qemu-e2k/scsi/qemu-pr-helper.c

1078 lines
30 KiB
C
Raw Normal View History

/*
* Privileged helper to handle persistent reservation commands for QEMU
*
* Copyright (C) 2017 Red Hat, Inc. <pbonzini@redhat.com>
*
* Author: Paolo Bonzini <pbonzini@redhat.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; under version 2 of the License.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
#include "qemu/osdep.h"
#include <getopt.h>
#include <sys/ioctl.h>
#include <linux/dm-ioctl.h>
#include <scsi/sg.h>
#ifdef CONFIG_LIBCAP_NG
#include <cap-ng.h>
#endif
#include <pwd.h>
#include <grp.h>
#ifdef CONFIG_MPATH
#include <libudev.h>
#include <mpath_cmd.h>
#include <mpath_persist.h>
#endif
#include "qemu/help-texts.h"
#include "qapi/error.h"
#include "qemu/cutils.h"
#include "qemu/main-loop.h"
#include "qemu/module.h"
#include "qemu/error-report.h"
#include "qemu/config-file.h"
#include "qemu/bswap.h"
#include "qemu/log.h"
#include "qemu/systemd.h"
#include "qapi/util.h"
#include "qapi/qmp/qstring.h"
#include "io/channel-socket.h"
#include "trace/control.h"
#include "qemu-version.h"
#include "block/aio.h"
#include "block/thread-pool.h"
#include "scsi/constants.h"
#include "scsi/utils.h"
#include "pr-helper.h"
#define PR_OUT_FIXED_PARAM_SIZE 24
static char *socket_path;
static char *pidfile;
static enum { RUNNING, TERMINATE, TERMINATING } state;
static QIOChannelSocket *server_ioc;
static int server_watch;
static int num_active_sockets = 1;
static int noisy;
static int verbose;
#ifdef CONFIG_LIBCAP_NG
static int uid = -1;
static int gid = -1;
#endif
static void compute_default_paths(void)
{
g_autofree char *state = qemu_get_local_state_dir();
socket_path = g_build_filename(state, "run", "qemu-pr-helper.sock", NULL);
pidfile = g_build_filename(state, "run", "qemu-pr-helper.pid", NULL);
}
static void usage(const char *name)
{
(printf) (
"Usage: %s [OPTIONS] FILE\n"
"Persistent Reservation helper program for QEMU\n"
"\n"
" -h, --help display this help and exit\n"
" -V, --version output version information and exit\n"
"\n"
" -d, --daemon run in the background\n"
" -f, --pidfile=PATH PID file when running as a daemon\n"
" (default '%s')\n"
" -k, --socket=PATH path to the unix socket\n"
" (default '%s')\n"
" -T, --trace [[enable=]<pattern>][,events=<file>][,file=<file>]\n"
" specify tracing options\n"
#ifdef CONFIG_LIBCAP_NG
" -u, --user=USER user to drop privileges to\n"
" -g, --group=GROUP group to drop privileges to\n"
#endif
"\n"
QEMU_HELP_BOTTOM "\n"
, name, pidfile, socket_path);
}
static void version(const char *name)
{
printf(
"%s " QEMU_FULL_VERSION "\n"
"Written by Paolo Bonzini.\n"
"\n"
QEMU_COPYRIGHT "\n"
"This is free software; see the source for copying conditions. There is NO\n"
"warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n"
, name);
}
/* SG_IO support */
typedef struct PRHelperSGIOData {
int fd;
const uint8_t *cdb;
uint8_t *sense;
uint8_t *buf;
int sz; /* input/output */
int dir;
} PRHelperSGIOData;
static int do_sgio_worker(void *opaque)
{
PRHelperSGIOData *data = opaque;
struct sg_io_hdr io_hdr;
int ret;
int status;
SCSISense sense_code;
memset(data->sense, 0, PR_HELPER_SENSE_SIZE);
memset(&io_hdr, 0, sizeof(io_hdr));
io_hdr.interface_id = 'S';
io_hdr.cmd_len = PR_HELPER_CDB_SIZE;
io_hdr.cmdp = (uint8_t *)data->cdb;
io_hdr.sbp = data->sense;
io_hdr.mx_sb_len = PR_HELPER_SENSE_SIZE;
io_hdr.timeout = 1;
io_hdr.dxfer_direction = data->dir;
io_hdr.dxferp = (char *)data->buf;
io_hdr.dxfer_len = data->sz;
ret = ioctl(data->fd, SG_IO, &io_hdr);
if (ret < 0) {
status = scsi_sense_from_errno(errno, &sense_code);
if (status == CHECK_CONDITION) {
scsi_build_sense(data->sense, sense_code);
}
} else if (io_hdr.host_status != SCSI_HOST_OK) {
status = scsi_sense_from_host_status(io_hdr.host_status, &sense_code);
if (status == CHECK_CONDITION) {
scsi_build_sense(data->sense, sense_code);
}
} else if (io_hdr.driver_status & SG_ERR_DRIVER_TIMEOUT) {
status = BUSY;
} else {
status = io_hdr.status;
}
if (status == GOOD) {
data->sz -= io_hdr.resid;
} else {
data->sz = 0;
}
return status;
}
static int coroutine_fn do_sgio(int fd, const uint8_t *cdb, uint8_t *sense,
uint8_t *buf, int *sz, int dir)
{
int r;
PRHelperSGIOData data = {
.fd = fd,
.cdb = cdb,
.sense = sense,
.buf = buf,
.sz = *sz,
.dir = dir,
};
r = thread_pool_submit_co(do_sgio_worker, &data);
*sz = data.sz;
return r;
}
/* Device mapper interface */
#ifdef CONFIG_MPATH
#define CONTROL_PATH "/dev/mapper/control"
typedef struct DMData {
struct dm_ioctl dm;
uint8_t data[1024];
} DMData;
static int control_fd;
static void *dm_ioctl(int ioc, struct dm_ioctl *dm)
{
static DMData d;
memcpy(&d.dm, dm, sizeof(d.dm));
QEMU_BUILD_BUG_ON(sizeof(d.data) < sizeof(struct dm_target_spec));
d.dm.version[0] = DM_VERSION_MAJOR;
d.dm.version[1] = 0;
d.dm.version[2] = 0;
d.dm.data_size = 1024;
d.dm.data_start = offsetof(DMData, data);
if (ioctl(control_fd, ioc, &d) < 0) {
return NULL;
}
memcpy(dm, &d.dm, sizeof(d.dm));
return &d.data;
}
static void *dm_dev_ioctl(int fd, int ioc, struct dm_ioctl *dm)
{
struct stat st;
int r;
r = fstat(fd, &st);
if (r < 0) {
perror("fstat");
exit(1);
}
dm->dev = st.st_rdev;
return dm_ioctl(ioc, dm);
}
static void dm_init(void)
{
control_fd = open(CONTROL_PATH, O_RDWR);
if (control_fd < 0) {
perror("Cannot open " CONTROL_PATH);
exit(1);
}
struct dm_ioctl dm = { };
if (!dm_ioctl(DM_VERSION, &dm)) {
perror("ioctl");
exit(1);
}
if (dm.version[0] != DM_VERSION_MAJOR) {
fprintf(stderr, "Unsupported device mapper interface");
exit(1);
}
}
/* Variables required by libmultipath and libmpathpersist. */
QEMU_BUILD_BUG_ON(PR_HELPER_DATA_SIZE > MPATH_MAX_PARAM_LEN);
static struct config *multipath_conf;
unsigned mpath_mx_alloc_len = PR_HELPER_DATA_SIZE;
int logsink;
struct udev *udev;
extern struct config *get_multipath_config(void);
struct config *get_multipath_config(void)
{
return multipath_conf;
}
extern void put_multipath_config(struct config *conf);
void put_multipath_config(struct config *conf)
{
}
static void multipath_pr_init(void)
{
udev = udev_new();
multipath_conf = mpath_lib_init();
}
static int is_mpath(int fd)
{
struct dm_ioctl dm = { .flags = DM_NOFLUSH_FLAG };
struct dm_target_spec *tgt;
tgt = dm_dev_ioctl(fd, DM_TABLE_STATUS, &dm);
if (!tgt) {
if (errno == ENXIO) {
return 0;
}
perror("ioctl");
exit(EXIT_FAILURE);
}
return !strncmp(tgt->target_type, "multipath", DM_MAX_TYPE_NAME);
}
static SCSISense mpath_generic_sense(int r)
{
switch (r) {
case MPATH_PR_SENSE_NOT_READY:
return SENSE_CODE(NOT_READY);
case MPATH_PR_SENSE_MEDIUM_ERROR:
return SENSE_CODE(READ_ERROR);
case MPATH_PR_SENSE_HARDWARE_ERROR:
return SENSE_CODE(TARGET_FAILURE);
case MPATH_PR_SENSE_ABORTED_COMMAND:
return SENSE_CODE(IO_ERROR);
default:
abort();
}
}
static int coroutine_fn mpath_reconstruct_sense(int fd, int r, uint8_t *sense)
{
switch (r) {
case MPATH_PR_SUCCESS:
return GOOD;
case MPATH_PR_SENSE_NOT_READY:
case MPATH_PR_SENSE_MEDIUM_ERROR:
case MPATH_PR_SENSE_HARDWARE_ERROR:
case MPATH_PR_SENSE_ABORTED_COMMAND:
{
/* libmpathpersist ate the exact sense. Try to find it by
* issuing TEST UNIT READY.
*/
uint8_t cdb[6] = { TEST_UNIT_READY };
int sz = 0;
qemu-pr-helper: fix crash in mpath_reconstruct_sense The 'r' variable was accidently shadowed, and because of this we were always passing 0 to mpath_generic_sense, instead of original return value, which triggers an abort() This is an attempt to fix the https://bugzilla.redhat.com/show_bug.cgi?id=1720047 although there might be other places in the code that trigger qemu-pr-helper crash, and this fix might not be the root cause. The crash was reproduced by creating an iscsi target on a test machine, and passing it twice to the guest like that: -blockdev node-name=idisk0,driver=iscsi,transport=...,target=... -device scsi-block,drive=idisk0,bus=scsi0.0,bootindex=-1,scsi-id=1,lun=0,share-rw=on -device scsi-block,drive=idisk0,bus=scsi0.0,bootindex=-1,scsi-id=1,lun=1,share-rw=on Then in the guest, both /dev/sda and /dev/sdb were aggregated by multipath to /dev/mpatha, which was passed to a nested guest like that -object pr-manager-helper,id=qemu_pr_helper,path=/root/work/vm/testvm/.run/pr_helper.socket -blockdev node-name=test,driver=host_device,filename=/dev/mapper/mpatha,pr-manager=qemu_pr_helper -device scsi-block,drive=test,bus=scsi0.0,bootindex=-1,scsi-id=0,lun=0 The nested guest run: sg_persist --no-inquiry -v --out --register --param-sark 0x1234 /dev/sda Strictly speaking this is wrong configuration since qemu is where the multipath was split, and thus the iscsi target was not aware of multipath, and thus when libmpathpersist code rightfully tried to register the PR key on all paths, it failed to do so. However qemu-pr-helper should not crash in this case. Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2019-09-24 01:38:48 +02:00
int ret = do_sgio(fd, cdb, sense, NULL, &sz, SG_DXFER_NONE);
qemu-pr-helper: fix crash in mpath_reconstruct_sense The 'r' variable was accidently shadowed, and because of this we were always passing 0 to mpath_generic_sense, instead of original return value, which triggers an abort() This is an attempt to fix the https://bugzilla.redhat.com/show_bug.cgi?id=1720047 although there might be other places in the code that trigger qemu-pr-helper crash, and this fix might not be the root cause. The crash was reproduced by creating an iscsi target on a test machine, and passing it twice to the guest like that: -blockdev node-name=idisk0,driver=iscsi,transport=...,target=... -device scsi-block,drive=idisk0,bus=scsi0.0,bootindex=-1,scsi-id=1,lun=0,share-rw=on -device scsi-block,drive=idisk0,bus=scsi0.0,bootindex=-1,scsi-id=1,lun=1,share-rw=on Then in the guest, both /dev/sda and /dev/sdb were aggregated by multipath to /dev/mpatha, which was passed to a nested guest like that -object pr-manager-helper,id=qemu_pr_helper,path=/root/work/vm/testvm/.run/pr_helper.socket -blockdev node-name=test,driver=host_device,filename=/dev/mapper/mpatha,pr-manager=qemu_pr_helper -device scsi-block,drive=test,bus=scsi0.0,bootindex=-1,scsi-id=0,lun=0 The nested guest run: sg_persist --no-inquiry -v --out --register --param-sark 0x1234 /dev/sda Strictly speaking this is wrong configuration since qemu is where the multipath was split, and thus the iscsi target was not aware of multipath, and thus when libmpathpersist code rightfully tried to register the PR key on all paths, it failed to do so. However qemu-pr-helper should not crash in this case. Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2019-09-24 01:38:48 +02:00
if (ret != GOOD) {
return ret;
}
scsi_build_sense(sense, mpath_generic_sense(r));
return CHECK_CONDITION;
}
case MPATH_PR_SENSE_UNIT_ATTENTION:
/* Congratulations libmpathpersist, you ruined the Unit Attention...
* Return a heavyweight one.
*/
scsi_build_sense(sense, SENSE_CODE(SCSI_BUS_RESET));
return CHECK_CONDITION;
case MPATH_PR_SENSE_INVALID_OP:
/* Only one valid sense. */
scsi_build_sense(sense, SENSE_CODE(INVALID_OPCODE));
return CHECK_CONDITION;
case MPATH_PR_ILLEGAL_REQ:
/* Guess. */
scsi_build_sense(sense, SENSE_CODE(INVALID_PARAM));
return CHECK_CONDITION;
case MPATH_PR_NO_SENSE:
scsi_build_sense(sense, SENSE_CODE(NO_SENSE));
return CHECK_CONDITION;
case MPATH_PR_RESERV_CONFLICT:
return RESERVATION_CONFLICT;
case MPATH_PR_OTHER:
default:
scsi_build_sense(sense, SENSE_CODE(LUN_COMM_FAILURE));
return CHECK_CONDITION;
}
}
static int coroutine_fn multipath_pr_in(int fd, const uint8_t *cdb, uint8_t *sense,
uint8_t *data, int sz)
{
int rq_servact = cdb[1];
struct prin_resp resp;
size_t written;
int r;
switch (rq_servact) {
case MPATH_PRIN_RKEY_SA:
case MPATH_PRIN_RRES_SA:
case MPATH_PRIN_RCAP_SA:
break;
case MPATH_PRIN_RFSTAT_SA:
/* Nobody implements it anyway, so bail out. */
default:
/* Cannot parse any other output. */
scsi_build_sense(sense, SENSE_CODE(INVALID_FIELD));
return CHECK_CONDITION;
}
r = mpath_persistent_reserve_in(fd, rq_servact, &resp, noisy, verbose);
if (r == MPATH_PR_SUCCESS) {
switch (rq_servact) {
case MPATH_PRIN_RKEY_SA:
case MPATH_PRIN_RRES_SA: {
struct prin_readdescr *out = &resp.prin_descriptor.prin_readkeys;
assert(sz >= 8);
written = MIN(out->additional_length + 8, sz);
stl_be_p(&data[0], out->prgeneration);
stl_be_p(&data[4], out->additional_length);
memcpy(&data[8], out->key_list, written - 8);
break;
}
case MPATH_PRIN_RCAP_SA: {
struct prin_capdescr *out = &resp.prin_descriptor.prin_readcap;
assert(sz >= 6);
written = 6;
stw_be_p(&data[0], out->length);
data[2] = out->flags[0];
data[3] = out->flags[1];
stw_be_p(&data[4], out->pr_type_mask);
break;
}
default:
scsi_build_sense(sense, SENSE_CODE(INVALID_OPCODE));
return CHECK_CONDITION;
}
assert(written <= sz);
memset(data + written, 0, sz - written);
}
return mpath_reconstruct_sense(fd, r, sense);
}
static int coroutine_fn multipath_pr_out(int fd, const uint8_t *cdb, uint8_t *sense,
const uint8_t *param, int sz)
{
int rq_servact = cdb[1];
int rq_scope = cdb[2] >> 4;
int rq_type = cdb[2] & 0xf;
g_autofree struct prout_param_descriptor *paramp = NULL;
char transportids[PR_HELPER_DATA_SIZE];
int r;
paramp = g_malloc0(sizeof(struct prout_param_descriptor)
+ sizeof(struct transportid *) * MPATH_MX_TIDS);
if (sz < PR_OUT_FIXED_PARAM_SIZE) {
/* Illegal request, Parameter list length error. This isn't fatal;
* we have read the data, send an error without closing the socket.
*/
scsi_build_sense(sense, SENSE_CODE(INVALID_PARAM_LEN));
return CHECK_CONDITION;
}
switch (rq_servact) {
case MPATH_PROUT_REG_SA:
case MPATH_PROUT_RES_SA:
case MPATH_PROUT_REL_SA:
case MPATH_PROUT_CLEAR_SA:
case MPATH_PROUT_PREE_SA:
case MPATH_PROUT_PREE_AB_SA:
case MPATH_PROUT_REG_IGN_SA:
break;
case MPATH_PROUT_REG_MOV_SA:
/* Not supported by struct prout_param_descriptor. */
default:
/* Cannot parse any other input. */
scsi_build_sense(sense, SENSE_CODE(INVALID_FIELD));
return CHECK_CONDITION;
}
/* Convert input data, especially transport IDs, to the structs
* used by libmpathpersist (which, of course, will immediately
* do the opposite).
*/
memcpy(&paramp->key, &param[0], 8);
memcpy(&paramp->sa_key, &param[8], 8);
paramp->sa_flags = param[20];
if (sz > PR_OUT_FIXED_PARAM_SIZE) {
size_t transportid_len;
int i, j;
if (sz < PR_OUT_FIXED_PARAM_SIZE + 4) {
scsi_build_sense(sense, SENSE_CODE(INVALID_PARAM_LEN));
return CHECK_CONDITION;
}
transportid_len = ldl_be_p(&param[24]) + PR_OUT_FIXED_PARAM_SIZE + 4;
if (transportid_len > sz) {
scsi_build_sense(sense, SENSE_CODE(INVALID_PARAM));
return CHECK_CONDITION;
}
for (i = PR_OUT_FIXED_PARAM_SIZE + 4, j = 0; i < transportid_len; ) {
struct transportid *id = (struct transportid *) &transportids[j];
int len;
id->format_code = param[i] & 0xc0;
id->protocol_id = param[i] & 0x0f;
switch (param[i] & 0xcf) {
case 0:
/* FC transport. */
if (i + 24 > transportid_len) {
goto illegal_req;
}
memcpy(id->n_port_name, &param[i + 8], 8);
j += offsetof(struct transportid, n_port_name[8]);
i += 24;
break;
case 5:
case 0x45:
/* iSCSI transport. */
len = lduw_be_p(&param[i + 2]);
if (len > 252 || (len & 3) || i + len + 4 > transportid_len) {
/* For format code 00, the standard says the maximum is 223
* plus the NUL terminator. For format code 01 there is no
* maximum length, but libmpathpersist ignores the first
* byte of id->iscsi_name so our maximum is 252.
*/
goto illegal_req;
}
if (memchr(&param[i + 4], 0, len) == NULL) {
goto illegal_req;
}
memcpy(id->iscsi_name, &param[i + 2], len + 2);
j += offsetof(struct transportid, iscsi_name[len + 2]);
i += len + 4;
break;
case 6:
/* SAS transport. */
if (i + 24 > transportid_len) {
goto illegal_req;
}
memcpy(id->sas_address, &param[i + 4], 8);
j += offsetof(struct transportid, sas_address[8]);
i += 24;
break;
default:
illegal_req:
scsi_build_sense(sense, SENSE_CODE(INVALID_PARAM));
return CHECK_CONDITION;
}
assert(paramp->num_transportid < MPATH_MX_TIDS);
paramp->trnptid_list[paramp->num_transportid++] = id;
}
}
r = mpath_persistent_reserve_out(fd, rq_servact, rq_scope, rq_type,
paramp, noisy, verbose);
return mpath_reconstruct_sense(fd, r, sense);
}
#endif
static int coroutine_fn do_pr_in(int fd, const uint8_t *cdb, uint8_t *sense,
uint8_t *data, int *resp_sz)
{
#ifdef CONFIG_MPATH
if (is_mpath(fd)) {
/* multipath_pr_in fills the whole input buffer. */
int r = multipath_pr_in(fd, cdb, sense, data, *resp_sz);
if (r != GOOD) {
*resp_sz = 0;
}
return r;
}
#endif
return do_sgio(fd, cdb, sense, data, resp_sz,
SG_DXFER_FROM_DEV);
}
static int coroutine_fn do_pr_out(int fd, const uint8_t *cdb, uint8_t *sense,
const uint8_t *param, int sz)
{
int resp_sz;
if ((fcntl(fd, F_GETFL) & O_ACCMODE) == O_RDONLY) {
scsi_build_sense(sense, SENSE_CODE(INVALID_OPCODE));
return CHECK_CONDITION;
}
#ifdef CONFIG_MPATH
if (is_mpath(fd)) {
return multipath_pr_out(fd, cdb, sense, param, sz);
}
#endif
resp_sz = sz;
return do_sgio(fd, cdb, sense, (uint8_t *)param, &resp_sz,
SG_DXFER_TO_DEV);
}
/* Client */
typedef struct PRHelperClient {
QIOChannelSocket *ioc;
Coroutine *co;
int fd;
uint8_t data[PR_HELPER_DATA_SIZE];
} PRHelperClient;
typedef struct PRHelperRequest {
int fd;
size_t sz;
uint8_t cdb[PR_HELPER_CDB_SIZE];
} PRHelperRequest;
static int coroutine_fn prh_read(PRHelperClient *client, void *buf, int sz,
Error **errp)
{
int ret = 0;
while (sz > 0) {
int *fds = NULL;
size_t nfds = 0;
int i;
struct iovec iov;
ssize_t n_read;
iov.iov_base = buf;
iov.iov_len = sz;
n_read = qio_channel_readv_full(QIO_CHANNEL(client->ioc), &iov, 1,
&fds, &nfds, 0, errp);
if (n_read == QIO_CHANNEL_ERR_BLOCK) {
qio_channel_yield(QIO_CHANNEL(client->ioc), G_IO_IN);
continue;
}
if (n_read <= 0) {
ret = n_read ? n_read : -1;
goto err;
}
/* Stash one file descriptor per request. */
if (nfds) {
bool too_many = false;
for (i = 0; i < nfds; i++) {
if (client->fd == -1) {
client->fd = fds[i];
} else {
close(fds[i]);
too_many = true;
}
}
g_free(fds);
if (too_many) {
ret = -1;
goto err;
}
}
buf += n_read;
sz -= n_read;
}
return 0;
err:
if (client->fd != -1) {
close(client->fd);
client->fd = -1;
}
return ret;
}
static int coroutine_fn prh_read_request(PRHelperClient *client,
PRHelperRequest *req,
PRHelperResponse *resp, Error **errp)
{
uint32_t sz;
if (prh_read(client, req->cdb, sizeof(req->cdb), NULL) < 0) {
return -1;
}
if (client->fd == -1) {
error_setg(errp, "No file descriptor in request.");
return -1;
}
if (req->cdb[0] != PERSISTENT_RESERVE_OUT &&
req->cdb[0] != PERSISTENT_RESERVE_IN) {
error_setg(errp, "Invalid CDB, closing socket.");
goto out_close;
}
sz = scsi_cdb_xfer(req->cdb);
if (sz > sizeof(client->data)) {
goto out_close;
}
if (req->cdb[0] == PERSISTENT_RESERVE_OUT) {
if (qio_channel_read_all(QIO_CHANNEL(client->ioc),
(char *)client->data, sz,
errp) < 0) {
goto out_close;
}
}
req->fd = client->fd;
req->sz = sz;
client->fd = -1;
return sz;
out_close:
close(client->fd);
client->fd = -1;
return -1;
}
static int coroutine_fn prh_write_response(PRHelperClient *client,
PRHelperRequest *req,
PRHelperResponse *resp, Error **errp)
{
ssize_t r;
size_t sz;
if (req->cdb[0] == PERSISTENT_RESERVE_IN && resp->result == GOOD) {
assert(resp->sz <= req->sz && resp->sz <= sizeof(client->data));
} else {
assert(resp->sz == 0);
}
sz = resp->sz;
resp->result = cpu_to_be32(resp->result);
resp->sz = cpu_to_be32(resp->sz);
r = qio_channel_write_all(QIO_CHANNEL(client->ioc),
(char *) resp, sizeof(*resp), errp);
if (r < 0) {
return r;
}
r = qio_channel_write_all(QIO_CHANNEL(client->ioc),
(char *) client->data,
sz, errp);
return r < 0 ? r : 0;
}
static void coroutine_fn prh_co_entry(void *opaque)
{
PRHelperClient *client = opaque;
Error *local_err = NULL;
uint32_t flags;
int r;
qio_channel_set_blocking(QIO_CHANNEL(client->ioc),
false, NULL);
io: follow coroutine AioContext in qio_channel_yield() The ongoing QEMU multi-queue block layer effort makes it possible for multiple threads to process I/O in parallel. The nbd block driver is not compatible with the multi-queue block layer yet because QIOChannel cannot be used easily from coroutines running in multiple threads. This series changes the QIOChannel API to make that possible. In the current API, calling qio_channel_attach_aio_context() sets the AioContext where qio_channel_yield() installs an fd handler prior to yielding: qio_channel_attach_aio_context(ioc, my_ctx); ... qio_channel_yield(ioc); // my_ctx is used here ... qio_channel_detach_aio_context(ioc); This API design has limitations: reading and writing must be done in the same AioContext and moving between AioContexts involves a cumbersome sequence of API calls that is not suitable for doing on a per-request basis. There is no fundamental reason why a QIOChannel needs to run within the same AioContext every time qio_channel_yield() is called. QIOChannel only uses the AioContext while inside qio_channel_yield(). The rest of the time, QIOChannel is independent of any AioContext. In the new API, qio_channel_yield() queries the AioContext from the current coroutine using qemu_coroutine_get_aio_context(). There is no need to explicitly attach/detach AioContexts anymore and qio_channel_attach_aio_context() and qio_channel_detach_aio_context() are gone. One coroutine can read from the QIOChannel while another coroutine writes from a different AioContext. This API change allows the nbd block driver to use QIOChannel from any thread. It's important to keep in mind that the block driver already synchronizes QIOChannel access and ensures that two coroutines never read simultaneously or write simultaneously. This patch updates all users of qio_channel_attach_aio_context() to the new API. Most conversions are simple, but vhost-user-server requires a new qemu_coroutine_yield() call to quiesce the vu_client_trip() coroutine when not attached to any AioContext. While the API is has become simpler, there is one wart: QIOChannel has a special case for the iohandler AioContext (used for handlers that must not run in nested event loops). I didn't find an elegant way preserve that behavior, so I added a new API called qio_channel_set_follow_coroutine_ctx(ioc, true|false) for opting in to the new AioContext model. By default QIOChannel uses the iohandler AioHandler. Code that formerly called qio_channel_attach_aio_context() now calls qio_channel_set_follow_coroutine_ctx(ioc, true) once after the QIOChannel is created. Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Acked-by: Daniel P. Berrangé <berrange@redhat.com> Message-ID: <20230830224802.493686-5-stefanha@redhat.com> [eblake: also fix migration/rdma.c] Signed-off-by: Eric Blake <eblake@redhat.com>
2023-08-31 00:48:02 +02:00
qio_channel_set_follow_coroutine_ctx(QIO_CHANNEL(client->ioc), true);
/* A very simple negotiation for future extensibility. No features
* are defined so write 0.
*/
flags = cpu_to_be32(0);
r = qio_channel_write_all(QIO_CHANNEL(client->ioc),
(char *) &flags, sizeof(flags), NULL);
if (r < 0) {
goto out;
}
r = qio_channel_read_all(QIO_CHANNEL(client->ioc),
(char *) &flags, sizeof(flags), NULL);
if (be32_to_cpu(flags) != 0 || r < 0) {
goto out;
}
while (qatomic_read(&state) == RUNNING) {
PRHelperRequest req;
PRHelperResponse resp;
int sz;
sz = prh_read_request(client, &req, &resp, &local_err);
if (sz < 0) {
break;
}
num_active_sockets++;
if (req.cdb[0] == PERSISTENT_RESERVE_OUT) {
r = do_pr_out(req.fd, req.cdb, resp.sense,
client->data, sz);
resp.sz = 0;
} else {
resp.sz = sizeof(client->data);
r = do_pr_in(req.fd, req.cdb, resp.sense,
client->data, &resp.sz);
resp.sz = MIN(resp.sz, sz);
}
num_active_sockets--;
close(req.fd);
if (r == -1) {
break;
}
resp.result = r;
if (prh_write_response(client, &req, &resp, &local_err) < 0) {
break;
}
}
if (local_err) {
if (verbose == 0) {
error_free(local_err);
} else {
error_report_err(local_err);
}
}
out:
object_unref(OBJECT(client->ioc));
g_free(client);
}
static gboolean accept_client(QIOChannel *ioc, GIOCondition cond, gpointer opaque)
{
QIOChannelSocket *cioc;
PRHelperClient *prh;
cioc = qio_channel_socket_accept(QIO_CHANNEL_SOCKET(ioc),
NULL);
if (!cioc) {
return TRUE;
}
prh = g_new(PRHelperClient, 1);
prh->ioc = cioc;
prh->fd = -1;
prh->co = qemu_coroutine_create(prh_co_entry, prh);
qemu_coroutine_enter(prh->co);
return TRUE;
}
static void termsig_handler(int signum)
{
qatomic_cmpxchg(&state, RUNNING, TERMINATE);
qemu_notify_event();
}
static void close_server_socket(void)
{
assert(server_ioc);
g_source_remove(server_watch);
server_watch = -1;
object_unref(OBJECT(server_ioc));
num_active_sockets--;
}
#ifdef CONFIG_LIBCAP_NG
static int drop_privileges(void)
{
/* clear all capabilities */
capng_clear(CAPNG_SELECT_BOTH);
if (capng_update(CAPNG_ADD, CAPNG_EFFECTIVE | CAPNG_PERMITTED,
CAP_SYS_RAWIO) < 0) {
return -1;
}
#ifdef CONFIG_MPATH
/* For /dev/mapper/control ioctls */
if (capng_update(CAPNG_ADD, CAPNG_EFFECTIVE | CAPNG_PERMITTED,
CAP_SYS_ADMIN) < 0) {
return -1;
}
#endif
/* Change user/group id, retaining the capabilities. Because file descriptors
* are passed via SCM_RIGHTS, we don't need supplementary groups (and in
* fact the helper can run as "nobody").
*/
if (capng_change_id(uid != -1 ? uid : getuid(),
gid != -1 ? gid : getgid(),
CAPNG_DROP_SUPP_GRP | CAPNG_CLEAR_BOUNDING)) {
return -1;
}
return 0;
}
#endif
int main(int argc, char **argv)
{
const char *sopt = "hVk:f:dT:u:g:vq";
struct option lopt[] = {
{ "help", no_argument, NULL, 'h' },
{ "version", no_argument, NULL, 'V' },
{ "socket", required_argument, NULL, 'k' },
{ "pidfile", required_argument, NULL, 'f' },
{ "daemon", no_argument, NULL, 'd' },
{ "trace", required_argument, NULL, 'T' },
{ "user", required_argument, NULL, 'u' },
{ "group", required_argument, NULL, 'g' },
{ "verbose", no_argument, NULL, 'v' },
{ "quiet", no_argument, NULL, 'q' },
{ NULL, 0, NULL, 0 }
};
int opt_ind = 0;
int loglevel = 1;
int quiet = 0;
int ch;
Error *local_err = NULL;
bool daemonize = false;
bool pidfile_specified = false;
bool socket_path_specified = false;
unsigned socket_activation;
struct sigaction sa_sigterm;
memset(&sa_sigterm, 0, sizeof(sa_sigterm));
sa_sigterm.sa_handler = termsig_handler;
sigaction(SIGTERM, &sa_sigterm, NULL);
sigaction(SIGINT, &sa_sigterm, NULL);
sigaction(SIGHUP, &sa_sigterm, NULL);
signal(SIGPIPE, SIG_IGN);
error_init(argv[0]);
module_call_init(MODULE_INIT_TRACE);
module_call_init(MODULE_INIT_QOM);
qemu_add_opts(&qemu_trace_opts);
qemu_init_exec_dir(argv[0]);
compute_default_paths();
while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) {
switch (ch) {
case 'k':
g_free(socket_path);
socket_path = g_strdup(optarg);
socket_path_specified = true;
if (socket_path[0] != '/') {
error_report("socket path must be absolute");
exit(EXIT_FAILURE);
}
break;
case 'f':
g_free(pidfile);
pidfile = g_strdup(optarg);
pidfile_specified = true;
break;
#ifdef CONFIG_LIBCAP_NG
case 'u': {
unsigned long res;
struct passwd *userinfo = getpwnam(optarg);
if (userinfo) {
uid = userinfo->pw_uid;
} else if (qemu_strtoul(optarg, NULL, 10, &res) == 0 &&
(uid_t)res == res) {
uid = res;
} else {
error_report("invalid user '%s'", optarg);
exit(EXIT_FAILURE);
}
break;
}
case 'g': {
unsigned long res;
struct group *groupinfo = getgrnam(optarg);
if (groupinfo) {
gid = groupinfo->gr_gid;
} else if (qemu_strtoul(optarg, NULL, 10, &res) == 0 &&
(gid_t)res == res) {
gid = res;
} else {
error_report("invalid group '%s'", optarg);
exit(EXIT_FAILURE);
}
break;
}
#else
case 'u':
case 'g':
error_report("-%c not supported by this %s", ch, argv[0]);
exit(1);
#endif
case 'd':
daemonize = true;
break;
case 'q':
quiet = 1;
break;
case 'v':
++loglevel;
break;
case 'T':
trace_opt_parse(optarg);
break;
case 'V':
version(argv[0]);
exit(EXIT_SUCCESS);
break;
case 'h':
usage(argv[0]);
exit(EXIT_SUCCESS);
break;
case '?':
error_report("Try `%s --help' for more information.", argv[0]);
exit(EXIT_FAILURE);
}
}
/* set verbosity */
noisy = !quiet && (loglevel >= 3);
verbose = quiet ? 0 : MIN(loglevel, 3);
if (!trace_init_backends()) {
exit(EXIT_FAILURE);
}
trace_init_file();
qemu_set_log(LOG_TRACE, &error_fatal);
#ifdef CONFIG_MPATH
dm_init();
multipath_pr_init();
#endif
socket_activation = check_socket_activation();
if (socket_activation == 0) {
SocketAddress saddr;
saddr = (SocketAddress){
.type = SOCKET_ADDRESS_TYPE_UNIX,
.u.q_unix.path = socket_path,
};
server_ioc = qio_channel_socket_new();
if (qio_channel_socket_listen_sync(server_ioc, &saddr,
1, &local_err) < 0) {
object_unref(OBJECT(server_ioc));
error_report_err(local_err);
return 1;
}
} else {
/* Using socket activation - check user didn't use -p etc. */
if (socket_path_specified) {
error_report("Unix socket can't be set when using socket activation");
exit(EXIT_FAILURE);
}
/* Can only listen on a single socket. */
if (socket_activation > 1) {
error_report("%s does not support socket activation with LISTEN_FDS > 1",
argv[0]);
exit(EXIT_FAILURE);
}
server_ioc = qio_channel_socket_new_fd(FIRST_SOCKET_ACTIVATION_FD,
&local_err);
if (server_ioc == NULL) {
error_reportf_err(local_err,
"Failed to use socket activation: ");
exit(EXIT_FAILURE);
}
}
qemu_init_main_loop(&error_fatal);
server_watch = qio_channel_add_watch(QIO_CHANNEL(server_ioc),
G_IO_IN,
accept_client,
NULL, NULL);
if (daemonize) {
if (daemon(0, 0) < 0) {
error_report("Failed to daemonize: %s", strerror(errno));
exit(EXIT_FAILURE);
}
}
if (daemonize || pidfile_specified) {
qemu_write_pidfile(pidfile, &error_fatal);
util: add qemu_write_pidfile() There are variants of qemu_create_pidfile() in qemu-pr-helper and qemu-ga. Let's have a common implementation in libqemuutil. The code is initially based from pr-helper write_pidfile(), with various improvements and suggestions from Daniel Berrangé: QEMU will leave the pidfile existing on disk when it exits which initially made me think it avoids the deletion race. The app managing QEMU, however, may well delete the pidfile after it has seen QEMU exit, and even if the app locks the pidfile before deleting it, there is still a race. eg consider the following sequence QEMU 1 libvirtd QEMU 2 1. lock(pidfile) 2. exit() 3. open(pidfile) 4. lock(pidfile) 5. open(pidfile) 6. unlink(pidfile) 7. close(pidfile) 8. lock(pidfile) IOW, at step 8 the new QEMU has successfully acquired the lock, but the pidfile no longer exists on disk because it was deleted after the original QEMU exited. While we could just say no external app should ever delete the pidfile, I don't think that is satisfactory as people don't read docs, and admins don't like stale pidfiles being left around on disk. To make this robust, I think we might want to copy libvirt's approach to pidfile acquisition which runs in a loop and checks that the file on disk /after/ acquiring the lock matches the file that was locked. Then we could in fact safely let QEMU delete its own pidfiles on clean exit.. Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com> Message-Id: <20180831145314.14736-2-marcandre.lureau@redhat.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2018-08-31 16:53:12 +02:00
}
#ifdef CONFIG_LIBCAP_NG
if (drop_privileges() < 0) {
error_report("Failed to drop privileges: %s", strerror(errno));
exit(EXIT_FAILURE);
}
#endif
state = RUNNING;
do {
main_loop_wait(false);
if (state == TERMINATE) {
state = TERMINATING;
close_server_socket();
}
} while (num_active_sockets > 0);
exit(EXIT_SUCCESS);
}