On 2020-04-03 9:38 a.m., Amber Lin wrote:
Thanks Felix. I'll make changes accordingly but please pay attention to my last reply inline.Ignore what I wrote about squishing events. I was thinking the output format.On 2020-04-02 7:51 p.m., Felix Kuehling wrote:Ok, I'll change back to 64 bits. I was thinking to use "data" as group if we have to run out of bits anyway, but with further thinking, we can squish several events into one since it's event "type", not "ID". 64 will be quite sufficient and no need to worry about running out of it.On 2020-04-02 4:46 p.m., Amber Lin wrote:When the compute is malfunctioning or performance drops, the system admin will use SMI (System Management Interface) tool to monitor/diagnostic whatwent wrong. This patch provides an event watch interface for the user space to register events they are interested. After the event is registered, the user can use annoymous file descriptor's poll functionwith wait-time specified to wait for the event to happen. Once the eventhappens, the user can use read() to retrieve information related to the event. VM fault event is done in this patch. v2: - remove UNREGISTER and add event ENABLE/DISABLE - correct kfifo usage - move event message API to kfd_ioctl.h Signed-off-by: Amber Lin <Amber.Lin@xxxxxxx> --- drivers/gpu/drm/amd/amdkfd/Makefile | 3 +- drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c | 2 + drivers/gpu/drm/amd/amdkfd/kfd_chardev.c | 30 ++++ drivers/gpu/drm/amd/amdkfd/kfd_device.c | 1 + drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c | 2 + drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 12 ++drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c | 177 +++++++++++++++++++++++drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h | 31 ++++ include/uapi/linux/kfd_ioctl.h | 30 +++- 9 files changed, 286 insertions(+), 2 deletions(-) create mode 100644 drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c create mode 100644 drivers/gpu/drm/amd/amdkfd/kfd_smi_events.hdiff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefileindex 6147462..cc98b4a 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -53,7 +53,8 @@ AMDKFD_FILES := $(AMDKFD_PATH)/kfd_module.o \ $(AMDKFD_PATH)/kfd_int_process_v9.o \ $(AMDKFD_PATH)/kfd_dbgdev.o \ $(AMDKFD_PATH)/kfd_dbgmgr.o \ - $(AMDKFD_PATH)/kfd_crat.o + $(AMDKFD_PATH)/kfd_crat.o \ + $(AMDKFD_PATH)/kfd_smi_events.o ifneq ($(CONFIG_AMD_IOMMU_V2),) AMDKFD_FILES += $(AMDKFD_PATH)/kfd_iommu.odiff --git a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.cindex 9f59ba9..24b4717 100644 --- a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c +++ b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c @@ -24,6 +24,7 @@ #include "kfd_events.h" #include "cik_int.h" #include "amdgpu_amdkfd.h" +#include "kfd_smi_events.h" static bool cik_event_interrupt_isr(struct kfd_dev *dev, const uint32_t *ih_ring_entry,@@ -107,6 +108,7 @@ static void cik_event_interrupt_wq(struct kfd_dev *dev,ihre->source_id == CIK_INTSRC_GFX_MEM_PROT_FAULT) { struct kfd_vm_fault_info info; + kfd_smi_event_update_vmfault(dev, pasid); kfd_process_vm_fault(dev->dqm, pasid); memset(&info, 0, sizeof(info));diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.cindex f8fa03a..591ac28 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -39,6 +39,7 @@ #include "kfd_device_queue_manager.h" #include "kfd_dbgmgr.h" #include "amdgpu_amdkfd.h" +#include "kfd_smi_events.h" static long kfd_ioctl(struct file *, unsigned int, unsigned long); static int kfd_open(struct inode *, struct file *);@@ -1243,6 +1244,32 @@ static int kfd_ioctl_acquire_vm(struct file *filep, struct kfd_process *p,return ret; } +/* Handle requests for watching SMI events */ +static int kfd_ioctl_smi_events(struct file *filep, + struct kfd_process *p, void *data) +{ + struct kfd_ioctl_smi_events_args *args = data; + struct kfd_dev *dev; + + dev = kfd_device_by_id(args->gpu_id); + if (!dev) + return -EINVAL; + + switch (args->op) { + case KFD_SMI_EVENTS_REGISTER: + /* register the device */ + return kfd_smi_event_register(dev, &args->data); + case KFD_SMI_EVENTS_ENABLE: + /* subscribe events to the device */ + return kfd_smi_event_enable(dev, args->events); + case KFD_SMI_EVENTS_DISABLE: + /* unsubscribe events */ + return kfd_smi_event_disable(dev, args->events); + } + + return -EINVAL; +} + bool kfd_dev_is_large_bar(struct kfd_dev *dev) { struct kfd_local_mem_info mem_info;@@ -1827,6 +1854,9 @@ static const struct amdkfd_ioctl_desc amdkfd_ioctls[] = {AMDKFD_IOCTL_DEF(AMDKFD_IOC_ALLOC_QUEUE_GWS, kfd_ioctl_alloc_queue_gws, 0), + + AMDKFD_IOCTL_DEF(AMDKFD_IOC_SMI_EVENTS, + kfd_ioctl_smi_events, 0), }; #define AMDKFD_CORE_IOCTL_COUNT ARRAY_SIZE(amdkfd_ioctls)diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.cindex 0491ab2..6ac6f31 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -532,6 +532,7 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, kfd->device_info = device_info; kfd->pdev = pdev; kfd->init_complete = false; + kfd->smi_events.registered = false; kfd->kfd2kgd = f2g; atomic_set(&kfd->compute_profile, 0);diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.cindex e05d75e..151e83e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c @@ -24,6 +24,7 @@ #include "kfd_events.h" #include "soc15_int.h" #include "kfd_device_queue_manager.h" +#include "kfd_smi_events.h" static bool event_interrupt_isr_v9(struct kfd_dev *dev, const uint32_t *ih_ring_entry,@@ -117,6 +118,7 @@ static void event_interrupt_wq_v9(struct kfd_dev *dev,info.prot_read = ring_id & 0x10; info.prot_write = ring_id & 0x20; + kfd_smi_event_update_vmfault(dev, pasid); kfd_process_vm_fault(dev->dqm, pasid); kfd_signal_vm_fault_event(dev, pasid, &info); }diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.hindex 43b888b..b10b665 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -223,6 +223,15 @@ struct kfd_vmid_info { uint32_t vmid_num_kfd; }; +struct kfd_smi_events { + /* device is registered to watch events */ + bool registered; + /* events enabled */ + uint32_t events;This should be 64-bit to allow for more future expansion. It doesn't matter as much here, because you can always change the internal header later. But in the ioctl API we can't change it later, so we should define it as 64-bit from the start.
As stated in the beginning, I'll use header+description where description is a string whose size rounds up to power of 4 (sizeof(uint32_t)).I'm not sure how to create mutex for smi_event so I referenced the debugger code.+ DECLARE_KFIFO_PTR(fifo, uint32_t); + wait_queue_head_t wait_queue; +}; + struct kfd_dev { struct kgd_dev *kgd; @@ -309,6 +318,9 @@ struct kfd_dev { /* Global GWS resource shared b/t processes*/ void *gws; + + /* if this device is in SMI events watch */ + struct kfd_smi_events smi_events; }; enum kfd_mempool {diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.cnew file mode 100644 index 0000000..aab4dac --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.c @@ -0,0 +1,177 @@ +/* + * Copyright 2020 Advanced Micro Devices, Inc. + *+ * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions:+ *+ * The above copyright notice and this permission notice shall be included in+ * all copies or substantial portions of the Software. + *+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR+ * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include <linux/poll.h> +#include <linux/wait.h> +#include <linux/anon_inodes.h> +#include <uapi/linux/kfd_ioctl.h> +#include "amdgpu_vm.h" +#include "kfd_priv.h" +#include "kfd_smi_events.h" + +static DEFINE_MUTEX(kfd_smi_mutex); +struct mutex *kfd_get_smi_mutex(void) +{ + return &kfd_smi_mutex; +}Why do you need a function for this. Just reference kfd_smi_mutex directly below. But since the fifo is per device, should there also be one mutex per device?++static __poll_t kfd_smi_ev_poll(struct file *, struct poll_table_struct *); +static ssize_t kfd_smi_ev_read(struct file *, char __user *, size_t, loff_t *);+static int kfd_smi_ev_release(struct inode *, struct file *); + +static uint32_t smi_fifo_len;This could be statically initialized rather than doing it in a function below because this can be calculated at compile time. It's also constant:static const uint32_t smi_fifo_len = 32 * (sizeof(struct kfd_smi_msg_vmfault)/sizeof(uint32_t) + 1);Or you could probably even make is a #define.True. I'll change it.+static const char kfd_smi_name[] = "kfd_smi_ev"; + +static const struct file_operations kfd_smi_ev_fops = { + .owner = THIS_MODULE, + .poll = kfd_smi_ev_poll, + .read = kfd_smi_ev_read, + .release = kfd_smi_ev_release +}; + +static __poll_t kfd_smi_ev_poll(struct file *filep, + struct poll_table_struct *wait) +{ + __poll_t mask; + struct kfd_dev *dev = filep->private_data; + + poll_wait(filep, &dev->smi_events.wait_queue, wait);+ mask = kfifo_is_empty(&dev->smi_events.fifo) ? 0: POLLIN | POLLRDNORM;+ + return mask; +} + +static ssize_t kfd_smi_ev_read(struct file *filep, char __user *user, + size_t size, loff_t *offset) +{ + int ret, copied = 0; + struct kfd_dev *dev = filep->private_data; + + mutex_lock(kfd_get_smi_mutex()); + ret = kfifo_to_user(&dev->smi_events.fifo, user, size, &copied); + mutex_unlock(kfd_get_smi_mutex()); + if (ret || !copied) { + pr_debug("smi-events: fail to send msg (%i) (%i)\n", + ret, copied); + return ret ? ret : -EAGAIN; + } + + return copied; +} + +static int kfd_smi_ev_release(struct inode *inode, struct file *filep) +{ + struct kfd_dev *dev = filep->private_data; + + dev->smi_events.events = 0; + kfifo_free(&dev->smi_events.fifo);This can cause race conditions. You should do this under the fifo mutex. Also set dev->smi_events.registered = false to allow a new registration after this.Good catch. I'll add them.+ + return 0; +} + +void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)dev->kgd; + struct amdgpu_task_info task_info; + struct kfd_smi_msg_vmfault msg; + uint32_t *fifo_in = (uint32_t *)&msg; + uint32_t fifo_in_len = sizeof(msg) / sizeof(uint32_t); + + if (!(dev->smi_events.events & KFD_SMI_EVENT_VMFAULT)) + return;You need to do this while holding the fifo mutex, otherwise the fifo can be destroyed before you access it below.Ok+ + amdgpu_vm_get_task_info(adev, pasid, &task_info); + msg.pid = task_info.pid; + strncpy(msg.task_name, task_info.task_name, 16); + + mutex_lock(kfd_get_smi_mutex()); + if ((fifo_in_len + 1) > + (smi_fifo_len - kfifo_len(&dev->smi_events.fifo))) {You could use kfifo_avail to get the space available in the fifo.Ah. Thanks.+ pr_err("smi_event: no space left for vmfault event\n"); + mutex_unlock(kfd_get_smi_mutex()); + return; + } + /* Always send the event type first to function as a header */ + kfifo_put(&dev->smi_events.fifo, KFD_SMI_EVENT_VMFAULT);+ /* Then the msg as the payload. Event type reveals the payload size. */+ kfifo_in(&dev->smi_events.fifo, fifo_in, fifo_in_len); + mutex_unlock(kfd_get_smi_mutex()); + + wake_up_all(&dev->smi_events.wait_queue); +} + +int kfd_smi_event_disable(struct kfd_dev *dev, uint64_t events) +{ + mutex_lock(kfd_get_smi_mutex()); + dev->smi_events.events &= ~events; + mutex_unlock(kfd_get_smi_mutex()); + + return 0; +} + +int kfd_smi_event_enable(struct kfd_dev *dev, uint64_t events) +{+ /* If the user didn't register SMI events for this device, kfifo is not+ * created to report events. + */ + if (!dev->smi_events.registered) + return -EINVAL; + + mutex_lock(kfd_get_smi_mutex()); + dev->smi_events.events |= events; + mutex_unlock(kfd_get_smi_mutex()); + + return 0; +} + +static void assign_fifo_len(void) +{+ /* When a new event is added into support and this new event msg uses+ * more memory, replace the msg struct here. + */+ uint32_t max_msg = sizeof(struct kfd_smi_msg_vmfault)/sizeof(uint32_t); + /* +1 for the event type in front of event message. up to 32 entries */+ smi_fifo_len = (++max_msg) * 32;See above. This could be statically initialized.+} + +int kfd_smi_event_register(struct kfd_dev *dev, int *fd) +{ + int ret = 0;This function should return failure if the event was already registered. You can only allow one FD to be created per device, because that FD owns the FIFO.This entire function must be under the fifo mutex.OkIt sounds a good idea, but I don't see how it improves the parsing... The user reads the content based on the event type. My test code on the user space is like this:+ + if (!smi_fifo_len) + assign_fifo_len(); ++ ret = kfifo_alloc(&dev->smi_events.fifo, smi_fifo_len, GFP_KERNEL);+ if (ret) { + pr_err("smi_event: fail to allocate kfifo\n"); + return ret; + } + init_waitqueue_head(&dev->smi_events.wait_queue); + dev->smi_events.events = 0; + dev->smi_events.registered = true; + + ret = anon_inode_getfd(kfd_smi_name, &kfd_smi_ev_fops, + (void *)dev, 0); + *fd = ret; + + return ret; +}diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.hnew file mode 100644 index 0000000..e10bb5d --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_smi_events.h @@ -0,0 +1,31 @@ +/* + * Copyright 2020 Advanced Micro Devices, Inc. + *+ * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions:+ *+ * The above copyright notice and this permission notice shall be included in+ * all copies or substantial portions of the Software. + *+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR+ * OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef KFD_SMI_EVENTS_H_INCLUDED +#define KFD_SMI_EVENTS_H_INCLUDED + +int kfd_smi_event_register(struct kfd_dev *dev, int *fd); +int kfd_smi_event_enable(struct kfd_dev *dev, uint64_t events); +int kfd_smi_event_disable(struct kfd_dev *dev, uint64_t events);+void kfd_smi_event_update_vmfault(struct kfd_dev *dev, uint16_t pasid);+ +#endifdiff --git a/include/uapi/linux/kfd_ioctl.h b/include/uapi/linux/kfd_ioctl.hindex 4f66764..dc9309e 100644 --- a/include/uapi/linux/kfd_ioctl.h +++ b/include/uapi/linux/kfd_ioctl.h @@ -442,6 +442,31 @@ struct kfd_ioctl_import_dmabuf_args { __u32 dmabuf_fd; /* to KFD */ }; +/* + * KFD SMI(System Management Interface) events + */ +enum kfd_smi_events_op { + KFD_SMI_EVENTS_REGISTER = 1, + KFD_SMI_EVENTS_ENABLE, + KFD_SMI_EVENTS_DISABLE +}; + +/* Event type (defined by bitmask) */ +#define KFD_SMI_EVENT_VMFAULT 0x00000001Since you can only have 64 events, you only need one byte. You have plenty of space in the header dword to add the message size. That would make it easier to parse the events and allow variable size events in the future. Then you should define the header as a struct. Something like this:struct kfd_smi_msg_header { __u8 type; __u8 pad; __u16 size; };read(fd, &event, sizeof(uint32_t)); //will change uint64_t switch (event) { case event_A: read(fd, &struct_event_A_variable, sizeof(struct event_A)); case event_B: read(fd, &struct_event_B_variable, sizeof(struct event_B));Not exactly like that, such as read() is done in a separate function, but you see the idea.This is what I mean by the event type tells the size itself. If I provide the header you suggested, the user still needs to decide which struct to use.... Unless we don't use struct for the message. I can change the output to become event:description_of_the event where description is a string. The header you suggest will apply to this method pretty well.
+ +struct kfd_ioctl_smi_events_args { + __u32 op; /* to KFD */ + __u32 events; /* to KFD */I thought this should be 64-bit.+ __u32 gpu_id; /* to KFD */ + __u32 data; /* from KFD */ +}; + +/* Message to user space for each event */ +struct kfd_smi_msg_vmfault {If you define the header like suggested above, you could include it here and at the start of all message structures. Might simplify the code a little:struct kfd_smi_msg_header header; Regards, Felix+ uint32_t pid; + char task_name[16]; +}; + /* Register offset inside the remapped mmio page */ enum kfd_mmio_remap { @@ -546,7 +571,10 @@ enum kfd_mmio_remap { #define AMDKFD_IOC_ALLOC_QUEUE_GWS \ AMDKFD_IOWR(0x1E, struct kfd_ioctl_alloc_queue_gws_args) +#define AMDKFD_IOC_SMI_EVENTS \ + AMDKFD_IOWR(0x1F, struct kfd_ioctl_smi_events_args) + #define AMDKFD_COMMAND_START 0x01 -#define AMDKFD_COMMAND_END 0x1F +#define AMDKFD_COMMAND_END 0x20 #endif
_______________________________________________ amd-gfx mailing list amd-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/amd-gfx
