernel/irqvectors all ### # Kernel objectsarch/x86/include/asm/irq_vectors.h +31
-44@@ -43,59 +43,46 @@*/ #define ISA_IRQ_VECTOR (irq) (((FIRST_EXTERNAL_VECTOR + 16) & ~15) + irq) +#ifndef __ASSEMBLY__/* -* Special IRQ vectors used by the SMP architecture, 0xf0-0xff -* -* some of the following vectors are 'rare', they are merged -* into a single vector (CALL_FUNCTION_VECTOR) to save vector space. -* TLB, reschedule and local APIC vectors are performance-critical. +* Special IRQ vectors used by the SMP architecture, 0xff and downwards */ +enum { + __SPURIOUS_APIC_VECTOR, + __ERROR_APIC_VECTOR, + __RESCHEDULE_VECTOR, + __CALL_FUNCTION_VECTOR, + __CALL_FUNCTION_SINGLE_VECTOR, + __THERMAL_APIC_VECTOR, + __THRESHOLD_APIC_VECTOR, + __REBOOT_VECTOR, + __X86_PLATFORM_IPI_VECTOR, + __IRQ_WORK_VECTOR, + __DEFERRED_ERROR_VECTOR, -#define SPURIOUS_APIC_VECTOR 0xff -/* -* Sanity check -*/ -#if ((SPURIOUS_APIC_VECTOR &0x0F ) !=0x0F ) -# error SPURIOUS_APIC_VECTOR definition error +#if IS_ENABLED (CONFIG_HYPERVISOR_GUEST ) + __HYPERVISOR_CALLBACK_VECTOR,#endif -#define ERROR_APIC_VECTOR 0xfe -#define RESCHEDULE_VECTOR 0xfd -#define CALL_FUNCTION_VECTOR 0xfc -#define CALL_FUNCTION_SINGLE_VECTOR 0xfb -#define THERMAL_APIC_VECTOR 0xfa -#define THRESHOLD_APIC_VECTOR 0xf9 -#define REBOOT_VECTOR 0xf8 - -/* -* Generic system vector for platform specific use -*/ -#define X86_PLATFORM_IPI_VECTOR 0xf7 - -/* -* IRQ work vector: -*/ -#define IRQ_WORK_VECTOR 0xf6 - -/* 0xf5 - unused, was UV_BAU_MESSAGE */ -#define DEFERRED_ERROR_VECTOR 0xf4 - -/* Vector on which hypervisor callbacks will be delivered */ -#define HYPERVISOR_CALLBACK_VECTOR 0xf3 - -/* Vector for KVM to deliver posted interrupt IPI */ -#define POSTED_INTR_VECTOR 0xf2 -#define POSTED_INTR_WAKEUP_VECTOR 0xf1 -#define POSTED_INTR_NESTED_VECTOR 0xf0 - -#define MANAGED_IRQ_SHUTDOWN_VECTOR 0xef +#if IS_ENABLED (CONFIG_KVM ) +/* Vector for KVM to deliver posted interrupt IPI */ + __POSTED_INTR_VECTOR, + __POSTED_INTR_WAKEUP_VECTOR, + __POSTED_INTR_NESTED_VECTOR, +#endif + __MANAGED_IRQ_SHUTDOWN_VECTOR,#if IS_ENABLED (CONFIG_HYPERV ) -#define HYPERV_REENLIGHTENMENT_VECTOR 0xee -#define HYPERV_STIMER0_VECTOR 0xed + __HYPERV_REENLIGHTENMENT_VECTOR, + __HYPERV_STIMER0_VECTOR,#endif + __LOCAL_TIMER_VECTOR, +}; +#endif /* !__ASSEMBLY__ */ -#define LOCAL_TIMER_VECTOR 0xec +#ifndef COMPILE_OFFSETS +#include <asm/irqvectors.h> +#endif #define NR_VECTORS 256arch/x86/kernel/irqvectors/Makefile new file +3
-0@@ -0,0 +1,11 @@ +# SPDX-License-Identifier: GPL-2.0 + +irqvectors-file :=arch/$(SRCARCH)/include/generated/asm/irqvectors.h +targets +=
arch/$(SRCARCH)/kernel/irqvectors/irqvectors.s + +$(irqvectors-file): arch/$(SRCARCH)/kernel/irqvectors/irqvectors.s FORCEarch/x86/kernel/irqvectors/Makefile new file +5
-0+ $(call filechk,offsets,__ASM_IRQVECTORS_H__) + +PHONY += all +all: $(irqvectors-file) + @:arch/x86/kernel/irqvectors/irqvectors.c new file +32
-0@@ -0,0 +1,42 @@ +// SPDX-License-Identifier: GPL-2.0 +#define COMPILE_OFFSETS + +#include <linux/kbuild.h> +#include <asm/irq_vectors.h> + +#define VECNR (v) (0xFF - __##v) +#define VECTOR (v) DEFINE(v, VECNR(v)) + +static void __used common (void ) +{ +VECTOR (SPURIOUS_APIC_VECTOR ); +VECTOR (ERROR_APIC_VECTOR ); +VECTOR (RESCHEDULE_VECTOR ); +VECTOR (CALL_FUNCTION_VECTOR ); +VECTOR (CALL_FUNCTION_SINGLE_VECTOR ); +VECTOR (THERMAL_APIC_VECTOR ); +VECTOR (THRESHOLD_APIC_VECTOR ); +VECTOR (REBOOT_VECTOR ); +VECTOR (X86_PLATFORM_IPI_VECTOR ); +VECTOR (IRQ_WORK_VECTOR ); +VECTOR (DEFERRED_ERROR_VECTOR ); + +#if IS_ENABLED (CONFIG_HYPERVISOR_GUEST ) +VECTOR (HYPERVISOR_CALLBACK_VECTOR ); +#endif + +#if IS_ENABLED (CONFIG_KVM ) +/* Vector for KVM to deliver posted interrupt IPI */ +VECTOR (POSTED_INTR_VECTOR ); +VECTOR (POSTED_INTR_WAKEUP_VECTOR ); +VECTOR (POSTED_INTR_NESTED_VECTOR );
arch/x86/kernel/irqvectors/irqvectors.c line 42
Quoted 10 lines
arch/x86/kernel/irqvectors/irqvectors.c new file +10
-0+#endif +VECTOR (MANAGED_IRQ_SHUTDOWN_VECTOR ); + +#if IS_ENABLED (CONFIG_HYPERV ) +VECTOR (HYPERV_REENLIGHTENMENT_VECTOR ); +VECTOR (HYPERV_STIMER0_VECTOR ); +#endif +VECTOR (LOCAL_TIMER_VECTOR ); +} +
Thanks,
Jacob
Re: [PATCH v2 05/13] x86/irq: Reserve a per CPU IDT vector for posted MSIs
Quoted 52 lines · Jacob Pan
Hi Thomas,
Quoted 29 lines · Thomas Gleixner
Quoted 19 lines · Jacob Pan
(quoted diff) +0
-0
b/arch/x86/include/asm/irq_vectors.hindexd18bfb238f66..1ee00be8218d100644 ---a/arch/x86/include/asm/irq_vectors.h
+++b/arch/x86/include/asm/irq_vectors.h(quoted diff) +2
-0@@ -97,9 +97,16 @@ #define LOCAL_TIMER_VECTOR 0xec +/* + * Posted interrupt notification vector for all device MSIs deliveredto
+ * the host kernel.
+ */(quoted diff) +3
-1+#define POSTED_MSI_NOTIFICATION_VECTOR 0xeb #define NR_VECTORS 256 -#ifdef CONFIG_X86_LOCAL_APIC +#ifdef CONFIG_X86_POSTED_MSI +#define FIRST_SYSTEM_VECTORQuoted 4 lines
POSTED_MSI_NOTIFICATION_VECTOR+#elifdefined(CONFIG_X86_LOCAL_APIC)
#defineFIRST_SYSTEM_VECTORLOCAL_TIMER_VECTOR
#else
#defineFIRST_SYSTEM_VECTORNR_VECTORSThis is horrible and we had attempts before to make the system vector space dense. They all did not work and making an exception for this is not what we want.
Quoted 3 lines
If we really care then we do it proper for _all_ of them. Something like the uncompiled below. There is certainly a smarter way to do the build thing, but my kbuild foo is rusty.
I too had the concern of the wasting system vectors, but did not know how to fix it. But now your code below works well. Tested without KVM in .config to show the gaps:
In VECTOR IRQ domain.
BEFORE:
System: 46: 0-31,50,235-236,244,246-255AFTER:
System: 46: 0-31,50,241-242,245-255The only gap is
MANAGED_IRQ_SHUTDOWN_VECTOR(243), which is expected on a running system.
Quoted 8 lines
Verified in irqvectors.s: .ascii "->MANAGED_IRQ_SHUTDOWN_VECTOR $243
POSTED MSI/first system vector moved up from 235 to 241 for this case.
Will try to let
tools/arch/x86/include/asm/irq_vectors.halso use it instead of manually copy over each time. Any suggestions greatly appreciated.
On a second thought, if we make system IRQ vector determined at compile time based on different CONFIG options, will it break userspace tools such as perf? More importantly the rule of not breaking userspace.
+Arnaldo
Quoted 171 lines
Quoted 157 lines
---
arch/x86/Makefile +1
-0@@ -245,6 +245,7 @@ archscripts: scripts_basic archheaders: $(Q)$(MAKE) $(build)=arch/x86/entry/syscalls all + $(Q)$(MAKE) $(build)=arch/x86/kernel/irqvectors all ### # Kernel objectsarch/x86/include/asm/irq_vectors.h +31
-44@@ -43,59 +43,46 @@*/ #define ISA_IRQ_VECTOR (irq) (((FIRST_EXTERNAL_VECTOR + 16) & ~15) + irq) +#ifndef __ASSEMBLY__/* -* Special IRQ vectors used by the SMP architecture, 0xf0-0xff -* -* some of the following vectors are 'rare', they are merged -* into a single vector (CALL_FUNCTION_VECTOR) to save vector space. -* TLB, reschedule and local APIC vectors are performance-critical. +* Special IRQ vectors used by the SMP architecture, 0xff and downwards */ +enum { + __SPURIOUS_APIC_VECTOR, + __ERROR_APIC_VECTOR, + __RESCHEDULE_VECTOR, + __CALL_FUNCTION_VECTOR, + __CALL_FUNCTION_SINGLE_VECTOR, + __THERMAL_APIC_VECTOR, + __THRESHOLD_APIC_VECTOR, + __REBOOT_VECTOR, + __X86_PLATFORM_IPI_VECTOR, + __IRQ_WORK_VECTOR, + __DEFERRED_ERROR_VECTOR, -#define SPURIOUS_APIC_VECTOR 0xff -/* -* Sanity check -*/ -#if ((SPURIOUS_APIC_VECTOR &0x0F ) !=0x0F ) -# error SPURIOUS_APIC_VECTOR definition error +#if IS_ENABLED (CONFIG_HYPERVISOR_GUEST ) + __HYPERVISOR_CALLBACK_VECTOR,#endif -#define ERROR_APIC_VECTOR 0xfe -#define RESCHEDULE_VECTOR 0xfd -#define CALL_FUNCTION_VECTOR 0xfc -#define CALL_FUNCTION_SINGLE_VECTOR 0xfb -#define THERMAL_APIC_VECTOR 0xfa -#define THRESHOLD_APIC_VECTOR 0xf9 -#define REBOOT_VECTOR 0xf8 - -/* -* Generic system vector for platform specific use -*/ -#define X86_PLATFORM_IPI_VECTOR 0xf7 - -/* -* IRQ work vector: -*/ -#define IRQ_WORK_VECTOR 0xf6 - -/* 0xf5 - unused, was UV_BAU_MESSAGE */ -#define DEFERRED_ERROR_VECTOR 0xf4 - -/* Vector on which hypervisor callbacks will be delivered */ -#define HYPERVISOR_CALLBACK_VECTOR 0xf3 - -/* Vector for KVM to deliver posted interrupt IPI */ -#define POSTED_INTR_VECTOR 0xf2 -#define POSTED_INTR_WAKEUP_VECTOR 0xf1 -#define POSTED_INTR_NESTED_VECTOR 0xf0 - -#define MANAGED_IRQ_SHUTDOWN_VECTOR 0xef +#if IS_ENABLED (CONFIG_KVM ) +/* Vector for KVM to deliver posted interrupt IPI */ + __POSTED_INTR_VECTOR, + __POSTED_INTR_WAKEUP_VECTOR, + __POSTED_INTR_NESTED_VECTOR, +#endif + __MANAGED_IRQ_SHUTDOWN_VECTOR,#if IS_ENABLED (CONFIG_HYPERV ) -#define HYPERV_REENLIGHTENMENT_VECTOR 0xee -#define HYPERV_STIMER0_VECTOR 0xed + __HYPERV_REENLIGHTENMENT_VECTOR, + __HYPERV_STIMER0_VECTOR,#endif + __LOCAL_TIMER_VECTOR, +}; +#endif /* !__ASSEMBLY__ */ -#define LOCAL_TIMER_VECTOR 0xec +#ifndef COMPILE_OFFSETS +#include <asm/irqvectors.h> +#endif #define NR_VECTORS 256arch/x86/kernel/irqvectors/Makefile new file +3
-0@@ -0,0 +1,11 @@ +# SPDX-License-Identifier: GPL-2.0 + +irqvectors-file :=arch/$(SRCARCH)/include/generated/asm/irqvectors.h +targets +=
arch/$(SRCARCH)/kernel/irqvectors/irqvectors.s + +$(irqvectors-file): arch/$(SRCARCH)/kernel/irqvectors/irqvectors.s FORCEarch/x86/kernel/irqvectors/Makefile new file +5
-0+ $(call filechk,offsets,__ASM_IRQVECTORS_H__) + +PHONY += all +all: $(irqvectors-file) + @:arch/x86/kernel/irqvectors/irqvectors.c new file +32
-0@@ -0,0 +1,42 @@ +// SPDX-License-Identifier: GPL-2.0 +#define COMPILE_OFFSETS + +#include <linux/kbuild.h> +#include <asm/irq_vectors.h> + +#define VECNR (v) (0xFF - __##v) +#define VECTOR (v) DEFINE(v, VECNR(v)) + +static void __used common (void ) +{ +VECTOR (SPURIOUS_APIC_VECTOR ); +VECTOR (ERROR_APIC_VECTOR ); +VECTOR (RESCHEDULE_VECTOR ); +VECTOR (CALL_FUNCTION_VECTOR ); +VECTOR (CALL_FUNCTION_SINGLE_VECTOR ); +VECTOR (THERMAL_APIC_VECTOR ); +VECTOR (THRESHOLD_APIC_VECTOR ); +VECTOR (REBOOT_VECTOR ); +VECTOR (X86_PLATFORM_IPI_VECTOR ); +VECTOR (IRQ_WORK_VECTOR ); +VECTOR (DEFERRED_ERROR_VECTOR ); + +#if IS_ENABLED (CONFIG_HYPERVISOR_GUEST ) +VECTOR (HYPERVISOR_CALLBACK_VECTOR ); +#endif + +#if IS_ENABLED (CONFIG_KVM ) +/* Vector for KVM to deliver posted interrupt IPI */ +VECTOR (POSTED_INTR_VECTOR ); +VECTOR (POSTED_INTR_WAKEUP_VECTOR ); +VECTOR (POSTED_INTR_NESTED_VECTOR );arch/x86/kernel/irqvectors/irqvectors.c line 42
Quoted 10 lines
arch/x86/kernel/irqvectors/irqvectors.c new file +10
-0+#endif +VECTOR (MANAGED_IRQ_SHUTDOWN_VECTOR ); + +#if IS_ENABLED (CONFIG_HYPERV ) +VECTOR (HYPERV_REENLIGHTENMENT_VECTOR ); +VECTOR (HYPERV_STIMER0_VECTOR ); +#endif +VECTOR (LOCAL_TIMER_VECTOR ); +} +Thanks,
Quoted 1 line
Jacob
Thanks,
Jacob
Re: [PATCH v2 05/13] x86/irq: Reserve a per CPU IDT vector for posted MSIs
Quoted 24 lines · Jacob Pan
Quoted 23 lines · Jacob Pan
Quoted 3 lines · Thomas Gleixner
If we really care then we do it proper for _all_ of them. Something like the uncompiled below. There is certainly a smarter way to do the build thing, but my kbuild foo is rusty.
I too had the concern of the wasting system vectors, but did not know how to fix it. But now your code below works well. Tested without KVM in .config to show the gaps:
In VECTOR IRQ domain.
BEFORE:
System: 46: 0-31,50,235-236,244,246-255AFTER:
System: 46: 0-31,50,241-242,245-255The only gap is
MANAGED_IRQ_SHUTDOWN_VECTOR(243), which is expected on a running system.Verified in irqvectors.s: .ascii "->MANAGED_IRQ_SHUTDOWN_VECTOR $243
POSTED MSI/first system vector moved up from 235 to 241 for this case.
Quoted 7 lines
Quoted 4 lines
Will try to let
tools/arch/x86/include/asm/irq_vectors.halso use it instead of manually copy over each time. Any suggestions greatly appreciated.On a second thought, if we make system IRQ vector determined at compile time based on different CONFIG options, will it break userspace tools such as perf? More importantly the rule of not breaking userspace.
tools/arch/x86/include/asm/irq_vectors.h is only used to generate the list of system vectors for pretty output. And your change already broke that.
The obvious solution to that is to expose that list in sysfs for consumption by perf.
But we don't have to do any of that right away. It's an orthogonal issue. Just waste the extra system vector to start with and then we can add the compile time dependend change on top if we really care about gaining back the vectors.
Thanks,
tglx
Re: [PATCH v2 05/13] x86/irq: Reserve a per CPU IDT vector for posted MSIs
Thomas Gleixner <[email protected]> writes:
Dropping CC of a bazillion people and adding Arnaldo who owns perf trace which has the vector decoding code.
Quoted 48 lines · Jacob Pan
Quoted 24 lines · Jacob Pan
Quoted 23 lines · Thomas Gleixner
Quoted 3 lines
If we really care then we do it proper for _all_ of them. Something like the uncompiled below. There is certainly a smarter way to do the build thing, but my kbuild foo is rusty.
I too had the concern of the wasting system vectors, but did not know how to fix it. But now your code below works well. Tested without KVM in .config to show the gaps:
In VECTOR IRQ domain.
BEFORE:
System: 46: 0-31,50,235-236,244,246-255AFTER:
System: 46: 0-31,50,241-242,245-255The only gap is
MANAGED_IRQ_SHUTDOWN_VECTOR(243), which is expected on a running system.Verified in irqvectors.s: .ascii "->MANAGED_IRQ_SHUTDOWN_VECTOR $243
POSTED MSI/first system vector moved up from 235 to 241 for this case.
Quoted 7 lines
Quoted 4 lines
Will try to let
tools/arch/x86/include/asm/irq_vectors.halso use it instead of manually copy over each time. Any suggestions greatly appreciated.On a second thought, if we make system IRQ vector determined at compile time based on different CONFIG options, will it break userspace tools such as perf? More importantly the rule of not breaking userspace.
tools/arch/x86/include/asm/irq_vectors.his only used to generate the list of system vectors for pretty output. And your change already broke that.The obvious solution to that is to expose that list in sysfs for consumption by perf.
But we don't have to do any of that right away. It's an orthogonal issue. Just waste the extra system vector to start with and then we can add the compile time dependend change on top if we really care about gaining back the vectors.
Thanks,
tglx
Re: [PATCH v2 05/13] x86/irq: Reserve a per CPU IDT vector for posted MSIs
Quoted 4 lines · Andi Kleen
Thomas Gleixner <[email protected]> writes:
Dropping CC of a bazillion people and adding Arnaldo who owns perf trace which has the vector decoding code.
He was already on Cc.
Re: [PATCH v2 05/13] x86/irq: Reserve a per CPU IDT vector for posted MSIs
In reply to Thomas GleixnerQuoted 25 lines · Thomas Gleixner
Quoted 24 lines · Jacob Pan
Quoted 23 lines · Jacob Pan
Quoted 3 lines · Thomas Gleixner
If we really care then we do it proper for _all_ of them. Something like the uncompiled below. There is certainly a smarter way to do the build thing, but my kbuild foo is rusty.
I too had the concern of the wasting system vectors, but did not know how to fix it. But now your code below works well. Tested without KVM in .config to show the gaps:
In VECTOR IRQ domain.
BEFORE:
System: 46: 0-31,50,235-236,244,246-255AFTER:
System: 46: 0-31,50,241-242,245-255The only gap is
MANAGED_IRQ_SHUTDOWN_VECTOR(243), which is expected on a running system.Verified in irqvectors.s: .ascii "->MANAGED_IRQ_SHUTDOWN_VECTOR $243
POSTED MSI/first system vector moved up from 235 to 241 for this case.
Quoted 7 lines
Quoted 7 lines
Quoted 4 lines
Will try to let
tools/arch/x86/include/asm/irq_vectors.halso use it instead of manually copy over each time. Any suggestions greatly appreciated.On a second thought, if we make system IRQ vector determined at compile time based on different CONFIG options, will it break userspace tools such as perf? More importantly the rule of not breaking userspace.
The rule for tools/perf is "don't impose _any requirement_ on the kernel developers, they don't have to test if any change they do outside of tools/ will break something inside tools/."
Quoted 3 lines
tools/arch/x86/include/asm/irq_vectors.his only used to generate the list of system vectors for pretty output. And your change already broke that.
Yeah, I even moved that from tools/arch/x86/include/asm/irq_vectors.h to tools/perf/trace/beauty/arch/x86/include/asm/irq_vectors.h (for next merge window).
Having it in tools/arch/x86/include/asm/irq_vectors.h was a bad decision as it, as you mentinoned, is only used to generate string tables:
⬢[acme@toolbox perf-tools-next]$ tools/perf/trace/beauty/tracepoints/x86_irq_vectors.sh
static const char *x86_irq_vectors[] = {
[0x02] = "NMI",
[0x80] = "IA32_SYSCALL",
[0xec] = "LOCAL_TIMER",
[0xed] = "HYPERV_STIMER0",
[0xee] = "HYPERV_REENLIGHTENMENT",
[0xef] = "MANAGED_IRQ_SHUTDOWN",
[0xf0] = "POSTED_INTR_NESTED",
[0xf1] = "POSTED_INTR_WAKEUP",
[0xf2] = "POSTED_INTR",
[0xf3] = "HYPERVISOR_CALLBACK",
[0xf4] = "DEFERRED_ERROR",
[0xf6] = "IRQ_WORK",
[0xf7] = "X86_PLATFORM_IPI",
[0xf8] = "REBOOT",
[0xf9] = "THRESHOLD_APIC",
[0xfa] = "THERMAL_APIC",
[0xfb] = "CALL_FUNCTION_SINGLE",
[0xfc] = "CALL_FUNCTION",
[0xfd] = "RESCHEDULE",
[0xfe] = "ERROR_APIC",
[0xff] = "SPURIOUS_APIC",
};⬢[acme@toolbox perf-tools-next]$
Used in:
root@number:~# perf trace -a -e irq_vectors:irq_work_entry/max-stack=32/ --max-events=1
0.000 kworker/u57:0-/9912 irq_vectors:irq_work_entry(vector: IRQ_WORK)
__sysvec_irq_work ([kernel.kallsyms])
__sysvec_irq_work ([kernel.kallsyms])
sysvec_irq_work ([kernel.kallsyms])
asm_sysvec_irq_work ([kernel.kallsyms])
_raw_spin_unlock_irqrestore ([kernel.kallsyms])
dma_fence_wait_timeout ([kernel.kallsyms])
intel_atomic_commit_tail ([kernel.kallsyms])
process_one_work ([kernel.kallsyms])
worker_thread ([kernel.kallsyms])
kthread ([kernel.kallsyms])
ret_from_fork ([kernel.kallsyms])
ret_from_fork_asm ([kernel.kallsyms])
root@number:~#But as the original cset introducing this explains, these irq_vectors:
tracepoins operate on just one of the vectors, so irq_work_entry(vector:IRQ_WORK), irq_vectors:reschedule_exit(vector: RESCHEDULE), etc.
Quoted 2 lines
The obvious solution to that is to expose that list in sysfs for consumption by perf.
nah, the best thing these days is stop using 'int' for vector and use 'enum irq_vector', then since we have BTF we can use that to do the enum -> string translation, like with (using /sys/kernel/btf/vmlinux, that is pretty much available everywhere these days):
root@number:~# pahole clocksource_ids
enum clocksource_ids {
CSID_GENERIC = 0,
CSID_ARM_ARCH_COUNTER = 1,
CSID_MAX = 2,
};root@number:~# pahole skb_drop_reason | head
enum skb_drop_reason {
SKB_NOT_DROPPED_YET = 0,
SKB_CONSUMED = 1,
SKB_DROP_REASON_NOT_SPECIFIED = 2,
SKB_DROP_REASON_NO_SOCKET = 3,
SKB_DROP_REASON_PKT_TOO_SMALL = 4,
SKB_DROP_REASON_TCP_CSUM = 5,
SKB_DROP_REASON_SOCKET_FILTER = 6,
SKB_DROP_REASON_UDP_CSUM = 7,
SKB_DROP_REASON_NETFILTER_DROP = 8,
root@number:~#Then its easy to go from 0 to CSID_GENERIC, etc.
⬢[acme@toolbox pahole]$ perf stat -e cycles pahole skb_drop_reason > /dev/null
Performance counter stats for 'pahole skb_drop_reason':
6,095,427 cpu_atom/cycles:u/ (2.82%)
103,694,633 cpu_core/cycles:u/ (97.18%) 0.039031759 seconds time elapsed
0.016028000 seconds user
0.023007000 seconds sys⬢[acme@toolbox pahole]$
- Arnaldo
Quoted 8 lines
But we don't have to do any of that right away. It's an orthogonal issue. Just waste the extra system vector to start with and then we can add the compile time dependend change on top if we really care about gaining back the vectors.
Thanks,
tglx
Re: [PATCH v2 05/13] x86/irq: Reserve a per CPU IDT vector for posted MSIs
Hi Arnaldo,
Quoted 9 lines · Arnaldo Carvalho de Melo
Quoted 4 lines
Quoted 4 lines
On a second thought, if we make system IRQ vector determined at compile time based on different CONFIG options, will it break userspace tools such as perf? More importantly the rule of not breaking userspace.
The rule for
tools/perfis "don't impose_anyrequirement_on the kernel developers, they don't have to test if any change they do outside oftools/will break something insidetools/."
Quoted 7 lines
Quoted 3 lines
tools/arch/x86/include/asm/irq_vectors.his only used to generate the list of system vectors for pretty output. And your change already broke that.Yeah, I even moved that from
tools/arch/x86/include/asm/irq_vectors.htotools/perf/trace/beauty/arch/x86/include/asm/irq_vectors.h(for next merge window).
So I will not add anything to the tools directory for my next version. Just a heads-up for adding this new vector.
Thanks,
Jacob
RE: [PATCH v2 05/13] x86/irq: Reserve a per CPU IDT vector for posted MSIs
In reply to Jacob Panarch/x86/include/asm/irq_vectors.h line 105
Quoted 10 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMarch/x86/include/asm/irq_vectors.h +5
-0+/* +* Posted interrupt notification vector for all device MSIs delivered to +* the host kernel. +*/ +#define POSTED_MSI_NOTIFICATION_VECTOR 0xeb#define NR_VECTORS 256
Every interrupt is kind of a notification.
Just call it POSTED_MSI_VECTOR
Sean Christopherson
maintainerRe: [PATCH v2 05/13] x86/irq: Reserve a per CPU IDT vector for posted MSIs
Quoted 12 lines · Kevin Tian
arch/x86/include/asm/irq_vectors.h line 105
Quoted 10 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMarch/x86/include/asm/irq_vectors.h +5
-0+/* +* Posted interrupt notification vector for all device MSIs delivered to +* the host kernel. +*/ +#define POSTED_MSI_NOTIFICATION_VECTOR 0xeb#define NR_VECTORS 256Every interrupt is kind of a notification.
FWIW, I find value in having "notification" in the name to differentiate between the IRQ that is notifying the CPU that there's a posted IRQ to be processed, and the posted IRQ itself.
RE: [PATCH v2 05/13] x86/irq: Reserve a per CPU IDT vector for posted MSIs
Quoted 17 lines
From: Sean Christopherson <[email protected]>
Sent: Friday, April 12, 2024 10:28 PMQuoted 12 lines · Kevin Tian
arch/x86/include/asm/irq_vectors.h line 105
Quoted 10 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMarch/x86/include/asm/irq_vectors.h +5
-0+/* +* Posted interrupt notification vector for all device MSIs delivered to +* the host kernel. +*/ +#define POSTED_MSI_NOTIFICATION_VECTOR 0xeb#define NR_VECTORS 256Every interrupt is kind of a notification.
Quoted 5 lines
FWIW, I find value in having "notification" in the name to differentiate between
the IRQ that is notifying the CPU that there's a posted IRQ to be processed, and
the posted IRQ itself.
IMHO one who knows posted msi doesn't need the extra 'notification' in the name to differentiate.
one who doesn't know what posted msi is anyway needs to look at the surrounding code including the above comment. having 'notification' in the name alone doesn't really help.
but I'd not hold strong on this... 😊
PATCH v2 6/13 x86/irq: Set up per host CPU posted interrupt descriptors
In reply to Jacob PanTo support posted MSIs, create a posted interrupt descriptor (PID) for each host CPU. Later on, when setting up IRQ CPU affinity, IOMMU's interrupt remapping table entry (IRTE) will point to the physical address of the matching CPU's PID.
Each PID is initialized with the owner CPU's physical APICID as the destination.
---
v2: Fix xAPIC destination ID assignment, Oliver Sang reported failture on system with x2apic optout BIOS option.
--- arch/x86/include/asm/hardirq.h | 3 +++ arch/x86/include/asm/posted_intr.h | 7 +++++++ arch/x86/kernel/cpu/common.c | 3 +++ arch/x86/kernel/irq.c | 23 +++++++++++++++++++++++ 4 files changed, 36 insertions(+)
arch/x86/include/asm/hardirq.h +3 -0
@@ -48,6 +48,9 @@ typedef struct {DECLARE_PER_CPU_SHARED_ALIGNED (irq_cpustat_t, irq_stat); +#ifdef CONFIG_X86_POSTED_MSI +DECLARE_PER_CPU_ALIGNED(struct pi_desc, posted_interrupt_desc); +#endif #define __ARCH_IRQ_STAT#define inc_irq_stat (member) this_cpu_inc(irq_stat.member)
arch/x86/include/asm/posted_intr.h +7 -0
@@ -80,4 +80,11 @@ static inline bool pi_test_sn(struct pi_desc *pi_desc) return test_bit(POSTED_INTR_SN, (unsigned long *)&pi_desc->control); } +#ifdef CONFIG_X86_POSTED_MSI +extern void intel_posted_msi_init (void ); + +#else +static inline void intel_posted_msi_init (void ) {}; + +#endif /* X86_POSTED_MSI */ #endif /* _X86_POSTED_INTR_H */
arch/x86/kernel/cpu/common.c +3 -0
@@ -68,6 +68,7 @@#include <asm/traps.h> #include <asm/sev.h> #include <asm/tdx.h> +#include <asm/posted_intr.h> #include "cpu.h"
@@ -2219,6 +2220,8 @@ void cpu_init(void)barrier ();x2apic_setup (); + +intel_posted_msi_init (); }mmgrab (&init_mm);
arch/x86/kernel/irq.c +23 -0
@@ -22,6 +22,8 @@#include <asm/desc.h> #include <asm/traps.h> #include <asm/thermal.h> +#include <asm/posted_intr.h> +#include <asm/irq_remapping.h> #define CREATE_TRACE_POINTS #include <asm/trace/irq_vectors.h>
@@ -334,6 +336,27 @@ DEFINE_IDTENTRY_SYSVEC_SIMPLE(sysvec_kvm_posted_intr_nested_ipi) }#endif +#ifdef CONFIG_X86_POSTED_MSI + +/* Posted Interrupt Descriptors for coalesced MSIs to be posted */ +DEFINE_PER_CPU_ALIGNED(struct pi_desc, posted_interrupt_desc); + +void intel_posted_msi_init (void ) +{ +u32 destination; +u32 apic_id; + +this_cpu_write (posted_interrupt_desc.nv,POSTED_MSI_NOTIFICATION_VECTOR ); + +/* +* APIC destination ID is stored in bit 8:15 while in XAPIC mode. +* VT-d spec. CH 9.11 +*/ + apic_id =this_cpu_read (x86_cpu_to_apicid); + destination =x2apic_enabled () ? apic_id : apic_id <<8 ; +this_cpu_write (posted_interrupt_desc.ndst, destination); +} +#endif /* X86_POSTED_MSI */ #ifdef CONFIG_HOTPLUG_CPU /* A cpu has been removed from cpu_online_mask. Reset irq affinities. */
Signature · 1 line
2.25.1
RE: [PATCH v2 06/13] x86/irq: Set up per host CPU posted interrupt descriptors
arch/x86/kernel/irq.c DEFINE_IDTENTRY_SYSVEC_SIMPLE(sysvec_kvm_posted_intr_nested_ipi) · line 342
Quoted 7 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMarch/x86/kernel/irq.c +4
-0+#ifdef CONFIG_X86_POSTED_MSI + +/* Posted Interrupt Descriptors for coalesced MSIs to be posted */ +DEFINE_PER_CPU_ALIGNED(struct pi_desc, posted_interrupt_desc);
'posted_msi_desc' to be more accurate?
Re: [PATCH v2 06/13] x86/irq: Set up per host CPU posted interrupt descriptors
Hi Kevin,
Quoted 9 lines · Kevin
arch/x86/kernel/irq.c DEFINE_IDTENTRY_SYSVEC_SIMPLE(sysvec_kvm_posted_intr_nested_ipi) · line 342
Quoted 7 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMarch/x86/kernel/irq.c +4
-0+#ifdef CONFIG_X86_POSTED_MSI + +/* Posted Interrupt Descriptors for coalesced MSIs to be posted */ +DEFINE_PER_CPU_ALIGNED(struct pi_desc, posted_interrupt_desc);'
posted_msi_desc' to be more accurate?
makes sense, will do.
Thanks,
Jacob
PATCH v2 7/13 x86/irq: Factor out calling ISR from common_interrupt
In reply to Jacob PanPrepare for calling external IRQ handlers directly from the posted MSI demultiplexing loop. Extract the common code with common interrupt to avoid code duplication.
--- arch/x86/kernel/irq.c | 23 ++++++++++++++--------- 1 file changed, 14 insertions(+), 9 deletions(-)
arch/x86/kernel/irq.c +14 -9
@@ -242,18 +242,10 @@ static __always_inline void handle_irq(struct irq_desc *desc,__handle_irq (desc, regs); } -/* -* common_interrupt() handles all normal device IRQ's (the special SMP -* cross-CPU interrupts have their own entry points). -*/ -DEFINE_IDTENTRY_IRQ (common_interrupt) +static __always_inline void call_irq_handler(int vector, struct pt_regs *regs) { -struct pt_regs *old_regs =set_irq_regs (regs);struct irq_desc *desc; -/* entry code tells RCU that we're not quiescent. Check it. */ -RCU_LOCKDEP_WARN (!rcu_is_watching (),"IRQ failed to wake up RCU" ); - desc =__this_cpu_read (vector_irq[vector]);if (likely (!IS_ERR_OR_NULL (desc))) {handle_irq (desc, regs);
@@ -268,7 +260,20 @@ DEFINE_IDTENTRY_IRQ(common_interrupt)__this_cpu_write (vector_irq[vector],VECTOR_UNUSED ); } } +} + +/* +* common_interrupt() handles all normal device IRQ's (the special SMP +* cross-CPU interrupts have their own entry points). +*/ +DEFINE_IDTENTRY_IRQ (common_interrupt) +{ +struct pt_regs *old_regs =set_irq_regs (regs); + +/* entry code tells RCU that we're not quiescent. Check it. */ +RCU_LOCKDEP_WARN (!rcu_is_watching (),"IRQ failed to wake up RCU" ); +call_irq_handler (vector, regs);set_irq_regs (old_regs); }
Signature · 1 line
2.25.1
RE: [PATCH v2 07/13] x86/irq: Factor out calling ISR from common_interrupt
Quoted 36 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMPrepare for calling external IRQ handlers directly from the posted MSI demultiplexing loop. Extract the common code with common interrupt to avoid code duplication.
Signed-off-by Jacob Pan <[email protected]>--- arch/x86/kernel/irq.c | 23 ++++++++++++++--------- 1 file changed, 14 insertions(+), 9 deletions(-)arch/x86/kernel/irq.c +1
-9@@ -242,18 +242,10 @@ static __always_inline void handle_irq(structirq_desc *desc,__handle_irq (desc ,regs ); } -/* -* common_interrupt() handles all normal device IRQ's (the special SMP -* cross-CPU interrupts have their own entry points). -*/ -DEFINE_IDTENTRY_IRQ (common_interrupt) +static __always_inline void call_irq_handler(int vector, struct pt_regs *regs) { -struct pt_regs *old_regs =set_irq_regs (regs);struct irq_desc *desc; -/* entry code tells RCU that we're not quiescent. Check it. */ -RCU_LOCKDEP_WARN (!rcu_is_watching (),"IRQ failed to wake up RCU"); - desc =__this_cpu_read (vector_irq[vector]);
arch/x86/kernel/irq.c static __always_inline void handle_irq(struct irq_desc *desc, · line 251
Quoted 2 lines
arch/x86/kernel/irq.c +0
-0if (likely (!IS_ERR_OR_NULL (desc))) {handle_irq (desc, regs);
the hidden lines has one problem:
} else {
apic_eoi();
if (desc == VECTOR_UNUSED) {
...there will be two EOI's for unused vectors, adding the one in sysvec_posted_msi_notification().
Re: [PATCH v2 07/13] x86/irq: Factor out calling ISR from common_interrupt
Hi Kevin,
Quoted 41 lines · Kevin
Quoted 36 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMPrepare for calling external IRQ handlers directly from the posted MSI demultiplexing loop. Extract the common code with common interrupt to avoid code duplication.
Signed-off-by Jacob Pan <[email protected]>--- arch/x86/kernel/irq.c | 23 ++++++++++++++--------- 1 file changed, 14 insertions(+), 9 deletions(-)arch/x86/kernel/irq.c +1
-9@@ -242,18 +242,10 @@ static __always_inline void handle_irq(structirq_desc *desc,__handle_irq (desc ,regs ); } -/* -* common_interrupt() handles all normal device IRQ's (the special SMP -* cross-CPU interrupts have their own entry points). -*/ -DEFINE_IDTENTRY_IRQ (common_interrupt) +static __always_inline void call_irq_handler(int vector, structpt_regs *regs) { -struct pt_regs *old_regs =set_irq_regs (regs);struct irq_desc *desc; -/* entry code tells RCU that we're not quiescent. Check it. */ -RCU_LOCKDEP_WARN (!rcu_is_watching (),"IRQ failed to wake up RCU"); - desc =__this_cpu_read (vector_irq[vector]);arch/x86/kernel/irq.c static __always_inline void handle_irq(struct irq_desc *desc, · line 251
Quoted 2 lines
arch/x86/kernel/irq.c +0
-0if (likely (!IS_ERR_OR_NULL (desc))) {handle_irq (desc, regs);the hidden lines has one problem:
Quoted 8 lines
} else { apic_eoi(); if (desc == VECTOR_UNUSED) { ...there will be two EOI's for unused vectors, adding the one in
sysvec_posted_msi_notification().
Indeed this unlikely case could cause lost interrupt. Imagine we have:
- IDT vector N (MSI notification), O, and P (other high-priority system vectors).
- Device MSI vector A which triggers N.
Action APIC IRR APIC ISR
---------------------------------------------------------
Device MSI A N
APIC accepts N - N
New IRQs arrive O,P Nhandle_irq(A)eoi() due to A's fault - O,P
eoi in post_msi - P
----------------------------------------------------------
The second EOI clears ISR for vector O but missed processing it.
Intel SDM 11.8.4 for background.
"The IRR contains the active interrupt requests that have been accepted, but not yet dispatched to the processor for servicing. When the local APIC accepts an interrupt, it sets the bit in the IRR that corresponds the vector of the accepted interrupt. When the processor core is ready to handle the next interrupt, the local APIC clears the highest priority IRR bit that is set and sets the corresponding ISR bit. The vector for the highest priority bit set in the ISR is then dispatched to the processor core for servicing.
While the processor is servicing the highest priority interrupt, the local APIC can send additional fixed interrupts by setting bits in the IRR. When the interrupt service routine issues a write to the EOI register (see Section 11.8.5, Signaling Interrupt Servicing Completion), the local APIC responds by clearing the highest priority ISR bit that is set. It then repeats the process of clearing the highest priority bit in the IRR and setting the corresponding bit in the ISR. The processor core then begins executing the service routing for the highest priority bit set in the ISR "
I need to avoid the duplicated EOI in this case and at minimum cost for the hot path.
Thanks,
Jacob
PATCH v2 8/13 x86/irq: Install posted MSI notification handler
In reply to Jacob PanAll MSI vectors are multiplexed into a single notification vector when posted MSI is enabled. It is the responsibility of the notification vector handler to demultiplex MSI vectors. In this handler, for each pending bit, MSI vector handlers are dispatched without IDT delivery.
For example, the interrupt flow will change as follows:
(3 MSIs of different vectors arrive in a a high frequency burst)
BEFORE:
interrupt(MSI)
irq_enter()
handler() /* EOI */
irq_exit()
process_softirq()
interrupt(MSI)
irq_enter()
handler() /* EOI */
irq_exit()
process_softirq()
interrupt(MSI)
irq_enter()
handler() /* EOI */
irq_exit()
process_softirq()AFTER:
interrupt /* Posted MSI notification vector */
irq_enter()
atomic_xchg(PIR)
handler()
handler()
handler()
pi_clear_on()
apic_eoi()
irq_exit()
process_softirq()Except for the leading MSI, CPU notifications are skipped/coalesced.
For MSIs arrive at a low frequency, the demultiplexing loop does not wait for more interrupts to coalesce. Therefore, there's no additional latency other than the processing time.
---
v2:
- Delete extra inline attribute
- Fix pir pointer in xchg (Zeng Guang)
--- arch/x86/include/asm/hardirq.h | 3 + arch/x86/include/asm/idtentry.h | 3 + arch/x86/kernel/idt.c | 3 + arch/x86/kernel/irq.c | 113 ++++++++++++++++++++++++++++++++ 4 files changed, 122 insertions(+)
arch/x86/include/asm/hardirq.h +3 -0
@@ -44,6 +44,9 @@ typedef struct {unsigned int irq_hv_reenlightenment_count;unsigned int hyperv_stimer0_count;#endif +#ifdef CONFIG_X86_POSTED_MSI +unsigned int posted_msi_notification_count; +#endif }____cacheline_aligned irq_cpustat_t;DECLARE_PER_CPU_SHARED_ALIGNED (irq_cpustat_t, irq_stat);
arch/x86/include/asm/idtentry.h +3 -0
@@ -701,6 +701,9 @@ DECLARE_IDTENTRY_SYSVEC(ERROR_APIC_VECTOR, sysvec_error_interrupt);DECLARE_IDTENTRY_SYSVEC (SPURIOUS_APIC_VECTOR , sysvec_spurious_apic_interrupt);DECLARE_IDTENTRY_SYSVEC (LOCAL_TIMER_VECTOR , sysvec_apic_timer_interrupt);DECLARE_IDTENTRY_SYSVEC (X86_PLATFORM_IPI_VECTOR , sysvec_x86_platform_ipi); +# ifdef CONFIG_X86_POSTED_MSI +DECLARE_IDTENTRY_SYSVEC (POSTED_MSI_NOTIFICATION_VECTOR , sysvec_posted_msi_notification); +# endif #endif #ifdef CONFIG_SMP
arch/x86/kernel/idt.c +3 -0
@@ -163,6 +163,9 @@ static const __initconst struct idt_data apic_idts[] = {# endif INTG(SPURIOUS_APIC_VECTOR, asm_sysvec_spurious_apic_interrupt), INTG(ERROR_APIC_VECTOR, asm_sysvec_error_interrupt), +# ifdef CONFIG_X86_POSTED_MSI + INTG(POSTED_MSI_NOTIFICATION_VECTOR, asm_sysvec_posted_msi_notification), +# endif #endif };
arch/x86/kernel/irq.c +113 -0
125 lines not on this page · raw message
Signature · 1 line
2.25.1
RE: [PATCH v2 08/13] x86/irq: Install posted MSI notification handler
Quoted 21 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMarch/x86/kernel/irq.c +4
-0+ +/* +* De-multiplexing posted interrupts is on the performance path, the code +* below is written to optimize the cache performance based on the following
+ * considerations:
+ * 1.Posted interrupt descriptor (PID) fits in a cache line that is frequently
+ * accessed by both CPU and IOMMU.
+ * 2.During posted MSI processing, the CPU needs to do 64-bit read and xchg
+ * for checking and clearing posted interrupt request (PIR), a 256 bit field
+ * within the PID.
+ * 3.On the other side, the IOMMU does atomic swaps of the entire PID cache
+ * line when posting interrupts and setting control bits.
+ * 4.The CPU can access the cache line a magnitude faster than the IOMMU.
+ * 5.Each time the IOMMU does interrupt posting to the PIR will evict the PID
Quoted 8 lines
+ * cache line. The cache line states after each operation are as follows: + * CPU IOMMU PID Cache line state + * ---------------------------------------------------------------+ *...read64 exclusive
+ *...lock xchg64 modified
+ *... post/atomic swap invalid
+ *...-------------------------------------------------------------
+ *
According to VT-d spec: 5.2.3 Interrupt-Posting Hardware Operation:
"
- Read contents of the Posted Interrupt Descriptor, claiming exclusive ownership of its hosting cache-line.
...
- Modify the following descriptor field values atomically:
...
- Promote the cache-line to be globally observable, so that the modifications are visible to other caching agents. Hardware may write-back the cache-line anytime after this step.
"
sounds that the PID cache line is not evicted after IOMMU posting?
Re: [PATCH v2 08/13] x86/irq: Install posted MSI notification handler
Hi Kevin,
Quoted 49 lines · Kevin
Quoted 27 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMarch/x86/kernel/irq.c +3
-0+ +/* +* De-multiplexing posted interrupts is on the performance path, the code
+ * below is written to optimize the cache performance based on the following
+ * considerations:
+ * 1.Posted interrupt descriptor (PID) fits in a cache line that is frequently
+ * accessed by both CPU and IOMMU.
+ * 2.During posted MSI processing, the CPU needs to do 64-bit read and xchg
+ * for checking and clearing posted interrupt request (PIR), a 256
bit field
+ * within the PID.
+ * 3.On the other side, the IOMMU does atomic swaps of the entire PID cache
+ * line when posting interrupts and setting control bits.
+ * 4.The CPU can access the cache line a magnitude faster than the IOMMU.
+ * 5.Each time the IOMMU does interrupt posting to the PIR will evict the PID
+ * cache line. The cache line states after each operation are as
follows:Quoted 8 lines
+ * CPU IOMMU PID Cache line
state
+ * ---------------------------------------------------------------
+ *...read64 exclusive
+ *...lock xchg64 modified
+ *... post/atomic swap invalid
+ *...-------------------------------------------------------------
+ *According to VT-d spec: 5.2.3 Interrupt-Posting Hardware Operation:
"
- Read contents of the Posted Interrupt Descriptor, claiming exclusive ownership of its hosting cache-line.
...
- Modify the following descriptor field values atomically:
...
- Promote the cache-line to be globally observable, so that the modifications are visible to other caching agents. Hardware may write-back the cache-line anytime after this step.
"
Quoted 1 line
sounds that the PID cache line is not evicted after IOMMU posting?
IOMMU follows the same MESI protocol defined in SDM. VT-d spec. also says "This atomic read-modify-write operation will always snoop processor caches"
So if the PID cache line is in modified state (caused by writing ON bit, clear PIR, etc.), IOMMU request ownership will evict the cache.
Thanks,
Jacob
Re: [PATCH v2 08/13] x86/irq: Install posted MSI notification handler
In reply to Jacob PanQuoted 7 lines · Jacob Pan
#ifdef
CONFIG_SMParch/x86/kernel/idt.c +0
-0
arch/x86/kernel/idt.c static const __initconst struct idt_data apic_idts[] = { · line 170
Quoted 8 lines
arch/x86/kernel/idt.c +3
-0@@ -163,6 +163,9 @@ static const __initconst struct idt_data apic_idts[] = {# endif INTG(SPURIOUS_APIC_VECTOR, asm_sysvec_spurious_apic_interrupt), INTG(ERROR_APIC_VECTOR, asm_sysvec_error_interrupt), +# ifdef CONFIG_X86_POSTED_MSI + INTG(POSTED_MSI_NOTIFICATION_VECTOR, asm_sysvec_posted_msi_notification), +# endif #endif };
Obviously lacks FRED support...
Re: [PATCH v2 08/13] x86/irq: Install posted MSI notification handler
Hi Thomas,
Quoted 11 lines · Thomas Gleixner
Quoted 10 lines · Jacob Pan
#ifdef
CONFIG_SMParch/x86/kernel/idt.c +0
-0@@ -163,6 +163,9 @@ static const __initconst struct idt_dataapic_idts[] = { # endif
INTG(SPURIOUS_APIC_VECTOR,asm_sysvec_spurious_apic_interrupt), INTG(ERROR_APIC_VECTOR,
Quoted 8 lines
Quoted 6 lines
asm_sysvec_error_interrupt), +# ifdefCONFIG_X86_POSTED_MSIarch/x86/kernel/idt.c +1
-0+ INTG(POSTED_MSI_NOTIFICATION_VECTOR,
asm_sysvec_posted_msi_notification), +# endif
#endif
};Obviously lacks FRED support...
Good point, forgot FRED is merged :)
Will add an entry to entry_fred.c. I would not be able to test performance though.
Thanks,
Jacob
PATCH v2 9/13 x86/irq: Factor out common code for checking pending interrupts
In reply to Jacob PanUse a common function for checking pending interrupt vector in APIC IRR instead of duplicated open coding them.
Additional checks for posted MSI vectors can then be contained in this function.
--- arch/x86/include/asm/apic.h | 11 +++++++++++ arch/x86/kernel/apic/vector.c | 5 ++--- arch/x86/kernel/irq.c | 5 ++--- 3 files changed, 15 insertions(+), 6 deletions(-)
arch/x86/include/asm/apic.h +11 -0
@@ -499,6 +499,17 @@ static inline bool lapic_vector_set_in_irr(unsigned int vector)return !!(irr & (1U << (vector %32 ))); } +static inline bool is_vector_pending (unsigned int vector) +{ +unsigned int irr; + + irr =apic_read (APIC_IRR + (vector /32 *0x10 )); +if (irr & (1 << (vector %32 ))) +return true; + +return false; +} +/* * Warm reset vector position: */
arch/x86/kernel/apic/vector.c +2 -3
@@ -965,7 +965,7 @@ static void __vector_cleanup(struct vector_cleanup *cl, bool check_irr)lockdep_assert_held (&vector_lock);hlist_for_each_entry_safe (apicd, tmp, &cl->head, clist) { -unsigned int irr, vector = apicd->prev_vector; +unsigned int vector = apicd->prev_vector;/* * Paranoia: Check if the vector that needs to be cleaned
@@ -979,8 +979,7 @@ static void __vector_cleanup(struct vector_cleanup *cl, bool check_irr)* fixup_irqs() was just called to scan IRR for set bits and * forward them to new destination CPUs via IPIs. */ - irr = check_irr ?apic_read (APIC_IRR + (vector /32 *0x10 )) :0 ; -if (irr & (1U << (vector %32 ))) { +if (check_irr &&is_vector_pending (vector)) {pr_warn_once ("Moved interrupt pending in old target APIC %u\n" , apicd->irq); rearm = true;continue ;
arch/x86/kernel/irq.c +2 -3
@@ -480,7 +480,7 @@ DEFINE_IDTENTRY_SYSVEC(sysvec_posted_msi_notification)/* A cpu has been removed from cpu_online_mask. Reset irq affinities. */ void fixup_irqs (void ) { -unsigned int irr, vector; +unsigned int vector;struct irq_desc *desc;struct irq_data *data;struct irq_chip *chip;
@@ -507,8 +507,7 @@ void fixup_irqs(void)if (IS_ERR_OR_NULL (__this_cpu_read (vector_irq[vector])))continue ; - irr =apic_read (APIC_IRR + (vector /32 *0x10 )); -if (irr & (1 << (vector %32 ))) { +if (is_vector_pending (vector)) { desc =__this_cpu_read (vector_irq[vector]);raw_spin_lock (&desc->lock);
Signature · 1 line
2.25.1
PATCH v2 10/13 x86/irq: Extend checks for pending vectors to posted interrupts
In reply to Jacob PanDuring interrupt affinity change, it is possible to have interrupts delivered to the old CPU after the affinity has changed to the new one. To prevent lost interrupts, local APIC IRR is checked on the old CPU. Similar checks must be done for posted MSIs given the same reason.
Consider the following scenario:
Device system agent iommu memory CPU/LAPIC
1 FEEX_XXXX
2 Interrupt request
3 Fetch IRTE ->
4 ->Atomic Swap PID.PIR(vec)
Push to Global Observable(GO)
5 if (ON*)
i done;* else
6 send a notification ->
* ON: outstanding notification, 1 will suppress new notifications
If the affinity change happens between 3 and 5 in IOMMU, the old CPU's posted interrupt request (PIR) could have pending bit set for the vector being moved.
This patch adds a helper function to check individual vector status. Then use the helper to check for pending interrupts on the source CPU's PID.
---
v2: Fold in helper function patch.
--- arch/x86/include/asm/apic.h | 3 ++- arch/x86/include/asm/posted_intr.h | 18 ++++++++++++++++++ 2 files changed, 20 insertions(+), 1 deletion(-)
arch/x86/include/asm/apic.h +2 -1
@@ -13,6 +13,7 @@#include <asm/mpspec.h> #include <asm/msr.h> #include <asm/hardirq.h> +#include <asm/posted_intr.h> #define ARCH_APICTIMER_STOPS_ON_C3 1
@@ -507,7 +508,7 @@ static inline bool is_vector_pending(unsigned int vector)if (irr & (1 << (vector %32 )))return true; -return false; +return pi_pending_this_cpu (vector); }/*
arch/x86/include/asm/posted_intr.h +18 -0
@@ -1,6 +1,7 @@/* SPDX-License-Identifier: GPL-2.0 */ #ifndef _X86_POSTED_INTR_H#define _X86_POSTED_INTR_H +#include <asm/irq_vectors.h> #define POSTED_INTR_ON 0#define POSTED_INTR_SN 1
@@ -81,9 +82,26 @@ static inline bool pi_test_sn(struct pi_desc *pi_desc) }#ifdef CONFIG_X86_POSTED_MSI +/* +* Not all external vectors are subject to interrupt remapping, e.g. IOMMU's +* own interrupts. Here we do not distinguish them since those vector bits in +* PIR will always be zero. +*/ +static inline bool pi_pending_this_cpu (unsigned int vector) +{ +struct pi_desc *pid =this_cpu_ptr (&posted_interrupt_desc); + +if (WARN_ON_ONCE (vector >NR_VECTORS || vector <FIRST_EXTERNAL_VECTOR )) +return false; + +return test_bit (vector, (unsigned long *)pid->pir); +} +extern void intel_posted_msi_init (void );#else +static inline bool pi_pending_this_cpu (unsigned int vector) {return false; } +static inline void intel_posted_msi_init (void ) {};#endif /* X86_POSTED_MSI */
Signature · 1 line
2.25.1
RE: [PATCH v2 10/13] x86/irq: Extend checks for pending vectors to posted interrupts
Quoted 18 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMDuring interrupt affinity change, it is possible to have interrupts delivered to the old CPU after the affinity has changed to the new one. To prevent lost interrupts, local APIC IRR is checked on the old CPU. Similar checks must be done for posted MSIs given the same reason.
Consider the following scenario:
Device system agent iommu memory
CPU/LAPIC
1FEEX_XXXX
2 Interrupt request
3 Fetch IRTE ->
4 ->Atomic SwapPID.PIR(vec)
Push to Global
Observable(GO)
5 if (ON*)
Quoted 1 line
i done;*
there is a stray 'i'
Quoted 8 lines
else
6 send a notification ->* ON: outstanding notification, 1 will suppress new notifications
If the affinity change happens between 3 and 5 in IOMMU, the old CPU's posted
interrupt request (PIR) could have pending bit set for the vector being moved.
how could affinity change be possible in 3/4 when the cache line is locked by IOMMU? Strictly speaking it's about a change after 4 and before 6.
Re: [PATCH v2 10/13] x86/irq: Extend checks for pending vectors to posted interrupts
Hi Kevin,
Quoted 20 lines · Kevin
Quoted 18 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMDuring interrupt affinity change, it is possible to have interrupts delivered to the old CPU after the affinity has changed to the new one. To prevent lost interrupts, local APIC IRR is checked on the old CPU. Similar checks must be done for posted MSIs given the same reason.
Consider the following scenario:
Device system agent iommu
memory CPU/LAPIC
1FEEX_XXXX
2 Interrupt request
3 Fetch IRTE ->
4 ->Atomic SwapPID.PIR(vec)Push to Global
Observable(GO)
5 if (ON*)Quoted 1 line
i done;*
Quoted 1 line
there is a stray 'i'
will fix, thanks
Quoted 7 lines
Quoted 7 lines
else
6 send a
notification ->* ON: outstanding notification, 1 will suppress new notifications
Quoted 8 lines
Quoted 4 lines
If the affinity change happens between 3 and 5 in IOMMU, the old CPU's posted
interrupt request (PIR) could have pending bit set for the vector being moved.how could affinity change be possible in 3/4 when the cache line is locked by IOMMU? Strictly speaking it's about a change after 4 and before 6.
SW can still perform affinity change on IRTE and do the flushing on IR cache after IOMMU fectched it (step 3). They are async events.
In step 4, the atomic swap is on the PID cacheline, not IRTE.
Thanks,
Jacob
RE: [PATCH v2 10/13] x86/irq: Extend checks for pending vectors to posted interrupts
Quoted 43 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 13, 2024 2:24 AMHi Kevin,
Quoted 20 lines · Kevin
Quoted 18 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMDuring interrupt affinity change, it is possible to have interrupts delivered to the old CPU after the affinity has changed to the new one. To prevent lost interrupts, local APIC IRR is checked on the old CPU. Similar checks must be done for posted MSIs given the same reason.
Consider the following scenario:
Device system agent iommu
memory CPU/LAPIC
1FEEX_XXXX
2 Interrupt request
3 Fetch IRTE ->
4 ->Atomic SwapPID.PIR(vec)Push to Global
Observable(GO)
5 if (ON*)Quoted 1 line
i done;*Quoted 1 line
there is a stray 'i'
will fix, thanks
Quoted 12 lines
Quoted 11 lines
else
6 send a
notification ->* ON: outstanding notification, 1 will suppress new notifications
If the affinity change happens between 3 and 5 in IOMMU, the old CPU's posted
interrupt request (PIR) could have pending bit set for the vector being moved.
Quoted 8 lines
Quoted 3 lines
how could affinity change be possible in 3/4 when the cache line is locked by IOMMU? Strictly speaking it's about a change after 4 and before 6.
SW can still perform affinity change on IRTE and do the flushing on IR cache after IOMMU fectched it (step 3). They are async events.
In step 4, the atomic swap is on the PID cacheline, not IRTE.
yeah, I mixed IRTE with PID.
PATCH v2 11/13 iommu/vt-d: Make posted MSI an opt-in cmdline option
In reply to Jacob PanAdd a command line opt-in option for posted MSI if CONFIG_X86_POSTED_MSI=y.
Also introduce a helper function for testing if posted MSI is supported on the platform.
--- Documentation/admin-guide/kernel-parameters.txt | 1 + arch/x86/include/asm/irq_remapping.h | 11 +++++++++++ drivers/iommu/irq_remapping.c | 13 ++++++++++++- 3 files changed, 24 insertions(+), 1 deletion(-)
Documentation/admin-guide/kernel-parameters.txt +1 -0
@@ -2251,6 +2251,7 @@ no_x2apic_optout BIOS x2APIC opt-out request will be ignored nopost disable Interrupt Posting + posted_msi enable MSIs delivered as posted interrupts iomem= Disable strict checking of access to MMIO memory strict regions from userspace.
arch/x86/include/asm/irq_remapping.h +11 -0
@@ -50,6 +50,17 @@ static inline struct irq_domain *arch_get_ir_parent_domain(void) return x86_vector_domain; } +#ifdef CONFIG_X86_POSTED_MSI +extern int enable_posted_msi; + +static inline bool posted_msi_supported (void ) +{ +return enable_posted_msi &&irq_remapping_cap (IRQ_POSTING_CAP ); +} +#else +static inline bool posted_msi_supported (void ) {return false; }; +#endif +#else /* CONFIG_IRQ_REMAP */ static inline bool irq_remapping_cap (enum irq_remap_cap cap) {return 0 ; }
drivers/iommu/irq_remapping.c +12 -1
@@ -24,6 +24,10 @@ int no_x2apic_optout;int disable_irq_post =0 ; +#ifdef CONFIG_X86_POSTED_MSI +int enable_posted_msi; +#endif +static int disable_irq_remap;static struct irq_remap_ops *remap_ops;
@@ -70,7 +74,14 @@ static __init int setup_irqremap(char *str) no_x2apic_optout =1 ; else if (!strncmp(str, "nopost", 6)) disable_irq_post =1 ; - +#ifdef CONFIG_X86_POSTED_MSI + else if (!strncmp(str, "posted_msi", 10)) { +if (disable_irq_post || disable_irq_remap) +pr_warn ("Posted MSI not enabled due to conflicting options!" ); +else + enable_posted_msi =1 ; + } +#endif str +=strcspn (str,"," );while (*str ==',' ) str++;
Signature · 1 line
2.25.1
Re: [PATCH v2 11/13] iommu/vt-d: Make posted MSI an opt-in cmdline option
Quoted 29 lines · Jacob Pan
Add a command line opt-in option for posted MSI if
CONFIG_X86_POSTED_MSI=y.Also introduce a helper function for testing if posted MSI is supported on the platform.
Signed-off-by Jacob Pan <[email protected]>--- Documentation/admin-guide/kernel-parameters.txt | 1 + arch/x86/include/asm/irq_remapping.h | 11 +++++++++++ drivers/iommu/irq_remapping.c | 13 ++++++++++++- 3 files changed, 24 insertions(+), 1 deletion(-)Documentation/admin-guide/kernel-parameters.txt +1
-0@@ -2251,6 +2251,7 @@ no_x2apic_optout BIOS x2APIC opt-out request will be ignored nopost disable Interrupt Posting + posted_msi enable MSIs delivered as posted interrupts iomem= Disable strict checking of access to MMIO memory strict regions from userspace.arch/x86/include/asm/irq_remapping.h +0
-0
arch/x86/include/asm/irq_remapping.h static inline struct irq_domain *arch_get_ir_parent_domain(void) · line 59
Quoted 10 lines
arch/x86/include/asm/irq_remapping.h +7
-0@@ -50,6 +50,17 @@ static inline struct irq_domain *arch_get_ir_parent_domain(void)return x86_vector_domain; } +#ifdef CONFIG_X86_POSTED_MSI +extern int enable_posted_msi; + +static inline bool posted_msi_supported (void ) +{ +return enable_posted_msi &&irq_remapping_cap (IRQ_POSTING_CAP ); +}
Out of this patch set's scope, but, dropping into irq_remappping_cap(), I'd like to bring this change for discussion:
drivers/iommu/irq_remapping.c +1 -1
@@ -98,7 +98,7 @@ void set_irq_remapping_broken(void)bool irq_remapping_cap (enum irq_remap_cap cap) { -if (!remap_ops || disable_irq_post) +if (!remap_ops || disable_irq_remap)return false;return (remap_ops->capability & (1 << cap));
1. irq_remapping_cap() is to exam some cap, though at present it has only 1 cap, i.e. IRQ_POSTING_CAP, simply return false just because of disable_irq_post isn't good. Instead, IRQ_REMAP is the foundation of all remapping caps.
2. disable_irq_post is used by Intel iommu code only, here irq_remapping_cap() is common code. e.g. AMD iommu code doesn't use it to judge set cap of irq_post or not.
Quoted 3 lines
arch/x86/include/asm/irq_remapping.h +3
-0+#else +static inline bool posted_msi_supported (void ) {return false; }; +#endif
Re: [PATCH v2 11/13] iommu/vt-d: Make posted MSI an opt-in cmdline option
Hi Robert,
Quoted 60 lines · Robert Hoo
Quoted 30 lines · Jacob Pan
Add a command line opt-in option for posted MSI if
CONFIG_X86_POSTED_MSI=y.Also introduce a helper function for testing if posted MSI is supported on the platform.
Signed-off-by Jacob Pan <[email protected]>--- Documentation/admin-guide/kernel-parameters.txt | 1 + arch/x86/include/asm/irq_remapping.h | 11 +++++++++++ drivers/iommu/irq_remapping.c | 13 ++++++++++++- 3 files changed, 24 insertions(+), 1 deletion(-)(quoted diff) +0
-0
b/Documentation/admin-guide/kernel-parameters.txtindexbb884c14b2f6..e5fd02423c4c100644 ---a/Documentation/admin-guide/kernel-parameters.txt+++b/Documentation/admin-guide/kernel-parameters.txt@@ -2251,6 +2251,7 @@
no_x2apic_optout
BIOS x2APIC opt-out request will be
ignored nopost disable Interrupt Posting(quoted diff) +1
-0+ posted_msi enable MSIs delivered as postedinterrupts
iomem= Disable strict checking of access to
MMIO memory strict regions from userspace.(quoted diff) +0
-0
b/arch/x86/include/asm/irq_remapping.hindex7a2ed154a5e1..e46bde61029b100644 ---a/arch/x86/include/asm/irq_remapping.h
+++b/arch/x86/include/asm/irq_remapping.hReview static inline struct irq_domain · line 59
Quoted 10 lines
(quoted diff) +7
-0@@ -50,6 +50,17 @@ static inline struct irq_domain *arch_get_ir_parent_domain(void) return x86_vector_domain; } +#ifdef CONFIG_X86_POSTED_MSI +extern int enable_posted_msi; + +static inline bool posted_msi_supported(void) +{ + return enable_posted_msi && irq_remapping_cap(IRQ_POSTING_CAP); +}Out of this patch set's scope, but, dropping into
irq_remappping_cap(), I'd like to bring this change for discussion:drivers/iommu/irq_remapping.c +1
-1@@ -98,7 +98,7 @@ void set_irq_remapping_broken(void)bool irq_remapping_cap (enum irq_remap_cap cap) { -if (!remap_ops || disable_irq_post) +if (!remap_ops || disable_irq_remap)return false;return (remap_ops->capability & (1 << cap));
Quoted 6 lines
1.
irq_remapping_cap()is to exam some cap, though at present it has only 1 cap, i.e.IRQ_POSTING_CAP, simply return false just because ofdisable_irq_postisn't good. Instead,IRQ_REMAPis the foundation of all remapping caps. 2.disable_irq_postis used by Intel iommu code only, hereirq_remapping_cap()is common code. e.g. AMD iommu code doesn't use it to judge set cap ofirq_postor not.
I agree, posting should be treated as a sub-capability of remapping. IRQ_POSTING_CAP is only set when remapping is on.
We need to delete this such that posting is always off when remapping is off.
drivers/iommu/intel/irq_remapping.c +1 -5
@@ -1038,11 +1038,7 @@ static void disable_irq_remapping(void)iommu_disable_irq_remapping (iommu); } -/* -* Clear Posted-Interrupts capability. -*/ -if (!disable_irq_post) - intel_irq_remap_ops.capability &= ~(1 <<IRQ_POSTING_CAP ); + intel_irq_remap_ops.capability &= ~(1 <<IRQ_POSTING_CAP ); }
Quoted 4 lines
Review static inline struct irq_domain · line 62
Quoted 3 lines
(quoted diff) +3
-0+#else +static inline bool posted_msi_supported(void) { return false; }; +#endif
Thanks,
Jacob
Re: [PATCH v2 11/13] iommu/vt-d: Make posted MSI an opt-in cmdline option
Quoted 81 lines · Jacob Pan
Hi Robert,
Quoted 60 lines · Robert Hoo
Quoted 30 lines · Jacob Pan
Add a command line opt-in option for posted MSI if
CONFIG_X86_POSTED_MSI=y.Also introduce a helper function for testing if posted MSI is supported on the platform.
Signed-off-by Jacob Pan <[email protected]>--- Documentation/admin-guide/kernel-parameters.txt | 1 + arch/x86/include/asm/irq_remapping.h | 11 +++++++++++ drivers/iommu/irq_remapping.c | 13 ++++++++++++- 3 files changed, 24 insertions(+), 1 deletion(-)(quoted diff) +0
-0
b/Documentation/admin-guide/kernel-parameters.txtindexbb884c14b2f6..e5fd02423c4c100644 ---a/Documentation/admin-guide/kernel-parameters.txt+++b/Documentation/admin-guide/kernel-parameters.txt@@ -2251,6 +2251,7 @@
no_x2apic_optout
BIOS x2APIC opt-out request will be
ignored nopost disable Interrupt Posting(quoted diff) +1
-0+ posted_msi enable MSIs delivered as postedinterrupts
iomem= Disable strict checking of access to
MMIO memory strict regions from userspace.(quoted diff) +0
-0
b/arch/x86/include/asm/irq_remapping.hindex7a2ed154a5e1..e46bde61029b100644 ---a/arch/x86/include/asm/irq_remapping.h
+++b/arch/x86/include/asm/irq_remapping.hReview static inline struct irq_domain · line 59
Quoted 10 lines
(quoted diff) +7
-0@@ -50,6 +50,17 @@ static inline struct irq_domain *arch_get_ir_parent_domain(void) return x86_vector_domain; } +#ifdef CONFIG_X86_POSTED_MSI +extern int enable_posted_msi; + +static inline bool posted_msi_supported(void) +{ + return enable_posted_msi && irq_remapping_cap(IRQ_POSTING_CAP); +}Out of this patch set's scope, but, dropping into
irq_remappping_cap(), I'd like to bring this change for discussion:drivers/iommu/irq_remapping.c +1
-1@@ -98,7 +98,7 @@ void set_irq_remapping_broken(void)bool irq_remapping_cap (enum irq_remap_cap cap) { -if (!remap_ops || disable_irq_post) +if (!remap_ops || disable_irq_remap)return false;return (remap_ops->capability & (1 << cap));Quoted 6 lines
1.
irq_remapping_cap()is to exam some cap, though at present it has only 1 cap, i.e.IRQ_POSTING_CAP, simply return false just because ofdisable_irq_postisn't good. Instead,IRQ_REMAPis the foundation of all remapping caps. 2.disable_irq_postis used by Intel iommu code only, hereirq_remapping_cap()is common code. e.g. AMD iommu code doesn't use it to judge set cap ofirq_postor not.I agree, posting should be treated as a sub-capability of remapping.
IRQ_POSTING_CAPis only set when remapping is on.We need to delete this such that posting is always off when remapping is off.
drivers/iommu/intel/irq_remapping.c +0
-0@@ -1038,11 +1038,7 @@ static void disable_irq_remapping(void)iommu_disable_irq_remapping (iommu);
drivers/iommu/intel/irq_remapping.c static void disable_irq_remapping(void) · line 1043
Quoted 10 lines
drivers/iommu/intel/irq_remapping.c +1
-5} -/* -* Clear Posted-Interrupts capability. -*/ -if (!disable_irq_post) - intel_irq_remap_ops.capability &= ~(1 <<IRQ_POSTING_CAP ); + intel_irq_remap_ops.capability &= ~(1 <<IRQ_POSTING_CAP ); }
Right.
RE: [PATCH v2 11/13] iommu/vt-d: Make posted MSI an opt-in cmdline option
In reply to Jacob Pandrivers/iommu/irq_remapping.c static __init int setup_irqremap(char *str) · line 84
Quoted 12 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMdrivers/iommu/irq_remapping.c +4
-0+#ifdef CONFIG_X86_POSTED_MSI + else if (!strncmp(str, "posted_msi", 10)) { +if (disable_irq_post || disable_irq_remap) + pr_warn("Posted MSI not enabled due toconflicting options!");
drivers/iommu/irq_remapping.c +4
-0+ else + enable_posted_msi = 1; + } +#endif
the check of disable_irq_remap is unnecessary. It's unlikely to have a configuration with disable_irq_post=0 while disable_irq_remap=1 given the latter has bigger scope.
but thinking more do we really need a check here? there is no order guarantee that "posted_msi" is parsed after the parameters deciding the value of two disable variables.
it probably makes more sense to just set enable_posted_msi here and then do all required checks when picking up the irqchip in intel_irq_remapping_alloc().
Re: [PATCH v2 11/13] iommu/vt-d: Make posted MSI an opt-in cmdline option
Hi Kevin,
Quoted 17 lines · Kevin
drivers/iommu/irq_remapping.c static __init int setup_irqremap(char *str) · line 84
Quoted 12 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AMdrivers/iommu/irq_remapping.c +4
-0+#ifdef CONFIG_X86_POSTED_MSI + else if (!strncmp(str, "posted_msi", 10)) { +if (disable_irq_post || disable_irq_remap) + pr_warn("Posted MSI not enabled due toconflicting options!");
drivers/iommu/irq_remapping.c +4
-0+ else + enable_posted_msi = 1; + } +#endif the check of
disable_irq_remapis unnecessary. It's unlikely to have a configuration with disable_irq_post=0 while disable_irq_remap=1 given the latter has bigger scope.
Quoted 7 lines
but thinking more do we really need a check here? there is no order guarantee that "
posted_msi" is parsed after the parameters deciding the value of two disable variables.it probably makes more sense to just set
enable_posted_msihere and then do all required checks when picking up the irqchip inintel_irq_remapping_alloc().
Makes sense, I have a helper function posted_msi_supported() called in intel_irq_remapping_alloc() already.
My intention was to alert negligent users, but is is not really necessary as you said.
Thanks,
Jacob
PATCH v2 12/13 iommu/vt-d: Add an irq_chip for posted MSIs
In reply to Jacob PanIntroduce a new irq_chip for posted MSIs, the key difference is in irq_ack where EOI is performed by the notification handler.
When posted MSI is enabled, MSI domain/chip hierarchy will look like this example:
domain: IR-PCI-MSIX-0000:50:00.0-12
hwirq: 0x29
chip: IR-PCI-MSIX-0000:50:00.0
flags: 0x430
IRQCHIP_SKIP_SET_WAKE
IRQCHIP_ONESHOT_SAFE
parent:
domain: INTEL-IR-10-13
hwirq: 0x2d0000
chip: INTEL-IR-POST
flags: 0x0
parent:
domain: VECTOR
hwirq: 0x77
chip: APIC--- drivers/iommu/intel/irq_remapping.c | 46 +++++++++++++++++++++++++++++ 1 file changed, 46 insertions(+)
drivers/iommu/intel/irq_remapping.c +46 -0
@@ -1233,6 +1233,52 @@ static struct irq_chip intel_ir_chip = { .irq_set_vcpu_affinity = intel_ir_set_vcpu_affinity, }; +static void dummy (struct irq_data *d) +{ +} + +/* +* With posted MSIs, all vectors are multiplexed into a single notification +* vector. Devices MSIs are then dispatched in a demux loop where +* EOIs can be coalesced as well. +* +* "INTEL-IR-POST" IRQ chip does not do EOI on ACK, thus the dummy irq_ack() +* function. Instead EOI is performed by the posted interrupt notification +* handler. +* +* For the example below, 3 MSIs are coalesced into one CPU notification. Only +* one apic_eoi() is needed. +* +* __sysvec_posted_msi_notification() +* irq_enter(); +* handle_edge_irq() +* irq_chip_ack_parent() +* dummy(); // No EOI +* handle_irq_event() +* driver_handler() +* irq_enter(); +* handle_edge_irq() +* irq_chip_ack_parent() +* dummy(); // No EOI +* handle_irq_event() +* driver_handler() +* irq_enter(); +* handle_edge_irq() +* irq_chip_ack_parent() +* dummy(); // No EOI +* handle_irq_event() +* driver_handler() +* apic_eoi() +* irq_exit() +*/ +static struct irq_chip intel_ir_chip_post_msi = { + .name ="INTEL-IR-POST" , + .irq_ack = dummy, + .irq_set_affinity = intel_ir_set_affinity, + .irq_compose_msi_msg = intel_ir_compose_msi_msg, + .irq_set_vcpu_affinity = intel_ir_set_vcpu_affinity, +}; +static void fill_msi_msg (struct msi_msg *msg,u32 index,u32 subhandle) {memset (msg,0 ,sizeof (*msg));
Signature · 1 line
2.25.1
RE: [PATCH v2 12/13] iommu/vt-d: Add an irq_chip for posted MSIs
Quoted 19 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AM+ *
+ * For the example below, 3 MSIs are coalesced into one CPU notification. Only
+ * oneapic_eoi()is needed.
+ *
+ *__sysvec_posted_msi_notification()
+ *irq_enter();
+ *handle_edge_irq()
+ *irq_chip_ack_parent()
+ *dummy(); // No EOI
+ *handle_irq_event()
+ *driver_handler()
+ *irq_enter();
+ *handle_edge_irq()
+ *irq_chip_ack_parent()
+ *dummy(); // No EOI
Quoted 8 lines
+ *
handle_irq_event()
+ *driver_handler()
+ *irq_enter();
+ *handle_edge_irq()
+ *irq_chip_ack_parent()
+ *dummy(); // No EOI
+ *handle_irq_event()
+ *driver_handler()
typo: you added three irq_enter()'s here
drivers/iommu/intel/irq_remapping.c static struct irq_chip intel_ir_chip = { · line 1280
Quoted 10 lines
+ *
apic_eoi()
+ *irq_exit()
+ */drivers/iommu/intel/irq_remapping.c +7
-0+static struct irq_chip intel_ir_chip_post_msi = { + .name ="INTEL-IR-POST" , + .irq_ack = dummy, + .irq_set_affinity = intel_ir_set_affinity, + .irq_compose_msi_msg = intel_ir_compose_msi_msg, + .irq_set_vcpu_affinity = intel_ir_set_vcpu_affinity, +};
What about putting this patch at end of the series (combining the change in intel_irq_remapping_alloc()) to finally enable this feature?
It reads slightly better to me to first get those callbacks extended to deal with the new mechanism (i.e. most changes in patch13) before using them in the new irqchip. 😊
Re: [PATCH v2 12/13] iommu/vt-d: Add an irq_chip for posted MSIs
Hi Kevin,
Quoted 28 lines · Kevin
Quoted 19 lines
From: Jacob Pan <[email protected]>
Sent: Saturday, April 6, 2024 6:31 AM+ *
+ * For the example below, 3 MSIs are coalesced into one CPU notification. Only
+ * oneapic_eoi()is needed.
+ *
+ *__sysvec_posted_msi_notification()
+ *irq_enter();
+ *handle_edge_irq()
+ *irq_chip_ack_parent()
+ *dummy(); // No EOI
+ *handle_irq_event()
+ *driver_handler()
+ *irq_enter();
+ *handle_edge_irq()
+ *irq_chip_ack_parent()
+ *dummy(); // No EOIQuoted 8 lines
+ *
handle_irq_event()
+ *driver_handler()
+ *irq_enter();
+ *handle_edge_irq()
+ *irq_chip_ack_parent()
+ *dummy(); // No EOI
+ *handle_irq_event()
+ *driver_handler()
Quoted 1 line
typo: you added three irq_enter()'s here
right, will remove the middle two.
Quoted 16 lines
drivers/iommu/intel/irq_remapping.c static struct irq_chip intel_ir_chip = { · line 1280
Quoted 11 lines
+ *
apic_eoi()
+ *irq_exit()
+ */drivers/iommu/intel/irq_remapping.c +7
-0+static struct irq_chip intel_ir_chip_post_msi = { + .name ="INTEL-IR-POST" , + .irq_ack = dummy, + .irq_set_affinity = intel_ir_set_affinity, + .irq_compose_msi_msg = intel_ir_compose_msi_msg, + .irq_set_vcpu_affinity = intel_ir_set_vcpu_affinity, +};What about putting this patch at end of the series (combining the change in
intel_irq_remapping_alloc()) to finally enable this feature?
Quoted 3 lines
It reads slightly better to me to first get those callbacks extended to deal with the new mechanism (i.e. most changes in patch13) before using them in the new irqchip. 😊
makes sense, will do.
Thanks,
Jacob
PATCH v2 13/13 iommu/vt-d: Enable posted mode for device MSIs
In reply to Jacob PanWith posted MSI feature enabled on the CPU side, iommu interrupt remapping table entries (IRTEs) for device MSI/x can be allocated, activated, and programed in posted mode. This means that IRTEs are linked with their respective PIDs of the target CPU.
Handlers for the posted MSI notification vector will de-multiplex device MSI handlers. CPU notifications are coalesced if interrupts arrive at a high frequency.
Excluding the following:
- legacy devices IOAPIC, HPET (may be needed for booting, not a source of high MSIs)
- VT-d's own IRQs (not remappable).
---
v2: Fold in helper function for retrieving PID address
v1: Added a warning if the effective affinity mask is not set up
--- drivers/iommu/intel/irq_remapping.c | 69 +++++++++++++++++++++++++++-- 1 file changed, 65 insertions(+), 4 deletions(-)
drivers/iommu/intel/irq_remapping.c +65 -4
@@ -19,6 +19,7 @@#include <asm/cpu.h> #include <asm/irq_remapping.h> #include <asm/pci-direct.h> +#include <asm/posted_intr.h> #include "iommu.h" #include "../irq_remapping.h"
@@ -49,6 +50,7 @@ struct irq_2_iommu {u16 sub_handle;u8 irte_mask;enum irq_mode mode; +bool posted_msi; };struct intel_ir_data {
@@ -1118,6 +1120,14 @@ static void prepare_irte(struct irte *irte, int vector, unsigned int dest) irte->redir_hint =1 ; } +static void prepare_irte_posted (struct irte *irte) +{ +memset (irte,0 ,sizeof (*irte)); + + irte->present =1 ; + irte->p_pst =1 ; +} +struct irq_remap_ops intel_irq_remap_ops = { .prepare = intel_prepare_irq_remapping, .enable = intel_enable_irq_remapping,
@@ -1126,6 +1136,47 @@ struct irq_remap_ops intel_irq_remap_ops = { .enable_faulting = enable_drhd_fault_handling, }; +#ifdef CONFIG_X86_POSTED_MSI + +static phys_addr_t get_pi_desc_addr (struct irq_data *irqd) +{ +int cpu =cpumask_first (irq_data_get_effective_affinity_mask (irqd)); + +if (WARN_ON (cpu >= nr_cpu_ids)) +return 0 ; + +return __pa (per_cpu_ptr (&posted_interrupt_desc, cpu)); +} + +static void intel_ir_reconfigure_irte_posted (struct irq_data *irqd) +{ +struct intel_ir_data *ir_data = irqd->chip_data; +struct irte *irte = &ir_data->irte_entry; +struct irte irte_pi; +u64 pid_addr; + + pid_addr =get_pi_desc_addr (irqd); + +if (!pid_addr) { +pr_warn ("Failed to setup IRQ %d for posted mode" , irqd->irq); +return ; + } + +memset (&irte_pi,0 ,sizeof (irte_pi)); + +/* The shared IRTE already be set up as posted during alloc_irte */ +dmar_copy_shared_irte (&irte_pi, irte); + + irte_pi.pda_l = (pid_addr >> (32 -PDA_LOW_BIT )) & ~(-1UL <<PDA_LOW_BIT ); + irte_pi.pda_h = (pid_addr >>32 ) & ~(-1UL <<PDA_HIGH_BIT ); + +modify_irte (&ir_data->irq_2_iommu, &irte_pi); +} + +#else +static inline void intel_ir_reconfigure_irte_posted (struct irq_data *irqd) {} +#endif +static void intel_ir_reconfigure_irte (struct irq_data *irqd,bool force) {struct intel_ir_data *ir_data = irqd->chip_data;
@@ -1139,8 +1190,9 @@ static void intel_ir_reconfigure_irte(struct irq_data *irqd, bool force) irte->vector = cfg->vector; irte->dest_id =IRTE_DEST (cfg->dest_apicid); -/* Update the hardware only if the interrupt is in remapped mode. */ -if (force || ir_data->irq_2_iommu.mode ==IRQ_REMAPPING ) +if (ir_data->irq_2_iommu.posted_msi) +intel_ir_reconfigure_irte_posted (irqd); +else if (force || ir_data->irq_2_iommu.mode ==IRQ_REMAPPING )modify_irte (&ir_data->irq_2_iommu, irte); }
@@ -1194,7 +1246,7 @@ static int intel_ir_set_vcpu_affinity(struct irq_data *data, void *info)struct intel_ir_data *ir_data = data->chip_data;struct vcpu_data *vcpu_pi_info = info; -/* stop posting interrupts, back to remapping mode */ +/* stop posting interrupts, back to the default mode */ if (!vcpu_pi_info) {modify_irte (&ir_data->irq_2_iommu, &ir_data->irte_entry); }else {
@@ -1320,6 +1372,11 @@ static void intel_irq_remapping_prepare_irte(struct intel_ir_data *data,break ;case X86_IRQ_ALLOC_TYPE_PCI_MSI :case X86_IRQ_ALLOC_TYPE_PCI_MSIX : +if (posted_msi_supported ()) { +prepare_irte_posted (irte); + data->irq_2_iommu.posted_msi =1 ; + } +set_msi_sid (irte,pci_real_dma_dev (msi_desc_to_pci_dev (info->desc)));break ;
@@ -1407,7 +1464,11 @@ static int intel_irq_remapping_alloc(struct irq_domain *domain, irq_data->hwirq = (index <<16 ) + i; irq_data->chip_data = ird; - irq_data->chip = &intel_ir_chip; +if (posted_msi_supported () && + ((info->type ==X86_IRQ_ALLOC_TYPE_PCI_MSI ) || (info->type ==X86_IRQ_ALLOC_TYPE_PCI_MSIX ))) + irq_data->chip = &intel_ir_chip_post_msi; +else + irq_data->chip = &intel_ir_chip;intel_irq_remapping_prepare_irte (ird, irq_cfg, info, index, i);irq_set_status_flags (virq + i,IRQ_MOVE_PCNTXT ); }
Signature · 1 line
2.25.1