[v2,07/20] util/dsa: Implement DSA device start and stop logic.

Commit Message

Hao Xiang Nov. 14, 2023, 5:40 a.m. UTC

* DSA device open and close.
* DSA group contains multiple DSA devices.
* DSA group configure/start/stop/clean.

Signed-off-by: Hao Xiang <[e-mail]>
Signed-off-by: Bryan Zhang <[e-mail]>
---
 include/qemu/dsa.h |  49 +++++++
 util/dsa.c         | 338 +++++++++++++++++++++++++++++++++++++++++++++
 util/meson.build   |   1 +
 3 files changed, 388 insertions(+)
 create mode 100644 include/qemu/dsa.h
 create mode 100644 util/dsa.c

Comments

Fabiano Rosas Dec. 11, 2023, 9:28 p.m. UTC | #1

Hao Xiang <[e-mail]> writes:

> * DSA device open and close.
> * DSA group contains multiple DSA devices.
> * DSA group configure/start/stop/clean.
>
> Signed-off-by: Hao Xiang <[e-mail]>
> Signed-off-by: Bryan Zhang <[e-mail]>
> ---
>  include/qemu/dsa.h |  49 +++++++
>  util/dsa.c         | 338 +++++++++++++++++++++++++++++++++++++++++++++
>  util/meson.build   |   1 +
>  3 files changed, 388 insertions(+)
>  create mode 100644 include/qemu/dsa.h
>  create mode 100644 util/dsa.c
>
> diff --git a/include/qemu/dsa.h b/include/qemu/dsa.h
> new file mode 100644
> index 0000000000..30246b507e
> --- /dev/null
> +++ b/include/qemu/dsa.h
> @@ -0,0 +1,49 @@
> +#ifndef QEMU_DSA_H
> +#define QEMU_DSA_H
> +
> +#include "qemu/thread.h"
> +#include "qemu/queue.h"
> +
> +#ifdef CONFIG_DSA_OPT
> +
> +#pragma GCC push_options
> +#pragma GCC target("enqcmd")
> +
> +#include <linux/idxd.h>
> +#include "x86intrin.h"
> +
> +#endif
> +
> +/**
> + * @brief Initializes DSA devices.
> + *
> + * @param dsa_parameter A list of DSA device path from migration parameter.

This code seems pretty generic, let's decouple this doc from migration.

> + * @return int Zero if successful, otherwise non zero.
> + */
> +int dsa_init(const char *dsa_parameter);
> +
> +/**
> + * @brief Start logic to enable using DSA.
> + */
> +void dsa_start(void);
> +
> +/**
> + * @brief Stop logic to clean up DSA by halting the device group and cleaning up
> + * the completion thread.

"Stop the device group and the completion thread"

The mention of "clean/cleaning up" makes this confusing because of
dsa_cleanup() below.

> + */
> +void dsa_stop(void);
> +
> +/**
> + * @brief Clean up system resources created for DSA offloading.
> + *        This function is called during QEMU process teardown.

This is not called during QEMU process teardown. It's called at the end
of migration AFAICS. Maybe just leave this sentence out.

> + */
> +void dsa_cleanup(void);
> +
> +/**
> + * @brief Check if DSA is running.
> + *
> + * @return True if DSA is running, otherwise false.
> + */
> +bool dsa_is_running(void);
> +
> +#endif
> \ No newline at end of file
> diff --git a/util/dsa.c b/util/dsa.c
> new file mode 100644
> index 0000000000..8edaa892ec
> --- /dev/null
> +++ b/util/dsa.c
> @@ -0,0 +1,338 @@
> +/*
> + * Use Intel Data Streaming Accelerator to offload certain background
> + * operations.
> + *
> + * Copyright (c) 2023 Hao Xiang <[e-mail]>
> + *                    Bryan Zhang <[e-mail]>
> + *
> + * 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 AUTHORS OR COPYRIGHT HOLDERS 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 "qemu/osdep.h"
> +#include "qemu/queue.h"
> +#include "qemu/memalign.h"
> +#include "qemu/lockable.h"
> +#include "qemu/cutils.h"
> +#include "qemu/dsa.h"
> +#include "qemu/bswap.h"
> +#include "qemu/error-report.h"
> +#include "qemu/rcu.h"
> +
> +#ifdef CONFIG_DSA_OPT
> +
> +#pragma GCC push_options
> +#pragma GCC target("enqcmd")
> +
> +#include <linux/idxd.h>
> +#include "x86intrin.h"
> +
> +#define DSA_WQ_SIZE 4096
> +#define MAX_DSA_DEVICES 16
> +
> +typedef QSIMPLEQ_HEAD(dsa_task_queue, buffer_zero_batch_task) dsa_task_queue;
> +
> +struct dsa_device {
> +    void *work_queue;
> +};
> +
> +struct dsa_device_group {
> +    struct dsa_device *dsa_devices;
> +    int num_dsa_devices;
> +    uint32_t index;
> +    bool running;
> +    QemuMutex task_queue_lock;
> +    QemuCond task_queue_cond;
> +    dsa_task_queue task_queue;
> +};
> +
> +uint64_t max_retry_count;
> +static struct dsa_device_group dsa_group;
> +
> +
> +/**
> + * @brief This function opens a DSA device's work queue and
> + *        maps the DSA device memory into the current process.
> + *
> + * @param dsa_wq_path A pointer to the DSA device work queue's file path.
> + * @return A pointer to the mapped memory.
> + */
> +static void *
> +map_dsa_device(const char *dsa_wq_path)
> +{
> +    void *dsa_device;
> +    int fd;
> +
> +    fd = open(dsa_wq_path, O_RDWR);
> +    if (fd < 0) {
> +        fprintf(stderr, "open %s failed with errno = %d.\n",
> +                dsa_wq_path, errno);

Use error_report and error_setg* for these. Throughout the series.

> +        return MAP_FAILED;
> +    }
> +    dsa_device = mmap(NULL, DSA_WQ_SIZE, PROT_WRITE,
> +                      MAP_SHARED | MAP_POPULATE, fd, 0);
> +    close(fd);
> +    if (dsa_device == MAP_FAILED) {
> +        fprintf(stderr, "mmap failed with errno = %d.\n", errno);
> +        return MAP_FAILED;
> +    }
> +    return dsa_device;
> +}
> +
> +/**
> + * @brief Initializes a DSA device structure.
> + *
> + * @param instance A pointer to the DSA device.
> + * @param work_queue  A pointer to the DSA work queue.
> + */
> +static void
> +dsa_device_init(struct dsa_device *instance,
> +                void *dsa_work_queue)
> +{
> +    instance->work_queue = dsa_work_queue;
> +}
> +
> +/**
> + * @brief Cleans up a DSA device structure.
> + *
> + * @param instance A pointer to the DSA device to cleanup.
> + */
> +static void
> +dsa_device_cleanup(struct dsa_device *instance)
> +{
> +    if (instance->work_queue != MAP_FAILED) {
> +        munmap(instance->work_queue, DSA_WQ_SIZE);
> +    }
> +}
> +
> +/**
> + * @brief Initializes a DSA device group.
> + *
> + * @param group A pointer to the DSA device group.
> + * @param num_dsa_devices The number of DSA devices this group will have.
> + *
> + * @return Zero if successful, non-zero otherwise.
> + */
> +static int
> +dsa_device_group_init(struct dsa_device_group *group,
> +                      const char *dsa_parameter)

The documentation doesn't match the signature. This happens in other
places as well, please review all of them.

> +{
> +    if (dsa_parameter == NULL || strlen(dsa_parameter) == 0) {
> +        return 0;
> +    }
> +
> +    int ret = 0;
> +    char *local_dsa_parameter = g_strdup(dsa_parameter);
> +    const char *dsa_path[MAX_DSA_DEVICES];
> +    int num_dsa_devices = 0;
> +    char delim[2] = " ";

So we're using space separated strings. Let's document this in this file
and also on the migration parameter documentation.

> +
> +    char *current_dsa_path = strtok(local_dsa_parameter, delim);
> +
> +    while (current_dsa_path != NULL) {
> +        dsa_path[num_dsa_devices++] = current_dsa_path;
> +        if (num_dsa_devices == MAX_DSA_DEVICES) {
> +            break;
> +        }
> +        current_dsa_path = strtok(NULL, delim);
> +    }
> +
> +    group->dsa_devices =
> +        malloc(sizeof(struct dsa_device) * num_dsa_devices);

Use g_new0() here.

> +    group->num_dsa_devices = num_dsa_devices;
> +    group->index = 0;
> +
> +    group->running = false;
> +    qemu_mutex_init(&group->task_queue_lock);
> +    qemu_cond_init(&group->task_queue_cond);
> +    QSIMPLEQ_INIT(&group->task_queue);
> +
> +    void *dsa_wq = MAP_FAILED;
> +    for (int i = 0; i < num_dsa_devices; i++) {
> +        dsa_wq = map_dsa_device(dsa_path[i]);
> +        if (dsa_wq == MAP_FAILED) {
> +            fprintf(stderr, "map_dsa_device failed MAP_FAILED, "
> +                    "using simulation.\n");

What does "using simulation" means? And how are doing it by returning -1
from this function?

> +            ret = -1;

What about the memory for group->dsa_devices in the failure case? We
should either free it here or make sure the client code calls the
cleanup routines.

> +            goto exit;
> +        }
> +        dsa_device_init(&dsa_group.dsa_devices[i], dsa_wq);
> +    }
> +
> +exit:
> +    g_free(local_dsa_parameter);
> +    return ret;
> +}
> +
> +/**
> + * @brief Starts a DSA device group.
> + *
> + * @param group A pointer to the DSA device group.
> + * @param dsa_path An array of DSA device path.
> + * @param num_dsa_devices The number of DSA devices in the device group.
> + */
> +static void
> +dsa_device_group_start(struct dsa_device_group *group)
> +{
> +    group->running = true;
> +}
> +
> +/**
> + * @brief Stops a DSA device group.
> + *
> + * @param group A pointer to the DSA device group.
> + */
> +__attribute__((unused))
> +static void
> +dsa_device_group_stop(struct dsa_device_group *group)
> +{
> +    group->running = false;
> +}
> +
> +/**
> + * @brief Cleans up a DSA device group.
> + *
> + * @param group A pointer to the DSA device group.
> + */
> +static void
> +dsa_device_group_cleanup(struct dsa_device_group *group)
> +{
> +    if (!group->dsa_devices) {
> +        return;
> +    }
> +    for (int i = 0; i < group->num_dsa_devices; i++) {
> +        dsa_device_cleanup(&group->dsa_devices[i]);
> +    }
> +    free(group->dsa_devices);
> +    group->dsa_devices = NULL;
> +
> +    qemu_mutex_destroy(&group->task_queue_lock);
> +    qemu_cond_destroy(&group->task_queue_cond);
> +}
> +
> +/**
> + * @brief Returns the next available DSA device in the group.
> + *
> + * @param group A pointer to the DSA device group.
> + *
> + * @return struct dsa_device* A pointer to the next available DSA device
> + *         in the group.
> + */
> +__attribute__((unused))
> +static struct dsa_device *
> +dsa_device_group_get_next_device(struct dsa_device_group *group)
> +{
> +    if (group->num_dsa_devices == 0) {
> +        return NULL;
> +    }
> +    uint32_t current = qatomic_fetch_inc(&group->index);

The name "index" alone feels a bit opaque. Is there a more
representative name we could give it?

> +    current %= group->num_dsa_devices;
> +    return &group->dsa_devices[current];
> +}
> +
> +/**
> + * @brief Check if DSA is running.
> + *
> + * @return True if DSA is running, otherwise false.
> + */
> +bool dsa_is_running(void)
> +{
> +    return false;
> +}
> +
> +static void
> +dsa_globals_init(void)
> +{
> +    max_retry_count = UINT64_MAX;
> +}
> +
> +/**
> + * @brief Initializes DSA devices.
> + *
> + * @param dsa_parameter A list of DSA device path from migration parameter.
> + * @return int Zero if successful, otherwise non zero.
> + */
> +int dsa_init(const char *dsa_parameter)
> +{
> +    dsa_globals_init();
> +
> +    return dsa_device_group_init(&dsa_group, dsa_parameter);
> +}
> +
> +/**
> + * @brief Start logic to enable using DSA.
> + *
> + */
> +void dsa_start(void)
> +{
> +    if (dsa_group.num_dsa_devices == 0) {
> +        return;
> +    }
> +    if (dsa_group.running) {
> +        return;
> +    }
> +    dsa_device_group_start(&dsa_group);
> +}
> +
> +/**
> + * @brief Stop logic to clean up DSA by halting the device group and cleaning up
> + * the completion thread.
> + *
> + */
> +void dsa_stop(void)
> +{
> +    struct dsa_device_group *group = &dsa_group;
> +
> +    if (!group->running) {
> +        return;
> +    }
> +}
> +
> +/**
> + * @brief Clean up system resources created for DSA offloading.
> + *        This function is called during QEMU process teardown.
> + *
> + */
> +void dsa_cleanup(void)
> +{
> +    dsa_stop();
> +    dsa_device_group_cleanup(&dsa_group);
> +}
> +
> +#else
> +
> +bool dsa_is_running(void)
> +{
> +    return false;
> +}
> +
> +int dsa_init(const char *dsa_parameter)
> +{
> +    fprintf(stderr, "Intel Data Streaming Accelerator is not supported "
> +                    "on this platform.\n");
> +    return -1;

Nothing checks this later in the series and we end up trying to start a
migration when we shouldn't. Fixing the configure step would already
stop this happening, but make sure you check this anyway and abort the
migration.

> +}
> +
> +void dsa_start(void) {}
> +
> +void dsa_stop(void) {}
> +
> +void dsa_cleanup(void) {}
> +
> +#endif

These could all be in the header.

> +
> diff --git a/util/meson.build b/util/meson.build
> index c2322ef6e7..f7277c5e9b 100644
> --- a/util/meson.build
> +++ b/util/meson.build
> @@ -85,6 +85,7 @@ if have_block or have_ga
>  endif
>  if have_block
>    util_ss.add(files('aio-wait.c'))
> +  util_ss.add(files('dsa.c'))

I find it clearer to add the file conditionally under CONFIG_DSA_OPT
here and remove the ifdef from the C file. I'm not sure if we have any
guidelines for this, so up to you.

>    util_ss.add(files('buffer.c'))
>    util_ss.add(files('bufferiszero.c'))
>    util_ss.add(files('hbitmap.c'))

Hao Xiang Dec. 19, 2023, 6:41 a.m. UTC | #2

On Mon, Dec 11, 2023 at 1:28 PM Fabiano Rosas <[e-mail]> wrote:
>
> Hao Xiang <[e-mail]> writes:
>
> > * DSA device open and close.
> > * DSA group contains multiple DSA devices.
> > * DSA group configure/start/stop/clean.
> >
> > Signed-off-by: Hao Xiang <[e-mail]>
> > Signed-off-by: Bryan Zhang <[e-mail]>
> > ---
> >  include/qemu/dsa.h |  49 +++++++
> >  util/dsa.c         | 338 +++++++++++++++++++++++++++++++++++++++++++++
> >  util/meson.build   |   1 +
> >  3 files changed, 388 insertions(+)
> >  create mode 100644 include/qemu/dsa.h
> >  create mode 100644 util/dsa.c
> >
> > diff --git a/include/qemu/dsa.h b/include/qemu/dsa.h
> > new file mode 100644
> > index 0000000000..30246b507e
> > --- /dev/null
> > +++ b/include/qemu/dsa.h
> > @@ -0,0 +1,49 @@
> > +#ifndef QEMU_DSA_H
> > +#define QEMU_DSA_H
> > +
> > +#include "qemu/thread.h"
> > +#include "qemu/queue.h"
> > +
> > +#ifdef CONFIG_DSA_OPT
> > +
> > +#pragma GCC push_options
> > +#pragma GCC target("enqcmd")
> > +
> > +#include <linux/idxd.h>
> > +#include "x86intrin.h"
> > +
> > +#endif
> > +
> > +/**
> > + * @brief Initializes DSA devices.
> > + *
> > + * @param dsa_parameter A list of DSA device path from migration parameter.
>
> This code seems pretty generic, let's decouple this doc from migration.
>
> > + * @return int Zero if successful, otherwise non zero.
> > + */
> > +int dsa_init(const char *dsa_parameter);
> > +
> > +/**
> > + * @brief Start logic to enable using DSA.
> > + */
> > +void dsa_start(void);
> > +
> > +/**
> > + * @brief Stop logic to clean up DSA by halting the device group and cleaning up
> > + * the completion thread.
>
> "Stop the device group and the completion thread"
>
> The mention of "clean/cleaning up" makes this confusing because of
> dsa_cleanup() below.

Fixed.

>
> > + */
> > +void dsa_stop(void);
> > +
> > +/**
> > + * @brief Clean up system resources created for DSA offloading.
> > + *        This function is called during QEMU process teardown.
>
> This is not called during QEMU process teardown. It's called at the end
> of migration AFAICS. Maybe just leave this sentence out.

Fixed.

>
> > + */
> > +void dsa_cleanup(void);
> > +
> > +/**
> > + * @brief Check if DSA is running.
> > + *
> > + * @return True if DSA is running, otherwise false.
> > + */
> > +bool dsa_is_running(void);
> > +
> > +#endif
> > \ No newline at end of file
> > diff --git a/util/dsa.c b/util/dsa.c
> > new file mode 100644
> > index 0000000000..8edaa892ec
> > --- /dev/null
> > +++ b/util/dsa.c
> > @@ -0,0 +1,338 @@
> > +/*
> > + * Use Intel Data Streaming Accelerator to offload certain background
> > + * operations.
> > + *
> > + * Copyright (c) 2023 Hao Xiang <[e-mail]>
> > + *                    Bryan Zhang <[e-mail]>
> > + *
> > + * 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 AUTHORS OR COPYRIGHT HOLDERS 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 "qemu/osdep.h"
> > +#include "qemu/queue.h"
> > +#include "qemu/memalign.h"
> > +#include "qemu/lockable.h"
> > +#include "qemu/cutils.h"
> > +#include "qemu/dsa.h"
> > +#include "qemu/bswap.h"
> > +#include "qemu/error-report.h"
> > +#include "qemu/rcu.h"
> > +
> > +#ifdef CONFIG_DSA_OPT
> > +
> > +#pragma GCC push_options
> > +#pragma GCC target("enqcmd")
> > +
> > +#include <linux/idxd.h>
> > +#include "x86intrin.h"
> > +
> > +#define DSA_WQ_SIZE 4096
> > +#define MAX_DSA_DEVICES 16
> > +
> > +typedef QSIMPLEQ_HEAD(dsa_task_queue, buffer_zero_batch_task) dsa_task_queue;
> > +
> > +struct dsa_device {
> > +    void *work_queue;
> > +};
> > +
> > +struct dsa_device_group {
> > +    struct dsa_device *dsa_devices;
> > +    int num_dsa_devices;
> > +    uint32_t index;
> > +    bool running;
> > +    QemuMutex task_queue_lock;
> > +    QemuCond task_queue_cond;
> > +    dsa_task_queue task_queue;
> > +};
> > +
> > +uint64_t max_retry_count;
> > +static struct dsa_device_group dsa_group;
> > +
> > +
> > +/**
> > + * @brief This function opens a DSA device's work queue and
> > + *        maps the DSA device memory into the current process.
> > + *
> > + * @param dsa_wq_path A pointer to the DSA device work queue's file path.
> > + * @return A pointer to the mapped memory.
> > + */
> > +static void *
> > +map_dsa_device(const char *dsa_wq_path)
> > +{
> > +    void *dsa_device;
> > +    int fd;
> > +
> > +    fd = open(dsa_wq_path, O_RDWR);
> > +    if (fd < 0) {
> > +        fprintf(stderr, "open %s failed with errno = %d.\n",
> > +                dsa_wq_path, errno);
>
> Use error_report and error_setg* for these. Throughout the series.

All converted to using error_report.

>
> > +        return MAP_FAILED;
> > +    }
> > +    dsa_device = mmap(NULL, DSA_WQ_SIZE, PROT_WRITE,
> > +                      MAP_SHARED | MAP_POPULATE, fd, 0);
> > +    close(fd);
> > +    if (dsa_device == MAP_FAILED) {
> > +        fprintf(stderr, "mmap failed with errno = %d.\n", errno);
> > +        return MAP_FAILED;
> > +    }
> > +    return dsa_device;
> > +}
> > +
> > +/**
> > + * @brief Initializes a DSA device structure.
> > + *
> > + * @param instance A pointer to the DSA device.
> > + * @param work_queue  A pointer to the DSA work queue.
> > + */
> > +static void
> > +dsa_device_init(struct dsa_device *instance,
> > +                void *dsa_work_queue)
> > +{
> > +    instance->work_queue = dsa_work_queue;
> > +}
> > +
> > +/**
> > + * @brief Cleans up a DSA device structure.
> > + *
> > + * @param instance A pointer to the DSA device to cleanup.
> > + */
> > +static void
> > +dsa_device_cleanup(struct dsa_device *instance)
> > +{
> > +    if (instance->work_queue != MAP_FAILED) {
> > +        munmap(instance->work_queue, DSA_WQ_SIZE);
> > +    }
> > +}
> > +
> > +/**
> > + * @brief Initializes a DSA device group.
> > + *
> > + * @param group A pointer to the DSA device group.
> > + * @param num_dsa_devices The number of DSA devices this group will have.
> > + *
> > + * @return Zero if successful, non-zero otherwise.
> > + */
> > +static int
> > +dsa_device_group_init(struct dsa_device_group *group,
> > +                      const char *dsa_parameter)
>
> The documentation doesn't match the signature. This happens in other
> places as well, please review all of them.
>
Fixed all cases.

> > +{
> > +    if (dsa_parameter == NULL || strlen(dsa_parameter) == 0) {
> > +        return 0;
> > +    }
> > +
> > +    int ret = 0;
> > +    char *local_dsa_parameter = g_strdup(dsa_parameter);
> > +    const char *dsa_path[MAX_DSA_DEVICES];
> > +    int num_dsa_devices = 0;
> > +    char delim[2] = " ";
>
> So we're using space separated strings. Let's document this in this file
> and also on the migration parameter documentation.

Fixed.

>
> > +
> > +    char *current_dsa_path = strtok(local_dsa_parameter, delim);
> > +
> > +    while (current_dsa_path != NULL) {
> > +        dsa_path[num_dsa_devices++] = current_dsa_path;
> > +        if (num_dsa_devices == MAX_DSA_DEVICES) {
> > +            break;
> > +        }
> > +        current_dsa_path = strtok(NULL, delim);
> > +    }
> > +
> > +    group->dsa_devices =
> > +        malloc(sizeof(struct dsa_device) * num_dsa_devices);
>
> Use g_new0() here.

Converted to use g_new0 and g_free accordingly.

>
> > +    group->num_dsa_devices = num_dsa_devices;
> > +    group->index = 0;
> > +
> > +    group->running = false;
> > +    qemu_mutex_init(&group->task_queue_lock);
> > +    qemu_cond_init(&group->task_queue_cond);
> > +    QSIMPLEQ_INIT(&group->task_queue);
> > +
> > +    void *dsa_wq = MAP_FAILED;
> > +    for (int i = 0; i < num_dsa_devices; i++) {
> > +        dsa_wq = map_dsa_device(dsa_path[i]);
> > +        if (dsa_wq == MAP_FAILED) {
> > +            fprintf(stderr, "map_dsa_device failed MAP_FAILED, "
> > +                    "using simulation.\n");
>
> What does "using simulation" means? And how are doing it by returning -1
> from this function?

* "using simulation" was a copy and paste mistake. Removed that.
* -1 is an error code and will be propagated from
dsa_device_group_init to dsa_init and eventually to
multifd_load_setup/multifd_save_setup.
multifd_load_setup/multifd_save_setup now checks the return code from
dsa_init and aborts the migration if dsa_init fails.

>
> > +            ret = -1;
>
> What about the memory for group->dsa_devices in the failure case? We
> should either free it here or make sure the client code calls the
> cleanup routines.

In the failure case, dsa_device_group_cleanup will free the
group->dsa_devices memory allocation. dsa_device_group_cleanup is
called by dsa_cleanup. multifd_load_cleanup/multifd_save_cleanup will
call the cleanup routines.

>
> > +            goto exit;
> > +        }
> > +        dsa_device_init(&dsa_group.dsa_devices[i], dsa_wq);
> > +    }
> > +
> > +exit:
> > +    g_free(local_dsa_parameter);
> > +    return ret;
> > +}
> > +
> > +/**
> > + * @brief Starts a DSA device group.
> > + *
> > + * @param group A pointer to the DSA device group.
> > + * @param dsa_path An array of DSA device path.
> > + * @param num_dsa_devices The number of DSA devices in the device group.
> > + */
> > +static void
> > +dsa_device_group_start(struct dsa_device_group *group)
> > +{
> > +    group->running = true;
> > +}
> > +
> > +/**
> > + * @brief Stops a DSA device group.
> > + *
> > + * @param group A pointer to the DSA device group.
> > + */
> > +__attribute__((unused))
> > +static void
> > +dsa_device_group_stop(struct dsa_device_group *group)
> > +{
> > +    group->running = false;
> > +}
> > +
> > +/**
> > + * @brief Cleans up a DSA device group.
> > + *
> > + * @param group A pointer to the DSA device group.
> > + */
> > +static void
> > +dsa_device_group_cleanup(struct dsa_device_group *group)
> > +{
> > +    if (!group->dsa_devices) {
> > +        return;
> > +    }
> > +    for (int i = 0; i < group->num_dsa_devices; i++) {
> > +        dsa_device_cleanup(&group->dsa_devices[i]);
> > +    }
> > +    free(group->dsa_devices);
> > +    group->dsa_devices = NULL;
> > +
> > +    qemu_mutex_destroy(&group->task_queue_lock);
> > +    qemu_cond_destroy(&group->task_queue_cond);
> > +}
> > +
> > +/**
> > + * @brief Returns the next available DSA device in the group.
> > + *
> > + * @param group A pointer to the DSA device group.
> > + *
> > + * @return struct dsa_device* A pointer to the next available DSA device
> > + *         in the group.
> > + */
> > +__attribute__((unused))
> > +static struct dsa_device *
> > +dsa_device_group_get_next_device(struct dsa_device_group *group)
> > +{
> > +    if (group->num_dsa_devices == 0) {
> > +        return NULL;
> > +    }
> > +    uint32_t current = qatomic_fetch_inc(&group->index);
>
> The name "index" alone feels a bit opaque. Is there a more
> representative name we could give it?

I renamed it to device_allocator_index and added a comment to explain the field.

>
> > +    current %= group->num_dsa_devices;
> > +    return &group->dsa_devices[current];
> > +}
> > +
> > +/**
> > + * @brief Check if DSA is running.
> > + *
> > + * @return True if DSA is running, otherwise false.
> > + */
> > +bool dsa_is_running(void)
> > +{
> > +    return false;
> > +}
> > +
> > +static void
> > +dsa_globals_init(void)
> > +{
> > +    max_retry_count = UINT64_MAX;
> > +}
> > +
> > +/**
> > + * @brief Initializes DSA devices.
> > + *
> > + * @param dsa_parameter A list of DSA device path from migration parameter.
> > + * @return int Zero if successful, otherwise non zero.
> > + */
> > +int dsa_init(const char *dsa_parameter)
> > +{
> > +    dsa_globals_init();
> > +
> > +    return dsa_device_group_init(&dsa_group, dsa_parameter);
> > +}
> > +
> > +/**
> > + * @brief Start logic to enable using DSA.
> > + *
> > + */
> > +void dsa_start(void)
> > +{
> > +    if (dsa_group.num_dsa_devices == 0) {
> > +        return;
> > +    }
> > +    if (dsa_group.running) {
> > +        return;
> > +    }
> > +    dsa_device_group_start(&dsa_group);
> > +}
> > +
> > +/**
> > + * @brief Stop logic to clean up DSA by halting the device group and cleaning up
> > + * the completion thread.
> > + *
> > + */
> > +void dsa_stop(void)
> > +{
> > +    struct dsa_device_group *group = &dsa_group;
> > +
> > +    if (!group->running) {
> > +        return;
> > +    }
> > +}
> > +
> > +/**
> > + * @brief Clean up system resources created for DSA offloading.
> > + *        This function is called during QEMU process teardown.
> > + *
> > + */
> > +void dsa_cleanup(void)
> > +{
> > +    dsa_stop();
> > +    dsa_device_group_cleanup(&dsa_group);
> > +}
> > +
> > +#else
> > +
> > +bool dsa_is_running(void)
> > +{
> > +    return false;
> > +}
> > +
> > +int dsa_init(const char *dsa_parameter)
> > +{
> > +    fprintf(stderr, "Intel Data Streaming Accelerator is not supported "
> > +                    "on this platform.\n");
> > +    return -1;
>
> Nothing checks this later in the series and we end up trying to start a
> migration when we shouldn't. Fixing the configure step would already
> stop this happening, but make sure you check this anyway and abort the
> migration.

multifd_load_setup/multifd_save_setup now checks the return code from
dsa_init and aborts the migration if dsa_init fails. The
non-CONFIG_DSA_OPT version of dsa_init should really just be a no-op.
Changed that.

>
> > +}
> > +
> > +void dsa_start(void) {}
> > +
> > +void dsa_stop(void) {}
> > +
> > +void dsa_cleanup(void) {}
> > +
> > +#endif
>
> These could all be in the header.

The function definitions are already in dsa.h Do you mean moving the
function implementations to the header as well?

>
> > +
> > diff --git a/util/meson.build b/util/meson.build
> > index c2322ef6e7..f7277c5e9b 100644
> > --- a/util/meson.build
> > +++ b/util/meson.build
> > @@ -85,6 +85,7 @@ if have_block or have_ga
> >  endif
> >  if have_block
> >    util_ss.add(files('aio-wait.c'))
> > +  util_ss.add(files('dsa.c'))
>
> I find it clearer to add the file conditionally under CONFIG_DSA_OPT
> here and remove the ifdef from the C file. I'm not sure if we have any
> guidelines for this, so up to you.
>
> >    util_ss.add(files('buffer.c'))
> >    util_ss.add(files('bufferiszero.c'))
> >    util_ss.add(files('hbitmap.c'))

Fabiano Rosas Dec. 19, 2023, 1:18 p.m. UTC | #3

Hao Xiang <[e-mail]> writes:

>>
>> > +}
>> > +
>> > +void dsa_start(void) {}
>> > +
>> > +void dsa_stop(void) {}
>> > +
>> > +void dsa_cleanup(void) {}
>> > +
>> > +#endif
>>
>> These could all be in the header.
>
> The function definitions are already in dsa.h Do you mean moving the
> function implementations to the header as well?
>

I mean the empty !CONFIG_DSA_OPT variants could be in the header as
static inline.

Hao Xiang Dec. 27, 2023, 6 a.m. UTC | #4

On Tue, Dec 19, 2023 at 5:19 AM Fabiano Rosas <[e-mail]> wrote:
>
> Hao Xiang <[e-mail]> writes:
>
> >>
> >> > +}
> >> > +
> >> > +void dsa_start(void) {}
> >> > +
> >> > +void dsa_stop(void) {}
> >> > +
> >> > +void dsa_cleanup(void) {}
> >> > +
> >> > +#endif
> >>
> >> These could all be in the header.
> >
> > The function definitions are already in dsa.h Do you mean moving the
> > function implementations to the header as well?
> >
>
> I mean the empty !CONFIG_DSA_OPT variants could be in the header as
> static inline.
>

Fixed.

Patch

diff --git a/include/qemu/dsa.h b/include/qemu/dsa.h
new file mode 100644
index 0000000000..30246b507e
--- /dev/null
+++ b/include/qemu/dsa.h
@@ -0,0 +1,49 @@ 
+#ifndef QEMU_DSA_H
+#define QEMU_DSA_H
+
+#include "qemu/thread.h"
+#include "qemu/queue.h"
+
+#ifdef CONFIG_DSA_OPT
+
+#pragma GCC push_options
+#pragma GCC target("enqcmd")
+
+#include <linux/idxd.h>
+#include "x86intrin.h"
+
+#endif
+
+/**
+ * @brief Initializes DSA devices.
+ *
+ * @param dsa_parameter A list of DSA device path from migration parameter.
+ * @return int Zero if successful, otherwise non zero.
+ */
+int dsa_init(const char *dsa_parameter);
+
+/**
+ * @brief Start logic to enable using DSA.
+ */
+void dsa_start(void);
+
+/**
+ * @brief Stop logic to clean up DSA by halting the device group and cleaning up
+ * the completion thread.
+ */
+void dsa_stop(void);
+
+/**
+ * @brief Clean up system resources created for DSA offloading.
+ *        This function is called during QEMU process teardown.
+ */
+void dsa_cleanup(void);
+
+/**
+ * @brief Check if DSA is running.
+ *
+ * @return True if DSA is running, otherwise false.
+ */
+bool dsa_is_running(void);
+
+#endif
\ No newline at end of file
diff --git a/util/dsa.c b/util/dsa.c
new file mode 100644
index 0000000000..8edaa892ec
--- /dev/null
+++ b/util/dsa.c
@@ -0,0 +1,338 @@ 
+/*
+ * Use Intel Data Streaming Accelerator to offload certain background
+ * operations.
+ *
+ * Copyright (c) 2023 Hao Xiang <[e-mail]>
+ *                    Bryan Zhang <[e-mail]>
+ *
+ * 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 AUTHORS OR COPYRIGHT HOLDERS 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 "qemu/osdep.h"
+#include "qemu/queue.h"
+#include "qemu/memalign.h"
+#include "qemu/lockable.h"
+#include "qemu/cutils.h"
+#include "qemu/dsa.h"
+#include "qemu/bswap.h"
+#include "qemu/error-report.h"
+#include "qemu/rcu.h"
+
+#ifdef CONFIG_DSA_OPT
+
+#pragma GCC push_options
+#pragma GCC target("enqcmd")
+
+#include <linux/idxd.h>
+#include "x86intrin.h"
+
+#define DSA_WQ_SIZE 4096
+#define MAX_DSA_DEVICES 16
+
+typedef QSIMPLEQ_HEAD(dsa_task_queue, buffer_zero_batch_task) dsa_task_queue;
+
+struct dsa_device {
+    void *work_queue;
+};
+
+struct dsa_device_group {
+    struct dsa_device *dsa_devices;
+    int num_dsa_devices;
+    uint32_t index;
+    bool running;
+    QemuMutex task_queue_lock;
+    QemuCond task_queue_cond;
+    dsa_task_queue task_queue;
+};
+
+uint64_t max_retry_count;
+static struct dsa_device_group dsa_group;
+
+
+/**
+ * @brief This function opens a DSA device's work queue and
+ *        maps the DSA device memory into the current process.
+ *
+ * @param dsa_wq_path A pointer to the DSA device work queue's file path.
+ * @return A pointer to the mapped memory.
+ */
+static void *
+map_dsa_device(const char *dsa_wq_path)
+{
+    void *dsa_device;
+    int fd;
+
+    fd = open(dsa_wq_path, O_RDWR);
+    if (fd < 0) {
+        fprintf(stderr, "open %s failed with errno = %d.\n",
+                dsa_wq_path, errno);
+        return MAP_FAILED;
+    }
+    dsa_device = mmap(NULL, DSA_WQ_SIZE, PROT_WRITE,
+                      MAP_SHARED | MAP_POPULATE, fd, 0);
+    close(fd);
+    if (dsa_device == MAP_FAILED) {
+        fprintf(stderr, "mmap failed with errno = %d.\n", errno);
+        return MAP_FAILED;
+    }
+    return dsa_device;
+}
+
+/**
+ * @brief Initializes a DSA device structure.
+ *
+ * @param instance A pointer to the DSA device.
+ * @param work_queue  A pointer to the DSA work queue.
+ */
+static void
+dsa_device_init(struct dsa_device *instance,
+                void *dsa_work_queue)
+{
+    instance->work_queue = dsa_work_queue;
+}
+
+/**
+ * @brief Cleans up a DSA device structure.
+ *
+ * @param instance A pointer to the DSA device to cleanup.
+ */
+static void
+dsa_device_cleanup(struct dsa_device *instance)
+{
+    if (instance->work_queue != MAP_FAILED) {
+        munmap(instance->work_queue, DSA_WQ_SIZE);
+    }
+}
+
+/**
+ * @brief Initializes a DSA device group.
+ *
+ * @param group A pointer to the DSA device group.
+ * @param num_dsa_devices The number of DSA devices this group will have.
+ *
+ * @return Zero if successful, non-zero otherwise.
+ */
+static int
+dsa_device_group_init(struct dsa_device_group *group,
+                      const char *dsa_parameter)
+{
+    if (dsa_parameter == NULL || strlen(dsa_parameter) == 0) {
+        return 0;
+    }
+
+    int ret = 0;
+    char *local_dsa_parameter = g_strdup(dsa_parameter);
+    const char *dsa_path[MAX_DSA_DEVICES];
+    int num_dsa_devices = 0;
+    char delim[2] = " ";
+
+    char *current_dsa_path = strtok(local_dsa_parameter, delim);
+
+    while (current_dsa_path != NULL) {
+        dsa_path[num_dsa_devices++] = current_dsa_path;
+        if (num_dsa_devices == MAX_DSA_DEVICES) {
+            break;
+        }
+        current_dsa_path = strtok(NULL, delim);
+    }
+
+    group->dsa_devices =
+        malloc(sizeof(struct dsa_device) * num_dsa_devices);
+    group->num_dsa_devices = num_dsa_devices;
+    group->index = 0;
+
+    group->running = false;
+    qemu_mutex_init(&group->task_queue_lock);
+    qemu_cond_init(&group->task_queue_cond);
+    QSIMPLEQ_INIT(&group->task_queue);
+
+    void *dsa_wq = MAP_FAILED;
+    for (int i = 0; i < num_dsa_devices; i++) {
+        dsa_wq = map_dsa_device(dsa_path[i]);
+        if (dsa_wq == MAP_FAILED) {
+            fprintf(stderr, "map_dsa_device failed MAP_FAILED, "
+                    "using simulation.\n");
+            ret = -1;
+            goto exit;
+        }
+        dsa_device_init(&dsa_group.dsa_devices[i], dsa_wq);
+    }
+
+exit:
+    g_free(local_dsa_parameter);
+    return ret;
+}
+
+/**
+ * @brief Starts a DSA device group.
+ *
+ * @param group A pointer to the DSA device group.
+ * @param dsa_path An array of DSA device path.
+ * @param num_dsa_devices The number of DSA devices in the device group.
+ */
+static void
+dsa_device_group_start(struct dsa_device_group *group)
+{
+    group->running = true;
+}
+
+/**
+ * @brief Stops a DSA device group.
+ *
+ * @param group A pointer to the DSA device group.
+ */
+__attribute__((unused))
+static void
+dsa_device_group_stop(struct dsa_device_group *group)
+{
+    group->running = false;
+}
+
+/**
+ * @brief Cleans up a DSA device group.
+ *
+ * @param group A pointer to the DSA device group.
+ */
+static void
+dsa_device_group_cleanup(struct dsa_device_group *group)
+{
+    if (!group->dsa_devices) {
+        return;
+    }
+    for (int i = 0; i < group->num_dsa_devices; i++) {
+        dsa_device_cleanup(&group->dsa_devices[i]);
+    }
+    free(group->dsa_devices);
+    group->dsa_devices = NULL;
+
+    qemu_mutex_destroy(&group->task_queue_lock);
+    qemu_cond_destroy(&group->task_queue_cond);
+}
+
+/**
+ * @brief Returns the next available DSA device in the group.
+ *
+ * @param group A pointer to the DSA device group.
+ *
+ * @return struct dsa_device* A pointer to the next available DSA device
+ *         in the group.
+ */
+__attribute__((unused))
+static struct dsa_device *
+dsa_device_group_get_next_device(struct dsa_device_group *group)
+{
+    if (group->num_dsa_devices == 0) {
+        return NULL;
+    }
+    uint32_t current = qatomic_fetch_inc(&group->index);
+    current %= group->num_dsa_devices;
+    return &group->dsa_devices[current];
+}
+
+/**
+ * @brief Check if DSA is running.
+ *
+ * @return True if DSA is running, otherwise false.
+ */
+bool dsa_is_running(void)
+{
+    return false;
+}
+
+static void
+dsa_globals_init(void)
+{
+    max_retry_count = UINT64_MAX;
+}
+
+/**
+ * @brief Initializes DSA devices.
+ *
+ * @param dsa_parameter A list of DSA device path from migration parameter.
+ * @return int Zero if successful, otherwise non zero.
+ */
+int dsa_init(const char *dsa_parameter)
+{
+    dsa_globals_init();
+
+    return dsa_device_group_init(&dsa_group, dsa_parameter);
+}
+
+/**
+ * @brief Start logic to enable using DSA.
+ *
+ */
+void dsa_start(void)
+{
+    if (dsa_group.num_dsa_devices == 0) {
+        return;
+    }
+    if (dsa_group.running) {
+        return;
+    }
+    dsa_device_group_start(&dsa_group);
+}
+
+/**
+ * @brief Stop logic to clean up DSA by halting the device group and cleaning up
+ * the completion thread.
+ *
+ */
+void dsa_stop(void)
+{
+    struct dsa_device_group *group = &dsa_group;
+
+    if (!group->running) {
+        return;
+    }
+}
+
+/**
+ * @brief Clean up system resources created for DSA offloading.
+ *        This function is called during QEMU process teardown.
+ *
+ */
+void dsa_cleanup(void)
+{
+    dsa_stop();
+    dsa_device_group_cleanup(&dsa_group);
+}
+
+#else
+
+bool dsa_is_running(void)
+{
+    return false;
+}
+
+int dsa_init(const char *dsa_parameter)
+{
+    fprintf(stderr, "Intel Data Streaming Accelerator is not supported "
+                    "on this platform.\n");
+    return -1;
+}
+
+void dsa_start(void) {}
+
+void dsa_stop(void) {}
+
+void dsa_cleanup(void) {}
+
+#endif
+
diff --git a/util/meson.build b/util/meson.build
index c2322ef6e7..f7277c5e9b 100644
--- a/util/meson.build
+++ b/util/meson.build
@@ -85,6 +85,7 @@  if have_block or have_ga
 endif
 if have_block
   util_ss.add(files('aio-wait.c'))
+  util_ss.add(files('dsa.c'))
   util_ss.add(files('buffer.c'))
   util_ss.add(files('bufferiszero.c'))
   util_ss.add(files('hbitmap.c'))