qemu/migration/multifd.c

1698 lines
48 KiB
C
Raw Normal View History

/*
* Multifd common code
*
* Copyright (c) 2019-2020 Red Hat Inc
*
* Authors:
* Juan Quintela <quintela@redhat.com>
*
* This work is licensed under the terms of the GNU GPL, version 2 or later.
* See the COPYING file in the top-level directory.
*/
#include "qemu/osdep.h"
#include "qemu/cutils.h"
#include "qemu/rcu.h"
#include "exec/target_page.h"
#include "sysemu/sysemu.h"
#include "exec/ramblock.h"
#include "qemu/error-report.h"
#include "qapi/error.h"
#include "file.h"
#include "migration.h"
#include "migration-stats.h"
#include "socket.h"
#include "tls.h"
#include "qemu-file.h"
#include "trace.h"
#include "multifd.h"
#include "threadinfo.h"
#include "options.h"
#include "qemu/yank.h"
#include "io/channel-file.h"
#include "io/channel-socket.h"
#include "yank_functions.h"
/* Multiple fd's */
#define MULTIFD_MAGIC 0x11223344U
#define MULTIFD_VERSION 1
typedef struct {
uint32_t magic;
uint32_t version;
unsigned char uuid[16]; /* QemuUUID */
uint8_t id;
uint8_t unused1[7]; /* Reserved for future use */
uint64_t unused2[4]; /* Reserved for future use */
} __attribute__((packed)) MultiFDInit_t;
migration/multifd: Fix MultiFDSendParams.packet_num race As reported correctly by Fabiano [1] (while per Fabiano, it sourced back to Elena's initial report in Oct 2023), MultiFDSendParams.packet_num is buggy to be assigned and stored. Consider two consequent operations of: (1) queue a job into multifd send thread X, then (2) queue another sync request to the same send thread X. Then the MultiFDSendParams.packet_num will be assigned twice, and the first assignment can get lost already. To avoid that, we move the packet_num assignment from p->packet_num into where the thread will fill in the packet. Use atomic operations to protect the field, making sure there's no race. Note that atomic fetch_add() may not be good for scaling purposes, however multifd should be fine as number of threads should normally not go beyond 16 threads. Let's leave that concern for later but fix the issue first. There's also a trick on how to make it always work even on 32 bit hosts for uint64_t packet number. Switching to uintptr_t as of now to simply the case. It will cause packet number to overflow easier on 32 bit, but that shouldn't be a major concern for now as 32 bit systems is not the major audience for any performance concerns like what multifd wants to address. We also need to move multifd_send_state definition upper, so that multifd_send_fill_packet() can reference it. [1] https://lore.kernel.org/r/87o7d1jlu5.fsf@suse.de Reported-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-23-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:56 +03:00
struct {
MultiFDSendParams *params;
/* array of pages to sent */
MultiFDPages_t *pages;
/*
* Global number of generated multifd packets.
*
* Note that we used 'uintptr_t' because it'll naturally support atomic
* operations on both 32bit / 64 bits hosts. It means on 32bit systems
* multifd will overflow the packet_num easier, but that should be
* fine.
*
* Another option is to use QEMU's Stat64 then it'll be 64 bits on all
* hosts, however so far it does not support atomic fetch_add() yet.
* Make it easy for now.
*/
uintptr_t packet_num;
/*
* Synchronization point past which no more channels will be
* created.
*/
QemuSemaphore channels_created;
migration/multifd: Fix MultiFDSendParams.packet_num race As reported correctly by Fabiano [1] (while per Fabiano, it sourced back to Elena's initial report in Oct 2023), MultiFDSendParams.packet_num is buggy to be assigned and stored. Consider two consequent operations of: (1) queue a job into multifd send thread X, then (2) queue another sync request to the same send thread X. Then the MultiFDSendParams.packet_num will be assigned twice, and the first assignment can get lost already. To avoid that, we move the packet_num assignment from p->packet_num into where the thread will fill in the packet. Use atomic operations to protect the field, making sure there's no race. Note that atomic fetch_add() may not be good for scaling purposes, however multifd should be fine as number of threads should normally not go beyond 16 threads. Let's leave that concern for later but fix the issue first. There's also a trick on how to make it always work even on 32 bit hosts for uint64_t packet number. Switching to uintptr_t as of now to simply the case. It will cause packet number to overflow easier on 32 bit, but that shouldn't be a major concern for now as 32 bit systems is not the major audience for any performance concerns like what multifd wants to address. We also need to move multifd_send_state definition upper, so that multifd_send_fill_packet() can reference it. [1] https://lore.kernel.org/r/87o7d1jlu5.fsf@suse.de Reported-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-23-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:56 +03:00
/* send channels ready */
QemuSemaphore channels_ready;
/*
* Have we already run terminate threads. There is a race when it
* happens that we got one error while we are exiting.
* We will use atomic operations. Only valid values are 0 and 1.
*/
int exiting;
/* multifd ops */
MultiFDMethods *ops;
} *multifd_send_state;
struct {
MultiFDRecvParams *params;
MultiFDRecvData *data;
/* number of created threads */
int count;
/*
* This is always posted by the recv threads, the migration thread
* uses it to wait for recv threads to finish assigned tasks.
*/
QemuSemaphore sem_sync;
/* global number of generated multifd packets */
uint64_t packet_num;
int exiting;
/* multifd ops */
MultiFDMethods *ops;
} *multifd_recv_state;
static bool multifd_use_packets(void)
{
return !migrate_mapped_ram();
}
void multifd_send_channel_created(void)
{
qemu_sem_post(&multifd_send_state->channels_created);
}
static void multifd_set_file_bitmap(MultiFDSendParams *p)
{
MultiFDPages_t *pages = p->pages;
assert(pages->block);
for (int i = 0; i < p->pages->normal_num; i++) {
ramblock_set_file_bmap_atomic(pages->block, pages->offset[i], true);
}
for (int i = p->pages->normal_num; i < p->pages->num; i++) {
ramblock_set_file_bmap_atomic(pages->block, pages->offset[i], false);
}
}
/* Multifd without compression */
/**
* nocomp_send_setup: setup send side
*
* @p: Params for the channel that we are using
* @errp: pointer to an error
*/
static int nocomp_send_setup(MultiFDSendParams *p, Error **errp)
{
if (migrate_zero_copy_send()) {
p->write_flags |= QIO_CHANNEL_WRITE_FLAG_ZERO_COPY;
}
if (multifd_use_packets()) {
/* We need one extra place for the packet header */
p->iov = g_new0(struct iovec, p->page_count + 1);
} else {
p->iov = g_new0(struct iovec, p->page_count);
}
return 0;
}
/**
* nocomp_send_cleanup: cleanup send side
*
* For no compression this function does nothing.
*
* @p: Params for the channel that we are using
* @errp: pointer to an error
*/
static void nocomp_send_cleanup(MultiFDSendParams *p, Error **errp)
{
g_free(p->iov);
p->iov = NULL;
return;
}
static void multifd_send_prepare_iovs(MultiFDSendParams *p)
{
MultiFDPages_t *pages = p->pages;
for (int i = 0; i < pages->normal_num; i++) {
p->iov[p->iovs_num].iov_base = pages->block->host + pages->offset[i];
p->iov[p->iovs_num].iov_len = p->page_size;
p->iovs_num++;
}
p->next_packet_size = pages->normal_num * p->page_size;
}
/**
* nocomp_send_prepare: prepare date to be able to send
*
* For no compression we just have to calculate the size of the
* packet.
*
* Returns 0 for success or -1 for error
*
* @p: Params for the channel that we are using
* @errp: pointer to an error
*/
static int nocomp_send_prepare(MultiFDSendParams *p, Error **errp)
{
bool use_zero_copy_send = migrate_zero_copy_send();
int ret;
multifd_send_zero_page_detect(p);
if (!multifd_use_packets()) {
multifd_send_prepare_iovs(p);
multifd_set_file_bitmap(p);
return 0;
}
if (!use_zero_copy_send) {
/*
* Only !zerocopy needs the header in IOV; zerocopy will
* send it separately.
*/
multifd_send_prepare_header(p);
}
multifd_send_prepare_iovs(p);
p->flags |= MULTIFD_FLAG_NOCOMP;
multifd_send_fill_packet(p);
if (use_zero_copy_send) {
/* Send header first, without zerocopy */
ret = qio_channel_write_all(p->c, (void *)p->packet,
p->packet_len, errp);
if (ret != 0) {
return -1;
}
}
return 0;
}
/**
* nocomp_recv_setup: setup receive side
*
* For no compression this function does nothing.
*
* Returns 0 for success or -1 for error
*
* @p: Params for the channel that we are using
* @errp: pointer to an error
*/
static int nocomp_recv_setup(MultiFDRecvParams *p, Error **errp)
{
p->iov = g_new0(struct iovec, p->page_count);
return 0;
}
/**
* nocomp_recv_cleanup: setup receive side
*
* For no compression this function does nothing.
*
* @p: Params for the channel that we are using
*/
static void nocomp_recv_cleanup(MultiFDRecvParams *p)
{
g_free(p->iov);
p->iov = NULL;
}
/**
* nocomp_recv: read the data from the channel
*
* For no compression we just need to read things into the correct place.
*
* Returns 0 for success or -1 for error
*
* @p: Params for the channel that we are using
* @errp: pointer to an error
*/
static int nocomp_recv(MultiFDRecvParams *p, Error **errp)
{
uint32_t flags;
if (!multifd_use_packets()) {
return multifd_file_recv_data(p, errp);
}
flags = p->flags & MULTIFD_FLAG_COMPRESSION_MASK;
if (flags != MULTIFD_FLAG_NOCOMP) {
error_setg(errp, "multifd %u: flags received %x flags expected %x",
p->id, flags, MULTIFD_FLAG_NOCOMP);
return -1;
}
multifd_recv_zero_page_process(p);
if (!p->normal_num) {
return 0;
}
for (int i = 0; i < p->normal_num; i++) {
p->iov[i].iov_base = p->host + p->normal[i];
p->iov[i].iov_len = p->page_size;
ramblock_recv_bitmap_set_offset(p->block, p->normal[i]);
}
return qio_channel_readv_all(p->c, p->iov, p->normal_num, errp);
}
static MultiFDMethods multifd_nocomp_ops = {
.send_setup = nocomp_send_setup,
.send_cleanup = nocomp_send_cleanup,
.send_prepare = nocomp_send_prepare,
.recv_setup = nocomp_recv_setup,
.recv_cleanup = nocomp_recv_cleanup,
.recv = nocomp_recv
};
static MultiFDMethods *multifd_ops[MULTIFD_COMPRESSION__MAX] = {
[MULTIFD_COMPRESSION_NONE] = &multifd_nocomp_ops,
};
void multifd_register_ops(int method, MultiFDMethods *ops)
{
assert(0 < method && method < MULTIFD_COMPRESSION__MAX);
multifd_ops[method] = ops;
}
/* Reset a MultiFDPages_t* object for the next use */
static void multifd_pages_reset(MultiFDPages_t *pages)
{
/*
* We don't need to touch offset[] array, because it will be
* overwritten later when reused.
*/
pages->num = 0;
pages->normal_num = 0;
pages->block = NULL;
}
static int multifd_send_initial_packet(MultiFDSendParams *p, Error **errp)
{
MultiFDInit_t msg = {};
size_t size = sizeof(msg);
int ret;
msg.magic = cpu_to_be32(MULTIFD_MAGIC);
msg.version = cpu_to_be32(MULTIFD_VERSION);
msg.id = p->id;
memcpy(msg.uuid, &qemu_uuid.data, sizeof(msg.uuid));
ret = qio_channel_write_all(p->c, (char *)&msg, size, errp);
if (ret != 0) {
return -1;
}
stat64_add(&mig_stats.multifd_bytes, size);
return 0;
}
static int multifd_recv_initial_packet(QIOChannel *c, Error **errp)
{
MultiFDInit_t msg;
int ret;
ret = qio_channel_read_all(c, (char *)&msg, sizeof(msg), errp);
if (ret != 0) {
return -1;
}
msg.magic = be32_to_cpu(msg.magic);
msg.version = be32_to_cpu(msg.version);
if (msg.magic != MULTIFD_MAGIC) {
error_setg(errp, "multifd: received packet magic %x "
"expected %x", msg.magic, MULTIFD_MAGIC);
return -1;
}
if (msg.version != MULTIFD_VERSION) {
error_setg(errp, "multifd: received packet version %u "
"expected %u", msg.version, MULTIFD_VERSION);
return -1;
}
if (memcmp(msg.uuid, &qemu_uuid, sizeof(qemu_uuid))) {
char *uuid = qemu_uuid_unparse_strdup(&qemu_uuid);
char *msg_uuid = qemu_uuid_unparse_strdup((const QemuUUID *)msg.uuid);
error_setg(errp, "multifd: received uuid '%s' and expected "
"uuid '%s' for channel %hhd", msg_uuid, uuid, msg.id);
g_free(uuid);
g_free(msg_uuid);
return -1;
}
if (msg.id > migrate_multifd_channels()) {
error_setg(errp, "multifd: received channel id %u is greater than "
"number of channels %u", msg.id, migrate_multifd_channels());
return -1;
}
return msg.id;
}
static MultiFDPages_t *multifd_pages_init(uint32_t n)
{
MultiFDPages_t *pages = g_new0(MultiFDPages_t, 1);
pages->allocated = n;
pages->offset = g_new0(ram_addr_t, n);
return pages;
}
static void multifd_pages_clear(MultiFDPages_t *pages)
{
multifd_pages_reset(pages);
pages->allocated = 0;
g_free(pages->offset);
pages->offset = NULL;
g_free(pages);
}
void multifd_send_fill_packet(MultiFDSendParams *p)
{
MultiFDPacket_t *packet = p->packet;
MultiFDPages_t *pages = p->pages;
migration/multifd: Fix MultiFDSendParams.packet_num race As reported correctly by Fabiano [1] (while per Fabiano, it sourced back to Elena's initial report in Oct 2023), MultiFDSendParams.packet_num is buggy to be assigned and stored. Consider two consequent operations of: (1) queue a job into multifd send thread X, then (2) queue another sync request to the same send thread X. Then the MultiFDSendParams.packet_num will be assigned twice, and the first assignment can get lost already. To avoid that, we move the packet_num assignment from p->packet_num into where the thread will fill in the packet. Use atomic operations to protect the field, making sure there's no race. Note that atomic fetch_add() may not be good for scaling purposes, however multifd should be fine as number of threads should normally not go beyond 16 threads. Let's leave that concern for later but fix the issue first. There's also a trick on how to make it always work even on 32 bit hosts for uint64_t packet number. Switching to uintptr_t as of now to simply the case. It will cause packet number to overflow easier on 32 bit, but that shouldn't be a major concern for now as 32 bit systems is not the major audience for any performance concerns like what multifd wants to address. We also need to move multifd_send_state definition upper, so that multifd_send_fill_packet() can reference it. [1] https://lore.kernel.org/r/87o7d1jlu5.fsf@suse.de Reported-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-23-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:56 +03:00
uint64_t packet_num;
uint32_t zero_num = pages->num - pages->normal_num;
int i;
packet->flags = cpu_to_be32(p->flags);
packet->pages_alloc = cpu_to_be32(p->pages->allocated);
packet->normal_pages = cpu_to_be32(pages->normal_num);
packet->zero_pages = cpu_to_be32(zero_num);
packet->next_packet_size = cpu_to_be32(p->next_packet_size);
migration/multifd: Fix MultiFDSendParams.packet_num race As reported correctly by Fabiano [1] (while per Fabiano, it sourced back to Elena's initial report in Oct 2023), MultiFDSendParams.packet_num is buggy to be assigned and stored. Consider two consequent operations of: (1) queue a job into multifd send thread X, then (2) queue another sync request to the same send thread X. Then the MultiFDSendParams.packet_num will be assigned twice, and the first assignment can get lost already. To avoid that, we move the packet_num assignment from p->packet_num into where the thread will fill in the packet. Use atomic operations to protect the field, making sure there's no race. Note that atomic fetch_add() may not be good for scaling purposes, however multifd should be fine as number of threads should normally not go beyond 16 threads. Let's leave that concern for later but fix the issue first. There's also a trick on how to make it always work even on 32 bit hosts for uint64_t packet number. Switching to uintptr_t as of now to simply the case. It will cause packet number to overflow easier on 32 bit, but that shouldn't be a major concern for now as 32 bit systems is not the major audience for any performance concerns like what multifd wants to address. We also need to move multifd_send_state definition upper, so that multifd_send_fill_packet() can reference it. [1] https://lore.kernel.org/r/87o7d1jlu5.fsf@suse.de Reported-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-23-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:56 +03:00
packet_num = qatomic_fetch_inc(&multifd_send_state->packet_num);
packet->packet_num = cpu_to_be64(packet_num);
if (pages->block) {
strncpy(packet->ramblock, pages->block->idstr, 256);
}
for (i = 0; i < pages->num; i++) {
/* there are architectures where ram_addr_t is 32 bit */
uint64_t temp = pages->offset[i];
packet->offset[i] = cpu_to_be64(temp);
}
p->packets_sent++;
p->total_normal_pages += pages->normal_num;
p->total_zero_pages += zero_num;
trace_multifd_send(p->id, packet_num, pages->normal_num, zero_num,
p->flags, p->next_packet_size);
}
static int multifd_recv_unfill_packet(MultiFDRecvParams *p, Error **errp)
{
MultiFDPacket_t *packet = p->packet;
int i;
packet->magic = be32_to_cpu(packet->magic);
if (packet->magic != MULTIFD_MAGIC) {
error_setg(errp, "multifd: received packet "
"magic %x and expected magic %x",
packet->magic, MULTIFD_MAGIC);
return -1;
}
packet->version = be32_to_cpu(packet->version);
if (packet->version != MULTIFD_VERSION) {
error_setg(errp, "multifd: received packet "
"version %u and expected version %u",
packet->version, MULTIFD_VERSION);
return -1;
}
p->flags = be32_to_cpu(packet->flags);
packet->pages_alloc = be32_to_cpu(packet->pages_alloc);
/*
* If we received a packet that is 100 times bigger than expected
* just stop migration. It is a magic number.
*/
if (packet->pages_alloc > p->page_count) {
error_setg(errp, "multifd: received packet "
"with size %u and expected a size of %u",
packet->pages_alloc, p->page_count) ;
return -1;
}
p->normal_num = be32_to_cpu(packet->normal_pages);
if (p->normal_num > packet->pages_alloc) {
error_setg(errp, "multifd: received packet "
"with %u normal pages and expected maximum pages are %u",
p->normal_num, packet->pages_alloc) ;
return -1;
}
p->zero_num = be32_to_cpu(packet->zero_pages);
if (p->zero_num > packet->pages_alloc - p->normal_num) {
error_setg(errp, "multifd: received packet "
"with %u zero pages and expected maximum zero pages are %u",
p->zero_num, packet->pages_alloc - p->normal_num) ;
return -1;
}
p->next_packet_size = be32_to_cpu(packet->next_packet_size);
p->packet_num = be64_to_cpu(packet->packet_num);
p->packets_recved++;
p->total_normal_pages += p->normal_num;
p->total_zero_pages += p->zero_num;
trace_multifd_recv(p->id, p->packet_num, p->normal_num, p->zero_num,
p->flags, p->next_packet_size);
if (p->normal_num == 0 && p->zero_num == 0) {
return 0;
}
/* make sure that ramblock is 0 terminated */
packet->ramblock[255] = 0;
p->block = qemu_ram_block_by_name(packet->ramblock);
if (!p->block) {
error_setg(errp, "multifd: unknown ram block %s",
packet->ramblock);
return -1;
}
p->host = p->block->host;
for (i = 0; i < p->normal_num; i++) {
uint64_t offset = be64_to_cpu(packet->offset[i]);
if (offset > (p->block->used_length - p->page_size)) {
error_setg(errp, "multifd: offset too long %" PRIu64
" (max " RAM_ADDR_FMT ")",
offset, p->block->used_length);
return -1;
}
p->normal[i] = offset;
}
for (i = 0; i < p->zero_num; i++) {
uint64_t offset = be64_to_cpu(packet->offset[p->normal_num + i]);
if (offset > (p->block->used_length - p->page_size)) {
error_setg(errp, "multifd: offset too long %" PRIu64
" (max " RAM_ADDR_FMT ")",
offset, p->block->used_length);
return -1;
}
p->zero[i] = offset;
}
return 0;
}
static bool multifd_send_should_exit(void)
{
return qatomic_read(&multifd_send_state->exiting);
}
static bool multifd_recv_should_exit(void)
{
return qatomic_read(&multifd_recv_state->exiting);
}
/*
* The migration thread can wait on either of the two semaphores. This
* function can be used to kick the main thread out of waiting on either of
* them. Should mostly only be called when something wrong happened with
* the current multifd send thread.
*/
static void multifd_send_kick_main(MultiFDSendParams *p)
{
qemu_sem_post(&p->sem_sync);
qemu_sem_post(&multifd_send_state->channels_ready);
}
/*
* How we use multifd_send_state->pages and channel->pages?
*
* We create a pages for each channel, and a main one. Each time that
* we need to send a batch of pages we interchange the ones between
* multifd_send_state and the channel that is sending it. There are
* two reasons for that:
* - to not have to do so many mallocs during migration
* - to make easier to know what to free at the end of migration
*
* This way we always know who is the owner of each "pages" struct,
* and we don't need any locking. It belongs to the migration thread
* or to the channel thread. Switching is safe because the migration
* thread is using the channel mutex when changing it, and the channel
* have to had finish with its own, otherwise pending_job can't be
* false.
*
* Returns true if succeed, false otherwise.
*/
static bool multifd_send_pages(void)
{
int i;
static int next_channel;
MultiFDSendParams *p = NULL; /* make happy gcc */
MultiFDPages_t *pages = multifd_send_state->pages;
if (multifd_send_should_exit()) {
return false;
}
/* We wait here, until at least one channel is ready */
qemu_sem_wait(&multifd_send_state->channels_ready);
/*
* next_channel can remain from a previous migration that was
* using more channels, so ensure it doesn't overflow if the
* limit is lower now.
*/
next_channel %= migrate_multifd_channels();
for (i = next_channel;; i = (i + 1) % migrate_multifd_channels()) {
if (multifd_send_should_exit()) {
return false;
}
p = &multifd_send_state->params[i];
/*
* Lockless read to p->pending_job is safe, because only multifd
* sender thread can clear it.
*/
migration/multifd: Separate SYNC request with normal jobs Multifd provide a threaded model for processing jobs. On sender side, there can be two kinds of job: (1) a list of pages to send, or (2) a sync request. The sync request is a very special kind of job. It never contains a page array, but only a multifd packet telling the dest side to synchronize with sent pages. Before this patch, both requests use the pending_job field, no matter what the request is, it will boost pending_job, while multifd sender thread will decrement it after it finishes one job. However this should be racy, because SYNC is special in that it needs to set p->flags with MULTIFD_FLAG_SYNC, showing that this is a sync request. Consider a sequence of operations where: - migration thread enqueue a job to send some pages, pending_job++ (0->1) - [...before the selected multifd sender thread wakes up...] - migration thread enqueue another job to sync, pending_job++ (1->2), setup p->flags=MULTIFD_FLAG_SYNC - multifd sender thread wakes up, found pending_job==2 - send the 1st packet with MULTIFD_FLAG_SYNC and list of pages - send the 2nd packet with flags==0 and no pages This is not expected, because MULTIFD_FLAG_SYNC should hopefully be done after all the pages are received. Meanwhile, the 2nd packet will be completely useless, which contains zero information. I didn't verify above, but I think this issue is still benign in that at least on the recv side we always receive pages before handling MULTIFD_FLAG_SYNC. However that's not always guaranteed and just tricky. One other reason I want to separate it is using p->flags to communicate between the two threads is also not clearly defined, it's very hard to read and understand why accessing p->flags is always safe; see the current impl of multifd_send_thread() where we tried to cache only p->flags. It doesn't need to be that complicated. This patch introduces pending_sync, a separate flag just to show that the requester needs a sync. Alongside, we remove the tricky caching of p->flags now because after this patch p->flags should only be used by multifd sender thread now, which will be crystal clear. So it is always thread safe to access p->flags. With that, we can also safely convert the pending_job into a boolean, because we don't support >1 pending jobs anyway. Always use atomic ops to access both flags to make sure no cache effect. When at it, drop the initial setting of "pending_job = 0" because it's always allocated using g_new0(). Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-7-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:40 +03:00
if (qatomic_read(&p->pending_job) == false) {
next_channel = (i + 1) % migrate_multifd_channels();
break;
}
}
/*
migration/multifd: Optimize sender side to be lockless When reviewing my attempt to refactor send_prepare(), Fabiano suggested we try out with dropping the mutex in multifd code [1]. I thought about that before but I never tried to change the code. Now maybe it's time to give it a stab. This only optimizes the sender side. The trick here is multifd has a clear provider/consumer model, that the migration main thread publishes requests (either pending_job/pending_sync), while the multifd sender threads are consumers. Here we don't have a lot of complicated data sharing, and the jobs can logically be submitted lockless. Arm the code with atomic weapons. Two things worth mentioning: - For multifd_send_pages(): we can use qatomic_load_acquire() when trying to find a free channel, but that's expensive if we attach one ACQUIRE per channel. Instead, keep the qatomic_read() on reading the pending_job flag as we do already, meanwhile use one smp_mb_acquire() after the loop to guarantee the memory ordering. - For pending_sync: it doesn't have any extra data required since now p->flags are never touched, it should be safe to not use memory barrier. That's different from pending_job. Provide rich comments for all the lockless operations to state how they are paired. With that, we can remove the mutex. [1] https://lore.kernel.org/r/87o7d1jlu5.fsf@suse.de Suggested-by: Fabiano Rosas <farosas@suse.de> Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-24-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:57 +03:00
* Make sure we read p->pending_job before all the rest. Pairs with
* qatomic_store_release() in multifd_send_thread().
*/
migration/multifd: Optimize sender side to be lockless When reviewing my attempt to refactor send_prepare(), Fabiano suggested we try out with dropping the mutex in multifd code [1]. I thought about that before but I never tried to change the code. Now maybe it's time to give it a stab. This only optimizes the sender side. The trick here is multifd has a clear provider/consumer model, that the migration main thread publishes requests (either pending_job/pending_sync), while the multifd sender threads are consumers. Here we don't have a lot of complicated data sharing, and the jobs can logically be submitted lockless. Arm the code with atomic weapons. Two things worth mentioning: - For multifd_send_pages(): we can use qatomic_load_acquire() when trying to find a free channel, but that's expensive if we attach one ACQUIRE per channel. Instead, keep the qatomic_read() on reading the pending_job flag as we do already, meanwhile use one smp_mb_acquire() after the loop to guarantee the memory ordering. - For pending_sync: it doesn't have any extra data required since now p->flags are never touched, it should be safe to not use memory barrier. That's different from pending_job. Provide rich comments for all the lockless operations to state how they are paired. With that, we can remove the mutex. [1] https://lore.kernel.org/r/87o7d1jlu5.fsf@suse.de Suggested-by: Fabiano Rosas <farosas@suse.de> Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-24-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:57 +03:00
smp_mb_acquire();
assert(!p->pages->num);
multifd_send_state->pages = p->pages;
p->pages = pages;
migration/multifd: Optimize sender side to be lockless When reviewing my attempt to refactor send_prepare(), Fabiano suggested we try out with dropping the mutex in multifd code [1]. I thought about that before but I never tried to change the code. Now maybe it's time to give it a stab. This only optimizes the sender side. The trick here is multifd has a clear provider/consumer model, that the migration main thread publishes requests (either pending_job/pending_sync), while the multifd sender threads are consumers. Here we don't have a lot of complicated data sharing, and the jobs can logically be submitted lockless. Arm the code with atomic weapons. Two things worth mentioning: - For multifd_send_pages(): we can use qatomic_load_acquire() when trying to find a free channel, but that's expensive if we attach one ACQUIRE per channel. Instead, keep the qatomic_read() on reading the pending_job flag as we do already, meanwhile use one smp_mb_acquire() after the loop to guarantee the memory ordering. - For pending_sync: it doesn't have any extra data required since now p->flags are never touched, it should be safe to not use memory barrier. That's different from pending_job. Provide rich comments for all the lockless operations to state how they are paired. With that, we can remove the mutex. [1] https://lore.kernel.org/r/87o7d1jlu5.fsf@suse.de Suggested-by: Fabiano Rosas <farosas@suse.de> Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-24-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:57 +03:00
/*
* Making sure p->pages is setup before marking pending_job=true. Pairs
* with the qatomic_load_acquire() in multifd_send_thread().
*/
qatomic_store_release(&p->pending_job, true);
qemu_sem_post(&p->sem);
return true;
}
static inline bool multifd_queue_empty(MultiFDPages_t *pages)
{
return pages->num == 0;
}
static inline bool multifd_queue_full(MultiFDPages_t *pages)
{
return pages->num == pages->allocated;
}
static inline void multifd_enqueue(MultiFDPages_t *pages, ram_addr_t offset)
{
pages->offset[pages->num++] = offset;
}
/* Returns true if enqueue successful, false otherwise */
bool multifd_queue_page(RAMBlock *block, ram_addr_t offset)
{
MultiFDPages_t *pages;
retry:
pages = multifd_send_state->pages;
/* If the queue is empty, we can already enqueue now */
if (multifd_queue_empty(pages)) {
pages->block = block;
multifd_enqueue(pages, offset);
return true;
}
/*
* Not empty, meanwhile we need a flush. It can because of either:
*
* (1) The page is not on the same ramblock of previous ones, or,
* (2) The queue is full.
*
* After flush, always retry.
*/
if (pages->block != block || multifd_queue_full(pages)) {
if (!multifd_send_pages()) {
return false;
}
goto retry;
}
/* Not empty, and we still have space, do it! */
multifd_enqueue(pages, offset);
return true;
}
/* Multifd send side hit an error; remember it and prepare to quit */
static void multifd_send_set_error(Error *err)
{
/*
* We don't want to exit each threads twice. Depending on where
* we get the error, or if there are two independent errors in two
* threads at the same time, we can end calling this function
* twice.
*/
if (qatomic_xchg(&multifd_send_state->exiting, 1)) {
return;
}
if (err) {
MigrationState *s = migrate_get_current();
migrate_set_error(s, err);
if (s->state == MIGRATION_STATUS_SETUP ||
s->state == MIGRATION_STATUS_PRE_SWITCHOVER ||
s->state == MIGRATION_STATUS_DEVICE ||
s->state == MIGRATION_STATUS_ACTIVE) {
migrate_set_state(&s->state, s->state,
MIGRATION_STATUS_FAILED);
}
}
}
static void multifd_send_terminate_threads(void)
{
int i;
trace_multifd_send_terminate_threads();
/*
* Tell everyone we're quitting. No xchg() needed here; we simply
* always set it.
*/
qatomic_set(&multifd_send_state->exiting, 1);
/*
* Firstly, kick all threads out; no matter whether they are just idle,
* or blocked in an IO system call.
*/
for (i = 0; i < migrate_multifd_channels(); i++) {
MultiFDSendParams *p = &multifd_send_state->params[i];
qemu_sem_post(&p->sem);
multifd: Shut down the QIO channels to avoid blocking the send threads when they are terminated. When doing live migration with multifd channels 8, 16 or larger number, the guest hangs in the presence of the network errors such as missing TCP ACKs. At sender's side: The main thread is blocked on qemu_thread_join, migration_fd_cleanup is called because one thread fails on qio_channel_write_all when the network problem happens and other send threads are blocked on sendmsg. They could not be terminated. So the main thread is blocked on qemu_thread_join to wait for the threads terminated. (gdb) bt 0 0x00007f30c8dcffc0 in __pthread_clockjoin_ex () at /lib64/libpthread.so.0 1 0x000055cbb716084b in qemu_thread_join (thread=0x55cbb881f418) at ../util/qemu-thread-posix.c:627 2 0x000055cbb6b54e40 in multifd_save_cleanup () at ../migration/multifd.c:542 3 0x000055cbb6b4de06 in migrate_fd_cleanup (s=0x55cbb8024000) at ../migration/migration.c:1808 4 0x000055cbb6b4dfb4 in migrate_fd_cleanup_bh (opaque=0x55cbb8024000) at ../migration/migration.c:1850 5 0x000055cbb7173ac1 in aio_bh_call (bh=0x55cbb7eb98e0) at ../util/async.c:141 6 0x000055cbb7173bcb in aio_bh_poll (ctx=0x55cbb7ebba80) at ../util/async.c:169 7 0x000055cbb715ba4b in aio_dispatch (ctx=0x55cbb7ebba80) at ../util/aio-posix.c:381 8 0x000055cbb7173ffe in aio_ctx_dispatch (source=0x55cbb7ebba80, callback=0x0, user_data=0x0) at ../util/async.c:311 9 0x00007f30c9c8cdf4 in g_main_context_dispatch () at /usr/lib64/libglib-2.0.so.0 10 0x000055cbb71851a2 in glib_pollfds_poll () at ../util/main-loop.c:232 11 0x000055cbb718521c in os_host_main_loop_wait (timeout=42251070366) at ../util/main-loop.c:255 12 0x000055cbb7185321 in main_loop_wait (nonblocking=0) at ../util/main-loop.c:531 13 0x000055cbb6e6ba27 in qemu_main_loop () at ../softmmu/runstate.c:726 14 0x000055cbb6ad6fd7 in main (argc=68, argv=0x7ffc0c578888, envp=0x7ffc0c578ab0) at ../softmmu/main.c:50 To make sure that the send threads could be terminated, IO channels should be shut down to avoid waiting IO. Signed-off-by: Li Zhang <lizhang@suse.de> Reviewed-by: Daniel P. Berrangé <berrange@redhat.com> Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com> Reviewed-by: Juan Quintela <quintela@redhat.com> Signed-off-by: Juan Quintela <quintela@redhat.com>
2021-12-03 14:55:33 +03:00
if (p->c) {
qio_channel_shutdown(p->c, QIO_CHANNEL_SHUTDOWN_BOTH, NULL);
}
}
/*
* Finally recycle all the threads.
*/
for (i = 0; i < migrate_multifd_channels(); i++) {
MultiFDSendParams *p = &multifd_send_state->params[i];
if (p->tls_thread_created) {
qemu_thread_join(&p->tls_thread);
}
migration/multifd: Remove p->running We currently only need p->running to avoid calling qemu_thread_join() on a non existent thread if the thread has never been created. However, there are at least two bugs in this logic: 1) On the sending side, p->running is set too early and qemu_thread_create() can be skipped due to an error during TLS handshake, leaving the flag set and leading to a crash when multifd_send_cleanup() calls qemu_thread_join(). 2) During exit, the multifd thread clears the flag while holding the channel lock. The counterpart at multifd_send_cleanup() reads the flag outside of the lock and might free the mutex while the multifd thread still has it locked. Fix the first issue by setting the flag right before creating the thread. Rename it from p->running to p->thread_created to clarify its usage. Fix the second issue by not clearing the flag at the multifd thread exit. We don't have any use for that. Note that these bugs are straight-forward logic issues and not race conditions. There is still a gap for races to affect this code due to multifd_send_cleanup() being allowed to run concurrently with the thread creation loop. This issue is solved in the next patches. Cc: qemu-stable <qemu-stable@nongnu.org> Fixes: 29647140157a ("migration/tls: add support for multifd tls-handshake") Reported-by: Avihai Horon <avihaih@nvidia.com> Reported-by: chenyuhui5@huawei.com Reviewed-by: Peter Xu <peterx@redhat.com> Signed-off-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240206215118.6171-3-farosas@suse.de Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-07 00:51:14 +03:00
if (p->thread_created) {
qemu_thread_join(&p->thread);
}
}
}
static bool multifd_send_cleanup_channel(MultiFDSendParams *p, Error **errp)
{
if (p->c) {
migration_ioc_unregister_yank(p->c);
/*
* The object_unref() cannot guarantee the fd will always be
* released because finalize() of the iochannel is only
* triggered on the last reference and it's not guaranteed
* that we always hold the last refcount when reaching here.
*
* Closing the fd explicitly has the benefit that if there is any
* registered I/O handler callbacks on such fd, that will get a
* POLLNVAL event and will further trigger the cleanup to finally
* release the IOC.
*
* FIXME: It should logically be guaranteed that all multifd
* channels have no I/O handler callback registered when reaching
* here, because migration thread will wait for all multifd channel
* establishments to complete during setup. Since
* migrate_fd_cleanup() will be scheduled in main thread too, all
* previous callbacks should guarantee to be completed when
* reaching here. See multifd_send_state.channels_created and its
* usage. In the future, we could replace this with an assert
* making sure we're the last reference, or simply drop it if above
* is more clear to be justified.
*/
qio_channel_close(p->c, &error_abort);
object_unref(OBJECT(p->c));
p->c = NULL;
}
qemu_sem_destroy(&p->sem);
qemu_sem_destroy(&p->sem_sync);
g_free(p->name);
p->name = NULL;
multifd_pages_clear(p->pages);
p->pages = NULL;
p->packet_len = 0;
g_free(p->packet);
p->packet = NULL;
multifd_send_state->ops->send_cleanup(p, errp);
return *errp == NULL;
}
static void multifd_send_cleanup_state(void)
{
file_cleanup_outgoing_migration();
socket_cleanup_outgoing_migration();
qemu_sem_destroy(&multifd_send_state->channels_created);
qemu_sem_destroy(&multifd_send_state->channels_ready);
g_free(multifd_send_state->params);
multifd_send_state->params = NULL;
multifd_pages_clear(multifd_send_state->pages);
multifd_send_state->pages = NULL;
g_free(multifd_send_state);
multifd_send_state = NULL;
}
void multifd_send_shutdown(void)
{
int i;
if (!migrate_multifd()) {
return;
}
multifd_send_terminate_threads();
for (i = 0; i < migrate_multifd_channels(); i++) {
MultiFDSendParams *p = &multifd_send_state->params[i];
Error *local_err = NULL;
if (!multifd_send_cleanup_channel(p, &local_err)) {
migrate_set_error(migrate_get_current(), local_err);
migration/multifd: Do error_free after migrate_set_error to avoid memleaks When error happen in multifd_send_thread, it use error_copy to set migrate error in multifd_send_terminate_threads(). We should call error_free after it. Similarly, fix another two places in multifd_recv_thread/multifd_save_cleanup. The leak stack: Direct leak of 48 byte(s) in 1 object(s) allocated from: #0 0x7f781af07cf0 in calloc (/lib64/libasan.so.5+0xefcf0) #1 0x7f781a2ce22d in g_malloc0 (/lib64/libglib-2.0.so.0+0x5322d) #2 0x55ee1d075c17 in error_setv /mnt/sdb/backup/qemu/util/error.c:61 #3 0x55ee1d076464 in error_setg_errno_internal /mnt/sdb/backup/qemu/util/error.c:109 #4 0x55ee1cef066e in qio_channel_socket_writev /mnt/sdb/backup/qemu/io/channel-socket.c:569 #5 0x55ee1cee806b in qio_channel_writev /mnt/sdb/backup/qemu/io/channel.c:207 #6 0x55ee1cee806b in qio_channel_writev_all /mnt/sdb/backup/qemu/io/channel.c:171 #7 0x55ee1cee8248 in qio_channel_write_all /mnt/sdb/backup/qemu/io/channel.c:257 #8 0x55ee1ca12c9a in multifd_send_thread /mnt/sdb/backup/qemu/migration/multifd.c:657 #9 0x55ee1d0607fc in qemu_thread_start /mnt/sdb/backup/qemu/util/qemu-thread-posix.c:519 #10 0x7f78159ae2dd in start_thread (/lib64/libpthread.so.0+0x82dd) #11 0x7f78156df4b2 in __GI___clone (/lib64/libc.so.6+0xfc4b2) Indirect leak of 52 byte(s) in 1 object(s) allocated from: #0 0x7f781af07f28 in __interceptor_realloc (/lib64/libasan.so.5+0xeff28) #1 0x7f78156f07d9 in __GI___vasprintf_chk (/lib64/libc.so.6+0x10d7d9) #2 0x7f781a30ea6c in g_vasprintf (/lib64/libglib-2.0.so.0+0x93a6c) #3 0x7f781a2e7cd0 in g_strdup_vprintf (/lib64/libglib-2.0.so.0+0x6ccd0) #4 0x7f781a2e7d8c in g_strdup_printf (/lib64/libglib-2.0.so.0+0x6cd8c) #5 0x55ee1d075c86 in error_setv /mnt/sdb/backup/qemu/util/error.c:65 #6 0x55ee1d076464 in error_setg_errno_internal /mnt/sdb/backup/qemu/util/error.c:109 #7 0x55ee1cef066e in qio_channel_socket_writev /mnt/sdb/backup/qemu/io/channel-socket.c:569 #8 0x55ee1cee806b in qio_channel_writev /mnt/sdb/backup/qemu/io/channel.c:207 #9 0x55ee1cee806b in qio_channel_writev_all /mnt/sdb/backup/qemu/io/channel.c:171 #10 0x55ee1cee8248 in qio_channel_write_all /mnt/sdb/backup/qemu/io/channel.c:257 #11 0x55ee1ca12c9a in multifd_send_thread /mnt/sdb/backup/qemu/migration/multifd.c:657 #12 0x55ee1d0607fc in qemu_thread_start /mnt/sdb/backup/qemu/util/qemu-thread-posix.c:519 #13 0x7f78159ae2dd in start_thread (/lib64/libpthread.so.0+0x82dd) #14 0x7f78156df4b2 in __GI___clone (/lib64/libc.so.6+0xfc4b2) Reported-by: Euler Robot <euler.robot@huawei.com> Signed-off-by: Pan Nengyuan <pannengyuan@huawei.com> Message-Id: <20200506095416.26099-3-pannengyuan@huawei.com> Reviewed-by: Juan Quintela <quintela@redhat.com> Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2020-05-06 12:54:16 +03:00
error_free(local_err);
}
}
multifd_send_cleanup_state();
}
static int multifd_zero_copy_flush(QIOChannel *c)
{
int ret;
Error *err = NULL;
ret = qio_channel_flush(c, &err);
if (ret < 0) {
error_report_err(err);
return -1;
}
if (ret == 1) {
stat64_add(&mig_stats.dirty_sync_missed_zero_copy, 1);
}
return ret;
}
int multifd_send_sync_main(void)
{
int i;
bool flush_zero_copy;
if (!migrate_multifd()) {
return 0;
}
if (multifd_send_state->pages->num) {
if (!multifd_send_pages()) {
error_report("%s: multifd_send_pages fail", __func__);
return -1;
}
}
flush_zero_copy = migrate_zero_copy_send();
for (i = 0; i < migrate_multifd_channels(); i++) {
MultiFDSendParams *p = &multifd_send_state->params[i];
if (multifd_send_should_exit()) {
return -1;
}
trace_multifd_send_sync_main_signal(p->id);
migration/multifd: Separate SYNC request with normal jobs Multifd provide a threaded model for processing jobs. On sender side, there can be two kinds of job: (1) a list of pages to send, or (2) a sync request. The sync request is a very special kind of job. It never contains a page array, but only a multifd packet telling the dest side to synchronize with sent pages. Before this patch, both requests use the pending_job field, no matter what the request is, it will boost pending_job, while multifd sender thread will decrement it after it finishes one job. However this should be racy, because SYNC is special in that it needs to set p->flags with MULTIFD_FLAG_SYNC, showing that this is a sync request. Consider a sequence of operations where: - migration thread enqueue a job to send some pages, pending_job++ (0->1) - [...before the selected multifd sender thread wakes up...] - migration thread enqueue another job to sync, pending_job++ (1->2), setup p->flags=MULTIFD_FLAG_SYNC - multifd sender thread wakes up, found pending_job==2 - send the 1st packet with MULTIFD_FLAG_SYNC and list of pages - send the 2nd packet with flags==0 and no pages This is not expected, because MULTIFD_FLAG_SYNC should hopefully be done after all the pages are received. Meanwhile, the 2nd packet will be completely useless, which contains zero information. I didn't verify above, but I think this issue is still benign in that at least on the recv side we always receive pages before handling MULTIFD_FLAG_SYNC. However that's not always guaranteed and just tricky. One other reason I want to separate it is using p->flags to communicate between the two threads is also not clearly defined, it's very hard to read and understand why accessing p->flags is always safe; see the current impl of multifd_send_thread() where we tried to cache only p->flags. It doesn't need to be that complicated. This patch introduces pending_sync, a separate flag just to show that the requester needs a sync. Alongside, we remove the tricky caching of p->flags now because after this patch p->flags should only be used by multifd sender thread now, which will be crystal clear. So it is always thread safe to access p->flags. With that, we can also safely convert the pending_job into a boolean, because we don't support >1 pending jobs anyway. Always use atomic ops to access both flags to make sure no cache effect. When at it, drop the initial setting of "pending_job = 0" because it's always allocated using g_new0(). Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-7-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:40 +03:00
/*
* We should be the only user so far, so not possible to be set by
* others concurrently.
*/
assert(qatomic_read(&p->pending_sync) == false);
qatomic_set(&p->pending_sync, true);
qemu_sem_post(&p->sem);
}
for (i = 0; i < migrate_multifd_channels(); i++) {
MultiFDSendParams *p = &multifd_send_state->params[i];
if (multifd_send_should_exit()) {
return -1;
}
qemu_sem_wait(&multifd_send_state->channels_ready);
trace_multifd_send_sync_main_wait(p->id);
qemu_sem_wait(&p->sem_sync);
if (flush_zero_copy && p->c && (multifd_zero_copy_flush(p->c) < 0)) {
return -1;
}
}
trace_multifd_send_sync_main(multifd_send_state->packet_num);
return 0;
}
static void *multifd_send_thread(void *opaque)
{
MultiFDSendParams *p = opaque;
MigrationThread *thread = NULL;
Error *local_err = NULL;
int ret = 0;
bool use_packets = multifd_use_packets();
thread = migration_threads_add(p->name, qemu_get_thread_id());
trace_multifd_send_thread_start(p->id);
rcu_register_thread();
if (use_packets) {
if (multifd_send_initial_packet(p, &local_err) < 0) {
ret = -1;
goto out;
}
}
while (true) {
qemu_sem_post(&multifd_send_state->channels_ready);
qemu_sem_wait(&p->sem);
if (multifd_send_should_exit()) {
break;
}
migration/multifd: Optimize sender side to be lockless When reviewing my attempt to refactor send_prepare(), Fabiano suggested we try out with dropping the mutex in multifd code [1]. I thought about that before but I never tried to change the code. Now maybe it's time to give it a stab. This only optimizes the sender side. The trick here is multifd has a clear provider/consumer model, that the migration main thread publishes requests (either pending_job/pending_sync), while the multifd sender threads are consumers. Here we don't have a lot of complicated data sharing, and the jobs can logically be submitted lockless. Arm the code with atomic weapons. Two things worth mentioning: - For multifd_send_pages(): we can use qatomic_load_acquire() when trying to find a free channel, but that's expensive if we attach one ACQUIRE per channel. Instead, keep the qatomic_read() on reading the pending_job flag as we do already, meanwhile use one smp_mb_acquire() after the loop to guarantee the memory ordering. - For pending_sync: it doesn't have any extra data required since now p->flags are never touched, it should be safe to not use memory barrier. That's different from pending_job. Provide rich comments for all the lockless operations to state how they are paired. With that, we can remove the mutex. [1] https://lore.kernel.org/r/87o7d1jlu5.fsf@suse.de Suggested-by: Fabiano Rosas <farosas@suse.de> Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-24-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:57 +03:00
/*
* Read pending_job flag before p->pages. Pairs with the
* qatomic_store_release() in multifd_send_pages().
*/
if (qatomic_load_acquire(&p->pending_job)) {
MultiFDPages_t *pages = p->pages;
p->iovs_num = 0;
assert(pages->num);
ret = multifd_send_state->ops->send_prepare(p, &local_err);
if (ret != 0) {
break;
}
if (migrate_mapped_ram()) {
ret = file_write_ramblock_iov(p->c, p->iov, p->iovs_num,
p->pages->block, &local_err);
} else {
ret = qio_channel_writev_full_all(p->c, p->iov, p->iovs_num,
NULL, 0, p->write_flags,
&local_err);
}
if (ret != 0) {
break;
}
stat64_add(&mig_stats.multifd_bytes,
p->next_packet_size + p->packet_len);
stat64_add(&mig_stats.normal_pages, pages->normal_num);
stat64_add(&mig_stats.zero_pages, pages->num - pages->normal_num);
multifd_pages_reset(p->pages);
p->next_packet_size = 0;
migration/multifd: Optimize sender side to be lockless When reviewing my attempt to refactor send_prepare(), Fabiano suggested we try out with dropping the mutex in multifd code [1]. I thought about that before but I never tried to change the code. Now maybe it's time to give it a stab. This only optimizes the sender side. The trick here is multifd has a clear provider/consumer model, that the migration main thread publishes requests (either pending_job/pending_sync), while the multifd sender threads are consumers. Here we don't have a lot of complicated data sharing, and the jobs can logically be submitted lockless. Arm the code with atomic weapons. Two things worth mentioning: - For multifd_send_pages(): we can use qatomic_load_acquire() when trying to find a free channel, but that's expensive if we attach one ACQUIRE per channel. Instead, keep the qatomic_read() on reading the pending_job flag as we do already, meanwhile use one smp_mb_acquire() after the loop to guarantee the memory ordering. - For pending_sync: it doesn't have any extra data required since now p->flags are never touched, it should be safe to not use memory barrier. That's different from pending_job. Provide rich comments for all the lockless operations to state how they are paired. With that, we can remove the mutex. [1] https://lore.kernel.org/r/87o7d1jlu5.fsf@suse.de Suggested-by: Fabiano Rosas <farosas@suse.de> Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-24-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:57 +03:00
/*
* Making sure p->pages is published before saying "we're
* free". Pairs with the smp_mb_acquire() in
* multifd_send_pages().
*/
qatomic_store_release(&p->pending_job, false);
} else {
migration/multifd: Optimize sender side to be lockless When reviewing my attempt to refactor send_prepare(), Fabiano suggested we try out with dropping the mutex in multifd code [1]. I thought about that before but I never tried to change the code. Now maybe it's time to give it a stab. This only optimizes the sender side. The trick here is multifd has a clear provider/consumer model, that the migration main thread publishes requests (either pending_job/pending_sync), while the multifd sender threads are consumers. Here we don't have a lot of complicated data sharing, and the jobs can logically be submitted lockless. Arm the code with atomic weapons. Two things worth mentioning: - For multifd_send_pages(): we can use qatomic_load_acquire() when trying to find a free channel, but that's expensive if we attach one ACQUIRE per channel. Instead, keep the qatomic_read() on reading the pending_job flag as we do already, meanwhile use one smp_mb_acquire() after the loop to guarantee the memory ordering. - For pending_sync: it doesn't have any extra data required since now p->flags are never touched, it should be safe to not use memory barrier. That's different from pending_job. Provide rich comments for all the lockless operations to state how they are paired. With that, we can remove the mutex. [1] https://lore.kernel.org/r/87o7d1jlu5.fsf@suse.de Suggested-by: Fabiano Rosas <farosas@suse.de> Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-24-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:57 +03:00
/*
* If not a normal job, must be a sync request. Note that
* pending_sync is a standalone flag (unlike pending_job), so
* it doesn't require explicit memory barriers.
*/
assert(qatomic_read(&p->pending_sync));
if (use_packets) {
p->flags = MULTIFD_FLAG_SYNC;
multifd_send_fill_packet(p);
ret = qio_channel_write_all(p->c, (void *)p->packet,
p->packet_len, &local_err);
if (ret != 0) {
break;
}
/* p->next_packet_size will always be zero for a SYNC packet */
stat64_add(&mig_stats.multifd_bytes, p->packet_len);
p->flags = 0;
}
migration/multifd: Separate SYNC request with normal jobs Multifd provide a threaded model for processing jobs. On sender side, there can be two kinds of job: (1) a list of pages to send, or (2) a sync request. The sync request is a very special kind of job. It never contains a page array, but only a multifd packet telling the dest side to synchronize with sent pages. Before this patch, both requests use the pending_job field, no matter what the request is, it will boost pending_job, while multifd sender thread will decrement it after it finishes one job. However this should be racy, because SYNC is special in that it needs to set p->flags with MULTIFD_FLAG_SYNC, showing that this is a sync request. Consider a sequence of operations where: - migration thread enqueue a job to send some pages, pending_job++ (0->1) - [...before the selected multifd sender thread wakes up...] - migration thread enqueue another job to sync, pending_job++ (1->2), setup p->flags=MULTIFD_FLAG_SYNC - multifd sender thread wakes up, found pending_job==2 - send the 1st packet with MULTIFD_FLAG_SYNC and list of pages - send the 2nd packet with flags==0 and no pages This is not expected, because MULTIFD_FLAG_SYNC should hopefully be done after all the pages are received. Meanwhile, the 2nd packet will be completely useless, which contains zero information. I didn't verify above, but I think this issue is still benign in that at least on the recv side we always receive pages before handling MULTIFD_FLAG_SYNC. However that's not always guaranteed and just tricky. One other reason I want to separate it is using p->flags to communicate between the two threads is also not clearly defined, it's very hard to read and understand why accessing p->flags is always safe; see the current impl of multifd_send_thread() where we tried to cache only p->flags. It doesn't need to be that complicated. This patch introduces pending_sync, a separate flag just to show that the requester needs a sync. Alongside, we remove the tricky caching of p->flags now because after this patch p->flags should only be used by multifd sender thread now, which will be crystal clear. So it is always thread safe to access p->flags. With that, we can also safely convert the pending_job into a boolean, because we don't support >1 pending jobs anyway. Always use atomic ops to access both flags to make sure no cache effect. When at it, drop the initial setting of "pending_job = 0" because it's always allocated using g_new0(). Reviewed-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240202102857.110210-7-peterx@redhat.com Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-02 13:28:40 +03:00
qatomic_set(&p->pending_sync, false);
qemu_sem_post(&p->sem_sync);
}
}
out:
if (ret) {
assert(local_err);
trace_multifd_send_error(p->id);
multifd_send_set_error(local_err);
multifd_send_kick_main(p);
error_free(local_err);
}
rcu_unregister_thread();
migration_threads_remove(thread);
trace_multifd_send_thread_end(p->id, p->packets_sent, p->total_normal_pages,
p->total_zero_pages);
return NULL;
}
static void multifd_new_send_channel_async(QIOTask *task, gpointer opaque);
typedef struct {
MultiFDSendParams *p;
QIOChannelTLS *tioc;
} MultiFDTLSThreadArgs;
static void *multifd_tls_handshake_thread(void *opaque)
{
MultiFDTLSThreadArgs *args = opaque;
qio_channel_tls_handshake(args->tioc,
multifd_new_send_channel_async,
args->p,
NULL,
NULL);
g_free(args);
return NULL;
}
static bool multifd_tls_channel_connect(MultiFDSendParams *p,
QIOChannel *ioc,
Error **errp)
{
MigrationState *s = migrate_get_current();
const char *hostname = s->hostname;
MultiFDTLSThreadArgs *args;
QIOChannelTLS *tioc;
tioc = migration_tls_client_create(ioc, hostname, errp);
if (!tioc) {
return false;
}
/*
* Ownership of the socket channel now transfers to the newly
* created TLS channel, which has already taken a reference.
*/
object_unref(OBJECT(ioc));
trace_multifd_tls_outgoing_handshake_start(ioc, tioc, hostname);
qio_channel_set_name(QIO_CHANNEL(tioc), "multifd-tls-outgoing");
args = g_new0(MultiFDTLSThreadArgs, 1);
args->tioc = tioc;
args->p = p;
p->tls_thread_created = true;
qemu_thread_create(&p->tls_thread, "mig/src/tls",
multifd_tls_handshake_thread, args,
QEMU_THREAD_JOINABLE);
return true;
}
void multifd_channel_connect(MultiFDSendParams *p, QIOChannel *ioc)
{
qio_channel_set_delay(ioc, false);
migration_ioc_register_yank(ioc);
/* Setup p->c only if the channel is completely setup */
p->c = ioc;
migration/multifd: Remove p->running We currently only need p->running to avoid calling qemu_thread_join() on a non existent thread if the thread has never been created. However, there are at least two bugs in this logic: 1) On the sending side, p->running is set too early and qemu_thread_create() can be skipped due to an error during TLS handshake, leaving the flag set and leading to a crash when multifd_send_cleanup() calls qemu_thread_join(). 2) During exit, the multifd thread clears the flag while holding the channel lock. The counterpart at multifd_send_cleanup() reads the flag outside of the lock and might free the mutex while the multifd thread still has it locked. Fix the first issue by setting the flag right before creating the thread. Rename it from p->running to p->thread_created to clarify its usage. Fix the second issue by not clearing the flag at the multifd thread exit. We don't have any use for that. Note that these bugs are straight-forward logic issues and not race conditions. There is still a gap for races to affect this code due to multifd_send_cleanup() being allowed to run concurrently with the thread creation loop. This issue is solved in the next patches. Cc: qemu-stable <qemu-stable@nongnu.org> Fixes: 29647140157a ("migration/tls: add support for multifd tls-handshake") Reported-by: Avihai Horon <avihaih@nvidia.com> Reported-by: chenyuhui5@huawei.com Reviewed-by: Peter Xu <peterx@redhat.com> Signed-off-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240206215118.6171-3-farosas@suse.de Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-07 00:51:14 +03:00
p->thread_created = true;
qemu_thread_create(&p->thread, p->name, multifd_send_thread, p,
QEMU_THREAD_JOINABLE);
}
/*
* When TLS is enabled this function is called once to establish the
* TLS connection and a second time after the TLS handshake to create
* the multifd channel. Without TLS it goes straight into the channel
* creation.
*/
static void multifd_new_send_channel_async(QIOTask *task, gpointer opaque)
{
MultiFDSendParams *p = opaque;
QIOChannel *ioc = QIO_CHANNEL(qio_task_get_source(task));
Error *local_err = NULL;
bool ret;
trace_multifd_new_send_channel_async(p->id);
if (qio_task_propagate_error(task, &local_err)) {
ret = false;
goto out;
}
trace_multifd_set_outgoing_channel(ioc, object_get_typename(OBJECT(ioc)),
migrate_get_current()->hostname);
if (migrate_channel_requires_tls_upgrade(ioc)) {
ret = multifd_tls_channel_connect(p, ioc, &local_err);
if (ret) {
return;
}
} else {
multifd_channel_connect(p, ioc);
ret = true;
}
out:
/*
* Here we're not interested whether creation succeeded, only that
* it happened at all.
*/
multifd_send_channel_created();
if (ret) {
return;
}
trace_multifd_new_send_channel_async_error(p->id, local_err);
multifd_send_set_error(local_err);
/*
* For error cases (TLS or non-TLS), IO channel is always freed here
* rather than when cleanup multifd: since p->c is not set, multifd
* cleanup code doesn't even know its existence.
*/
object_unref(OBJECT(ioc));
error_free(local_err);
}
static bool multifd_new_send_channel_create(gpointer opaque, Error **errp)
{
if (!multifd_use_packets()) {
return file_send_channel_create(opaque, errp);
}
socket_send_channel_create(multifd_new_send_channel_async, opaque);
return true;
}
bool multifd_send_setup(void)
{
MigrationState *s = migrate_get_current();
Error *local_err = NULL;
int thread_count, ret = 0;
uint32_t page_count = MULTIFD_PACKET_SIZE / qemu_target_page_size();
bool use_packets = multifd_use_packets();
uint8_t i;
if (!migrate_multifd()) {
return true;
}
thread_count = migrate_multifd_channels();
multifd_send_state = g_malloc0(sizeof(*multifd_send_state));
multifd_send_state->params = g_new0(MultiFDSendParams, thread_count);
multifd_send_state->pages = multifd_pages_init(page_count);
qemu_sem_init(&multifd_send_state->channels_created, 0);
qemu_sem_init(&multifd_send_state->channels_ready, 0);
qatomic_set(&multifd_send_state->exiting, 0);
multifd_send_state->ops = multifd_ops[migrate_multifd_compression()];
for (i = 0; i < thread_count; i++) {
MultiFDSendParams *p = &multifd_send_state->params[i];
qemu_sem_init(&p->sem, 0);
qemu_sem_init(&p->sem_sync, 0);
p->id = i;
p->pages = multifd_pages_init(page_count);
if (use_packets) {
p->packet_len = sizeof(MultiFDPacket_t)
+ sizeof(uint64_t) * page_count;
p->packet = g_malloc0(p->packet_len);
p->packet->magic = cpu_to_be32(MULTIFD_MAGIC);
p->packet->version = cpu_to_be32(MULTIFD_VERSION);
}
p->name = g_strdup_printf("mig/src/send_%d", i);
p->page_size = qemu_target_page_size();
p->page_count = page_count;
p->write_flags = 0;
if (!multifd_new_send_channel_create(p, &local_err)) {
return false;
}
}
/*
* Wait until channel creation has started for all channels. The
* creation can still fail, but no more channels will be created
* past this point.
*/
for (i = 0; i < thread_count; i++) {
qemu_sem_wait(&multifd_send_state->channels_created);
}
for (i = 0; i < thread_count; i++) {
MultiFDSendParams *p = &multifd_send_state->params[i];
ret = multifd_send_state->ops->send_setup(p, &local_err);
if (ret) {
break;
}
}
if (ret) {
migrate_set_error(s, local_err);
error_report_err(local_err);
migrate_set_state(&s->state, MIGRATION_STATUS_SETUP,
MIGRATION_STATUS_FAILED);
return false;
}
return true;
}
bool multifd_recv(void)
{
int i;
static int next_recv_channel;
MultiFDRecvParams *p = NULL;
MultiFDRecvData *data = multifd_recv_state->data;
/*
* next_channel can remain from a previous migration that was
* using more channels, so ensure it doesn't overflow if the
* limit is lower now.
*/
next_recv_channel %= migrate_multifd_channels();
for (i = next_recv_channel;; i = (i + 1) % migrate_multifd_channels()) {
if (multifd_recv_should_exit()) {
return false;
}
p = &multifd_recv_state->params[i];
if (qatomic_read(&p->pending_job) == false) {
next_recv_channel = (i + 1) % migrate_multifd_channels();
break;
}
}
/*
* Order pending_job read before manipulating p->data below. Pairs
* with qatomic_store_release() at multifd_recv_thread().
*/
smp_mb_acquire();
assert(!p->data->size);
multifd_recv_state->data = p->data;
p->data = data;
/*
* Order p->data update before setting pending_job. Pairs with
* qatomic_load_acquire() at multifd_recv_thread().
*/
qatomic_store_release(&p->pending_job, true);
qemu_sem_post(&p->sem);
return true;
}
MultiFDRecvData *multifd_get_recv_data(void)
{
return multifd_recv_state->data;
}
static void multifd_recv_terminate_threads(Error *err)
{
int i;
trace_multifd_recv_terminate_threads(err != NULL);
if (qatomic_xchg(&multifd_recv_state->exiting, 1)) {
return;
}
if (err) {
MigrationState *s = migrate_get_current();
migrate_set_error(s, err);
if (s->state == MIGRATION_STATUS_SETUP ||
s->state == MIGRATION_STATUS_ACTIVE) {
migrate_set_state(&s->state, s->state,
MIGRATION_STATUS_FAILED);
}
}
for (i = 0; i < migrate_multifd_channels(); i++) {
MultiFDRecvParams *p = &multifd_recv_state->params[i];
/*
* The migration thread and channels interact differently
* depending on the presence of packets.
*/
if (multifd_use_packets()) {
/*
* The channel receives as long as there are packets. When
* packets end (i.e. MULTIFD_FLAG_SYNC is reached), the
* channel waits for the migration thread to sync. If the
* sync never happens, do it here.
*/
qemu_sem_post(&p->sem_sync);
} else {
/*
* The channel waits for the migration thread to give it
* work. When the migration thread runs out of work, it
* releases the channel and waits for any pending work to
* finish. If we reach here (e.g. due to error) before the
* work runs out, release the channel.
*/
qemu_sem_post(&p->sem);
}
/*
* We could arrive here for two reasons:
* - normal quit, i.e. everything went fine, just finished
* - error quit: We close the channels so the channel threads
* finish the qio_channel_read_all_eof()
*/
if (p->c) {
qio_channel_shutdown(p->c, QIO_CHANNEL_SHUTDOWN_BOTH, NULL);
}
}
}
void multifd_recv_shutdown(void)
migration/multifd: Move load_cleanup inside incoming_state_destroy Currently running migration_incoming_state_destroy() without first running multifd_load_cleanup() will cause a yank error: qemu-system-x86_64: ../util/yank.c:107: yank_unregister_instance: Assertion `QLIST_EMPTY(&entry->yankfns)' failed. (core dumped) The above error happens in the target host, when multifd is being used for precopy, and then postcopy is triggered and the migration finishes. This will crash the VM in the target host. To avoid that, move multifd_load_cleanup() inside migration_incoming_state_destroy(), so that the load cleanup becomes part of the incoming state destroying process. Running multifd_load_cleanup() twice can become an issue, though, but the only scenario it could be ran twice is on process_incoming_migration_bh(). So removing this extra call is necessary. On the other hand, this multifd_load_cleanup() call happens way before the migration_incoming_state_destroy() and having this happening before dirty_bitmap_mig_before_vm_start() and vm_start() may be a need. So introduce a new function multifd_load_shutdown() that will mainly stop all multifd threads and close their QIOChannels. Then use this function instead of multifd_load_cleanup() to make sure nothing else is received before dirty_bitmap_mig_before_vm_start(). Fixes: b5eea99ec2 ("migration: Add yank feature") Reported-by: Li Xiaohui <xiaohli@redhat.com> Signed-off-by: Leonardo Bras <leobras@redhat.com> Reviewed-by: Juan Quintela <quintela@redhat.com> Reviewed-by: Peter Xu <peterx@redhat.com> Signed-off-by: Juan Quintela <quintela@redhat.com>
2023-02-10 09:36:31 +03:00
{
if (migrate_multifd()) {
migration/multifd: Move load_cleanup inside incoming_state_destroy Currently running migration_incoming_state_destroy() without first running multifd_load_cleanup() will cause a yank error: qemu-system-x86_64: ../util/yank.c:107: yank_unregister_instance: Assertion `QLIST_EMPTY(&entry->yankfns)' failed. (core dumped) The above error happens in the target host, when multifd is being used for precopy, and then postcopy is triggered and the migration finishes. This will crash the VM in the target host. To avoid that, move multifd_load_cleanup() inside migration_incoming_state_destroy(), so that the load cleanup becomes part of the incoming state destroying process. Running multifd_load_cleanup() twice can become an issue, though, but the only scenario it could be ran twice is on process_incoming_migration_bh(). So removing this extra call is necessary. On the other hand, this multifd_load_cleanup() call happens way before the migration_incoming_state_destroy() and having this happening before dirty_bitmap_mig_before_vm_start() and vm_start() may be a need. So introduce a new function multifd_load_shutdown() that will mainly stop all multifd threads and close their QIOChannels. Then use this function instead of multifd_load_cleanup() to make sure nothing else is received before dirty_bitmap_mig_before_vm_start(). Fixes: b5eea99ec2 ("migration: Add yank feature") Reported-by: Li Xiaohui <xiaohli@redhat.com> Signed-off-by: Leonardo Bras <leobras@redhat.com> Reviewed-by: Juan Quintela <quintela@redhat.com> Reviewed-by: Peter Xu <peterx@redhat.com> Signed-off-by: Juan Quintela <quintela@redhat.com>
2023-02-10 09:36:31 +03:00
multifd_recv_terminate_threads(NULL);
}
}
static void multifd_recv_cleanup_channel(MultiFDRecvParams *p)
{
migration_ioc_unregister_yank(p->c);
object_unref(OBJECT(p->c));
p->c = NULL;
qemu_mutex_destroy(&p->mutex);
qemu_sem_destroy(&p->sem_sync);
qemu_sem_destroy(&p->sem);
g_free(p->name);
p->name = NULL;
p->packet_len = 0;
g_free(p->packet);
p->packet = NULL;
g_free(p->normal);
p->normal = NULL;
g_free(p->zero);
p->zero = NULL;
multifd_recv_state->ops->recv_cleanup(p);
}
static void multifd_recv_cleanup_state(void)
{
qemu_sem_destroy(&multifd_recv_state->sem_sync);
g_free(multifd_recv_state->params);
multifd_recv_state->params = NULL;
g_free(multifd_recv_state->data);
multifd_recv_state->data = NULL;
g_free(multifd_recv_state);
multifd_recv_state = NULL;
}
void multifd_recv_cleanup(void)
{
int i;
if (!migrate_multifd()) {
return;
}
multifd_recv_terminate_threads(NULL);
for (i = 0; i < migrate_multifd_channels(); i++) {
MultiFDRecvParams *p = &multifd_recv_state->params[i];
migration/multifd: Remove p->running We currently only need p->running to avoid calling qemu_thread_join() on a non existent thread if the thread has never been created. However, there are at least two bugs in this logic: 1) On the sending side, p->running is set too early and qemu_thread_create() can be skipped due to an error during TLS handshake, leaving the flag set and leading to a crash when multifd_send_cleanup() calls qemu_thread_join(). 2) During exit, the multifd thread clears the flag while holding the channel lock. The counterpart at multifd_send_cleanup() reads the flag outside of the lock and might free the mutex while the multifd thread still has it locked. Fix the first issue by setting the flag right before creating the thread. Rename it from p->running to p->thread_created to clarify its usage. Fix the second issue by not clearing the flag at the multifd thread exit. We don't have any use for that. Note that these bugs are straight-forward logic issues and not race conditions. There is still a gap for races to affect this code due to multifd_send_cleanup() being allowed to run concurrently with the thread creation loop. This issue is solved in the next patches. Cc: qemu-stable <qemu-stable@nongnu.org> Fixes: 29647140157a ("migration/tls: add support for multifd tls-handshake") Reported-by: Avihai Horon <avihaih@nvidia.com> Reported-by: chenyuhui5@huawei.com Reviewed-by: Peter Xu <peterx@redhat.com> Signed-off-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240206215118.6171-3-farosas@suse.de Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-07 00:51:14 +03:00
if (p->thread_created) {
qemu_thread_join(&p->thread);
}
}
for (i = 0; i < migrate_multifd_channels(); i++) {
multifd_recv_cleanup_channel(&multifd_recv_state->params[i]);
}
multifd_recv_cleanup_state();
}
void multifd_recv_sync_main(void)
{
int thread_count = migrate_multifd_channels();
bool file_based = !multifd_use_packets();
int i;
if (!migrate_multifd()) {
return;
}
/*
* File-based channels don't use packets and therefore need to
* wait for more work. Release them to start the sync.
*/
if (file_based) {
for (i = 0; i < thread_count; i++) {
MultiFDRecvParams *p = &multifd_recv_state->params[i];
trace_multifd_recv_sync_main_signal(p->id);
qemu_sem_post(&p->sem);
}
}
/*
* Initiate the synchronization by waiting for all channels.
*
* For socket-based migration this means each channel has received
* the SYNC packet on the stream.
*
* For file-based migration this means each channel is done with
* the work (pending_job=false).
*/
for (i = 0; i < thread_count; i++) {
trace_multifd_recv_sync_main_wait(i);
qemu_sem_wait(&multifd_recv_state->sem_sync);
}
if (file_based) {
/*
* For file-based loading is done in one iteration. We're
* done.
*/
return;
}
/*
* Sync done. Release the channels for the next iteration.
*/
for (i = 0; i < thread_count; i++) {
MultiFDRecvParams *p = &multifd_recv_state->params[i];
WITH_QEMU_LOCK_GUARD(&p->mutex) {
if (multifd_recv_state->packet_num < p->packet_num) {
multifd_recv_state->packet_num = p->packet_num;
}
}
trace_multifd_recv_sync_main_signal(p->id);
qemu_sem_post(&p->sem_sync);
}
trace_multifd_recv_sync_main(multifd_recv_state->packet_num);
}
static void *multifd_recv_thread(void *opaque)
{
MultiFDRecvParams *p = opaque;
Error *local_err = NULL;
bool use_packets = multifd_use_packets();
int ret;
trace_multifd_recv_thread_start(p->id);
rcu_register_thread();
while (true) {
uint32_t flags = 0;
bool has_data = false;
p->normal_num = 0;
if (use_packets) {
if (multifd_recv_should_exit()) {
break;
}
ret = qio_channel_read_all_eof(p->c, (void *)p->packet,
p->packet_len, &local_err);
if (ret == 0 || ret == -1) { /* 0: EOF -1: Error */
break;
}
qemu_mutex_lock(&p->mutex);
ret = multifd_recv_unfill_packet(p, &local_err);
if (ret) {
qemu_mutex_unlock(&p->mutex);
break;
}
flags = p->flags;
/* recv methods don't know how to handle the SYNC flag */
p->flags &= ~MULTIFD_FLAG_SYNC;
has_data = p->normal_num || p->zero_num;
qemu_mutex_unlock(&p->mutex);
} else {
/*
* No packets, so we need to wait for the vmstate code to
* give us work.
*/
qemu_sem_wait(&p->sem);
if (multifd_recv_should_exit()) {
break;
}
/* pairs with qatomic_store_release() at multifd_recv() */
if (!qatomic_load_acquire(&p->pending_job)) {
/*
* Migration thread did not send work, this is
* equivalent to pending_sync on the sending
* side. Post sem_sync to notify we reached this
* point.
*/
qemu_sem_post(&multifd_recv_state->sem_sync);
continue;
}
has_data = !!p->data->size;
}
if (has_data) {
ret = multifd_recv_state->ops->recv(p, &local_err);
if (ret != 0) {
break;
}
}
if (use_packets) {
if (flags & MULTIFD_FLAG_SYNC) {
qemu_sem_post(&multifd_recv_state->sem_sync);
qemu_sem_wait(&p->sem_sync);
}
} else {
p->total_normal_pages += p->data->size / qemu_target_page_size();
p->data->size = 0;
/*
* Order data->size update before clearing
* pending_job. Pairs with smp_mb_acquire() at
* multifd_recv().
*/
qatomic_store_release(&p->pending_job, false);
}
}
if (local_err) {
multifd_recv_terminate_threads(local_err);
migration/multifd: Do error_free after migrate_set_error to avoid memleaks When error happen in multifd_send_thread, it use error_copy to set migrate error in multifd_send_terminate_threads(). We should call error_free after it. Similarly, fix another two places in multifd_recv_thread/multifd_save_cleanup. The leak stack: Direct leak of 48 byte(s) in 1 object(s) allocated from: #0 0x7f781af07cf0 in calloc (/lib64/libasan.so.5+0xefcf0) #1 0x7f781a2ce22d in g_malloc0 (/lib64/libglib-2.0.so.0+0x5322d) #2 0x55ee1d075c17 in error_setv /mnt/sdb/backup/qemu/util/error.c:61 #3 0x55ee1d076464 in error_setg_errno_internal /mnt/sdb/backup/qemu/util/error.c:109 #4 0x55ee1cef066e in qio_channel_socket_writev /mnt/sdb/backup/qemu/io/channel-socket.c:569 #5 0x55ee1cee806b in qio_channel_writev /mnt/sdb/backup/qemu/io/channel.c:207 #6 0x55ee1cee806b in qio_channel_writev_all /mnt/sdb/backup/qemu/io/channel.c:171 #7 0x55ee1cee8248 in qio_channel_write_all /mnt/sdb/backup/qemu/io/channel.c:257 #8 0x55ee1ca12c9a in multifd_send_thread /mnt/sdb/backup/qemu/migration/multifd.c:657 #9 0x55ee1d0607fc in qemu_thread_start /mnt/sdb/backup/qemu/util/qemu-thread-posix.c:519 #10 0x7f78159ae2dd in start_thread (/lib64/libpthread.so.0+0x82dd) #11 0x7f78156df4b2 in __GI___clone (/lib64/libc.so.6+0xfc4b2) Indirect leak of 52 byte(s) in 1 object(s) allocated from: #0 0x7f781af07f28 in __interceptor_realloc (/lib64/libasan.so.5+0xeff28) #1 0x7f78156f07d9 in __GI___vasprintf_chk (/lib64/libc.so.6+0x10d7d9) #2 0x7f781a30ea6c in g_vasprintf (/lib64/libglib-2.0.so.0+0x93a6c) #3 0x7f781a2e7cd0 in g_strdup_vprintf (/lib64/libglib-2.0.so.0+0x6ccd0) #4 0x7f781a2e7d8c in g_strdup_printf (/lib64/libglib-2.0.so.0+0x6cd8c) #5 0x55ee1d075c86 in error_setv /mnt/sdb/backup/qemu/util/error.c:65 #6 0x55ee1d076464 in error_setg_errno_internal /mnt/sdb/backup/qemu/util/error.c:109 #7 0x55ee1cef066e in qio_channel_socket_writev /mnt/sdb/backup/qemu/io/channel-socket.c:569 #8 0x55ee1cee806b in qio_channel_writev /mnt/sdb/backup/qemu/io/channel.c:207 #9 0x55ee1cee806b in qio_channel_writev_all /mnt/sdb/backup/qemu/io/channel.c:171 #10 0x55ee1cee8248 in qio_channel_write_all /mnt/sdb/backup/qemu/io/channel.c:257 #11 0x55ee1ca12c9a in multifd_send_thread /mnt/sdb/backup/qemu/migration/multifd.c:657 #12 0x55ee1d0607fc in qemu_thread_start /mnt/sdb/backup/qemu/util/qemu-thread-posix.c:519 #13 0x7f78159ae2dd in start_thread (/lib64/libpthread.so.0+0x82dd) #14 0x7f78156df4b2 in __GI___clone (/lib64/libc.so.6+0xfc4b2) Reported-by: Euler Robot <euler.robot@huawei.com> Signed-off-by: Pan Nengyuan <pannengyuan@huawei.com> Message-Id: <20200506095416.26099-3-pannengyuan@huawei.com> Reviewed-by: Juan Quintela <quintela@redhat.com> Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
2020-05-06 12:54:16 +03:00
error_free(local_err);
}
rcu_unregister_thread();
trace_multifd_recv_thread_end(p->id, p->packets_recved,
p->total_normal_pages,
p->total_zero_pages);
return NULL;
}
int multifd_recv_setup(Error **errp)
{
int thread_count;
uint32_t page_count = MULTIFD_PACKET_SIZE / qemu_target_page_size();
bool use_packets = multifd_use_packets();
uint8_t i;
2022-12-20 21:44:18 +03:00
/*
* Return successfully if multiFD recv state is already initialised
* or multiFD is not enabled.
*/
if (multifd_recv_state || !migrate_multifd()) {
return 0;
}
2022-12-20 21:44:18 +03:00
thread_count = migrate_multifd_channels();
multifd_recv_state = g_malloc0(sizeof(*multifd_recv_state));
multifd_recv_state->params = g_new0(MultiFDRecvParams, thread_count);
multifd_recv_state->data = g_new0(MultiFDRecvData, 1);
multifd_recv_state->data->size = 0;
qatomic_set(&multifd_recv_state->count, 0);
qatomic_set(&multifd_recv_state->exiting, 0);
qemu_sem_init(&multifd_recv_state->sem_sync, 0);
multifd_recv_state->ops = multifd_ops[migrate_multifd_compression()];
for (i = 0; i < thread_count; i++) {
MultiFDRecvParams *p = &multifd_recv_state->params[i];
qemu_mutex_init(&p->mutex);
qemu_sem_init(&p->sem_sync, 0);
qemu_sem_init(&p->sem, 0);
p->pending_job = false;
p->id = i;
p->data = g_new0(MultiFDRecvData, 1);
p->data->size = 0;
if (use_packets) {
p->packet_len = sizeof(MultiFDPacket_t)
+ sizeof(uint64_t) * page_count;
p->packet = g_malloc0(p->packet_len);
}
p->name = g_strdup_printf("mig/dst/recv_%d", i);
p->normal = g_new0(ram_addr_t, page_count);
p->zero = g_new0(ram_addr_t, page_count);
p->page_count = page_count;
p->page_size = qemu_target_page_size();
}
for (i = 0; i < thread_count; i++) {
MultiFDRecvParams *p = &multifd_recv_state->params[i];
int ret;
ret = multifd_recv_state->ops->recv_setup(p, errp);
if (ret) {
return ret;
}
}
return 0;
}
bool multifd_recv_all_channels_created(void)
{
int thread_count = migrate_multifd_channels();
if (!migrate_multifd()) {
return true;
}
if (!multifd_recv_state) {
/* Called before any connections created */
return false;
}
return thread_count == qatomic_read(&multifd_recv_state->count);
}
/*
* Try to receive all multifd channels to get ready for the migration.
2022-12-20 21:44:18 +03:00
* Sets @errp when failing to receive the current channel.
*/
2022-12-20 21:44:18 +03:00
void multifd_recv_new_channel(QIOChannel *ioc, Error **errp)
{
MultiFDRecvParams *p;
Error *local_err = NULL;
bool use_packets = multifd_use_packets();
int id;
if (use_packets) {
id = multifd_recv_initial_packet(ioc, &local_err);
if (id < 0) {
multifd_recv_terminate_threads(local_err);
error_propagate_prepend(errp, local_err,
"failed to receive packet"
" via multifd channel %d: ",
qatomic_read(&multifd_recv_state->count));
return;
}
trace_multifd_recv_new_channel(id);
} else {
id = qatomic_read(&multifd_recv_state->count);
}
p = &multifd_recv_state->params[id];
if (p->c != NULL) {
error_setg(&local_err, "multifd: received id '%d' already setup'",
id);
multifd_recv_terminate_threads(local_err);
error_propagate(errp, local_err);
2022-12-20 21:44:18 +03:00
return;
}
p->c = ioc;
object_ref(OBJECT(ioc));
migration/multifd: Remove p->running We currently only need p->running to avoid calling qemu_thread_join() on a non existent thread if the thread has never been created. However, there are at least two bugs in this logic: 1) On the sending side, p->running is set too early and qemu_thread_create() can be skipped due to an error during TLS handshake, leaving the flag set and leading to a crash when multifd_send_cleanup() calls qemu_thread_join(). 2) During exit, the multifd thread clears the flag while holding the channel lock. The counterpart at multifd_send_cleanup() reads the flag outside of the lock and might free the mutex while the multifd thread still has it locked. Fix the first issue by setting the flag right before creating the thread. Rename it from p->running to p->thread_created to clarify its usage. Fix the second issue by not clearing the flag at the multifd thread exit. We don't have any use for that. Note that these bugs are straight-forward logic issues and not race conditions. There is still a gap for races to affect this code due to multifd_send_cleanup() being allowed to run concurrently with the thread creation loop. This issue is solved in the next patches. Cc: qemu-stable <qemu-stable@nongnu.org> Fixes: 29647140157a ("migration/tls: add support for multifd tls-handshake") Reported-by: Avihai Horon <avihaih@nvidia.com> Reported-by: chenyuhui5@huawei.com Reviewed-by: Peter Xu <peterx@redhat.com> Signed-off-by: Fabiano Rosas <farosas@suse.de> Link: https://lore.kernel.org/r/20240206215118.6171-3-farosas@suse.de Signed-off-by: Peter Xu <peterx@redhat.com>
2024-02-07 00:51:14 +03:00
p->thread_created = true;
qemu_thread_create(&p->thread, p->name, multifd_recv_thread, p,
QEMU_THREAD_JOINABLE);
qatomic_inc(&multifd_recv_state->count);
}
bool multifd_send_prepare_common(MultiFDSendParams *p)
{
multifd_send_zero_page_detect(p);
if (!p->pages->normal_num) {
p->next_packet_size = 0;
return false;
}
multifd_send_prepare_header(p);
return true;
}