numam-spdk/lib/iscsi/iscsi.h

507 lines
15 KiB
C
Raw Normal View History

/*-
* BSD LICENSE
*
* Copyright (C) 2008-2012 Daisuke Aoyama <aoyama@peach.ne.jp>.
* Copyright (c) Intel Corporation.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
* * Neither the name of Intel Corporation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#ifndef SPDK_ISCSI_H
#define SPDK_ISCSI_H
#include "spdk/stdinc.h"
#include "spdk/env.h"
#include "spdk/bdev.h"
#include "spdk/iscsi_spec.h"
#include "spdk/thread.h"
#include "spdk/sock.h"
#include "spdk/scsi.h"
#include "iscsi/param.h"
#include "spdk/assert.h"
#include "spdk/dif.h"
#include "spdk/util.h"
#define SPDK_ISCSI_DEFAULT_NODEBASE "iqn.2016-06.io.spdk"
#define DEFAULT_MAXR2T 4
#define MAX_INITIATOR_PORT_NAME 256
#define MAX_INITIATOR_NAME 223
#define MAX_TARGET_NAME 223
#define MAX_PORTAL 1024
#define MAX_INITIATOR 256
#define MAX_NETMASK 256
#define MAX_ISCSI_CONNECTIONS 1024
#define MAX_PORTAL_ADDR 256
#define MAX_PORTAL_PORT 32
#define DEFAULT_PORT 3260
#define DEFAULT_MAX_SESSIONS 128
#define DEFAULT_MAX_CONNECTIONS_PER_SESSION 2
#define DEFAULT_MAXOUTSTANDINGR2T 1
#define DEFAULT_DEFAULTTIME2WAIT 2
#define DEFAULT_DEFAULTTIME2RETAIN 20
#define DEFAULT_INITIALR2T true
#define DEFAULT_IMMEDIATEDATA true
#define DEFAULT_DATAPDUINORDER true
#define DEFAULT_DATASEQUENCEINORDER true
#define DEFAULT_ERRORRECOVERYLEVEL 0
#define DEFAULT_TIMEOUT 60
#define MAX_NOPININTERVAL 60
#define DEFAULT_NOPININTERVAL 30
/*
* SPDK iSCSI target currently only supports 64KB as the maximum data segment length
* it can receive from initiators. Other values may work, but no guarantees.
*/
#define SPDK_ISCSI_MAX_RECV_DATA_SEGMENT_LENGTH 65536
/*
* Defines maximum number of data out buffers each connection can have in
* use at any given time.
*/
#define MAX_DATA_OUT_PER_CONNECTION 16
/*
lib/iscsi: Make the max number of read subtasks for large read I/O configurable For some use case that there is heavy large read I/O, the performance bottleneck due to MAX_LARGE_DATAIN_PER_CONNECTION was reported. The following assumes that all I/Os are large read. Large read primary task whose I/O size is more than SPDK_BDEV_LARGE_BUF_MAX_SIZE (=64KB) is split into multiple read subtasks. spdk_iscsi_globals::MaxQueueDepth limits maximum number of outstanding read primary tasks, and MAX_LARGE_DATAIN_PER_CONNECTION (=64) limits maximum number of outstanding read subtasks. MAX_LARGE_DATAIN_PER_CONNECTION is also used to calculate PDU pool. To remove the performance bottleneck, change the macro constant MAX_LARGE_DATAIN_PER_CONNECTION to a global variable spdk_iscsi_globals::MaxLargeDataInPerConnection. We don't see any negative side effect if we set spdk_iscsi_globals::MaxLargeDataInPerConnection to 64. The use case that reported the performance issue will change the value of spdk_iscsi_globals::MaxLargeDataInPerConnection by its own responsibility. The next patch will add the value of spdk_iscsi_globals::MaxLargeDataInPerConnection to iSCSI options, and make it configurable by JSON RPC. Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com> Change-Id: Ifc30cdb8e00d50f4d3755ff399263cf5d0b681b6 Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3755 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2020-08-11 15:29:25 +00:00
* Defines default maximum number of data in buffers each connection can have in
* use at any given time. So this limit does not affect I/O smaller than
* SPDK_BDEV_SMALL_BUF_MAX_SIZE.
*/
lib/iscsi: Make the max number of read subtasks for large read I/O configurable For some use case that there is heavy large read I/O, the performance bottleneck due to MAX_LARGE_DATAIN_PER_CONNECTION was reported. The following assumes that all I/Os are large read. Large read primary task whose I/O size is more than SPDK_BDEV_LARGE_BUF_MAX_SIZE (=64KB) is split into multiple read subtasks. spdk_iscsi_globals::MaxQueueDepth limits maximum number of outstanding read primary tasks, and MAX_LARGE_DATAIN_PER_CONNECTION (=64) limits maximum number of outstanding read subtasks. MAX_LARGE_DATAIN_PER_CONNECTION is also used to calculate PDU pool. To remove the performance bottleneck, change the macro constant MAX_LARGE_DATAIN_PER_CONNECTION to a global variable spdk_iscsi_globals::MaxLargeDataInPerConnection. We don't see any negative side effect if we set spdk_iscsi_globals::MaxLargeDataInPerConnection to 64. The use case that reported the performance issue will change the value of spdk_iscsi_globals::MaxLargeDataInPerConnection by its own responsibility. The next patch will add the value of spdk_iscsi_globals::MaxLargeDataInPerConnection to iSCSI options, and make it configurable by JSON RPC. Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com> Change-Id: Ifc30cdb8e00d50f4d3755ff399263cf5d0b681b6 Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3755 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2020-08-11 15:29:25 +00:00
#define DEFAULT_MAX_LARGE_DATAIN_PER_CONNECTION 64
#define SPDK_ISCSI_MAX_BURST_LENGTH \
(SPDK_ISCSI_MAX_RECV_DATA_SEGMENT_LENGTH * MAX_DATA_OUT_PER_CONNECTION)
/*
* Defines default maximum amount in bytes of unsolicited data the iSCSI
* initiator may send to the SPDK iSCSI target during the execution of
* a single SCSI command. And it is smaller than the MaxBurstLength.
*/
#define SPDK_ISCSI_FIRST_BURST_LENGTH 8192
/*
* Defines minimum amount in bytes of unsolicited data the iSCSI initiator
* may send to the SPDK iSCSI target during the execution of a single
* SCSI command.
*/
#define SPDK_ISCSI_MIN_FIRST_BURST_LENGTH 512
#define SPDK_ISCSI_MAX_FIRST_BURST_LENGTH 16777215
/*
* Defines default maximum queue depth per connection and this can be
* changed by configuration file.
*/
#define DEFAULT_MAX_QUEUE_DEPTH 64
/** Defines how long we should wait for a logout request when the target
* requests logout to the initiator asynchronously.
*/
#define ISCSI_LOGOUT_REQUEST_TIMEOUT 30 /* in seconds */
/** Defines how long we should wait for a TCP close after responding to a
* logout request, before terminating the connection ourselves.
*/
#define ISCSI_LOGOUT_TIMEOUT 5 /* in seconds */
/** Defines how long we should wait until login process completes. */
#define ISCSI_LOGIN_TIMEOUT 30 /* in seconds */
/* For spdk_iscsi_login_in related function use, we need to avoid the conflict
* with other errors
* */
#define SPDK_ISCSI_LOGIN_ERROR_RESPONSE -1000
#define SPDK_ISCSI_LOGIN_ERROR_PARAMETER -1001
#define SPDK_ISCSI_PARAMETER_EXCHANGE_NOT_ONCE -1002
#define ISCSI_AHS_LEN 60
struct spdk_mobj {
struct spdk_mempool *mp;
void *buf;
uint32_t data_len;
};
/*
* Maximum number of SGL elements, i.e.,
* BHS, AHS, Header Digest, Data Segment and Data Digest.
*/
#define SPDK_ISCSI_MAX_SGL_DESCRIPTORS (5)
#define SPDK_CRC32C_INITIAL 0xffffffffUL
#define SPDK_CRC32C_XOR 0xffffffffUL
typedef void (*iscsi_conn_xfer_complete_cb)(void *cb_arg);
struct spdk_iscsi_pdu {
struct iscsi_bhs bhs;
lib/iscsi: Merge multiple Data-OUT PDUs up to 64KB in a sequence Some iSCSI initiators send a Data-OUT PDU sequence whose PDUs do not have block size multiples data. SPDK iSCSI target had replied SCSI write error to such initiators because previously we had sent a write subtask per Data-OUT PDU. SPDK SCSI library had rejected the write subtask because its data was not block size multiples. This patch fixes the issue. The idea is to aggregate multiple Data-OUT PDUs into a single write subtask up to 64KB or until F bit is set. MaxRecvDataSegmentLength is 64KB but MaxBurstLength is 1MB. Hence one Data-OUT PDU data may be split into multiple data buffers, but the maximum number of split is two. When processing the data segment of the Data-OUT PDU, save the data buffer of the current PDU to the current task if the data buffer is not full and F bit is not set. In this case, write subtask is not submitted. When processing the header of the Data-OUT PDU, if the current task saves the data buffer from the last Data-OUT PDU, it passes the data buffer to the Data-OUT PDU. When reading the data segment of the current PDU, attach the second data buffer to the current PDU if the first data buffer becomes full. These are enabled only if DIF is disabled. Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com> Change-Id: Ib9cfb53fe8c0807a63e58c61bed3bb52f60f4830 Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6439 Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Community-CI: Mellanox Build Bot
2021-02-22 04:14:49 +00:00
/* Merge multiple Data-OUT PDUs in a sequence into a subtask up to
* SPDK_ISCSI_MAX_RECV_DATA_SEGMENT_LENGTH or the final PDU comes.
*
* Both the size of a data buffer and MaxRecvDataSegmentLength are
* SPDK_ISCSI_MAX_RECV_DATA_SEGMENT_LENGTH at most. Hence the data segment of
* a Data-OUT PDU can be split into two data buffers at most.
*/
struct spdk_mobj *mobj[2];
bool is_rejected;
uint8_t *data;
uint8_t header_digest[ISCSI_DIGEST_LEN];
uint8_t data_digest[ISCSI_DIGEST_LEN];
size_t data_segment_len;
int bhs_valid_bytes;
int ahs_valid_bytes;
uint32_t data_valid_bytes;
int hdigest_valid_bytes;
int ddigest_valid_bytes;
int ref;
bool data_from_mempool; /* indicate whether the data buffer is allocated from mempool */
struct spdk_iscsi_task *task; /* data tied to a task buffer */
uint32_t cmd_sn;
uint32_t writev_offset;
uint32_t data_buf_len;
lib/iscsi: Merge multiple Data-OUT PDUs up to 64KB in a sequence Some iSCSI initiators send a Data-OUT PDU sequence whose PDUs do not have block size multiples data. SPDK iSCSI target had replied SCSI write error to such initiators because previously we had sent a write subtask per Data-OUT PDU. SPDK SCSI library had rejected the write subtask because its data was not block size multiples. This patch fixes the issue. The idea is to aggregate multiple Data-OUT PDUs into a single write subtask up to 64KB or until F bit is set. MaxRecvDataSegmentLength is 64KB but MaxBurstLength is 1MB. Hence one Data-OUT PDU data may be split into multiple data buffers, but the maximum number of split is two. When processing the data segment of the Data-OUT PDU, save the data buffer of the current PDU to the current task if the data buffer is not full and F bit is not set. In this case, write subtask is not submitted. When processing the header of the Data-OUT PDU, if the current task saves the data buffer from the last Data-OUT PDU, it passes the data buffer to the Data-OUT PDU. When reading the data segment of the current PDU, attach the second data buffer to the current PDU if the first data buffer becomes full. These are enabled only if DIF is disabled. Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com> Change-Id: Ib9cfb53fe8c0807a63e58c61bed3bb52f60f4830 Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6439 Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Community-CI: Mellanox Build Bot
2021-02-22 04:14:49 +00:00
uint32_t data_offset;
uint32_t crc32c;
bool dif_insert_or_strip;
struct spdk_dif_ctx dif_ctx;
struct spdk_iscsi_conn *conn;
iscsi_conn_xfer_complete_cb cb_fn;
void *cb_arg;
/* The sock request ends with a 0 length iovec. Place the actual iovec immediately
* after it. There is a static assert below to check if the compiler inserted
* any unwanted padding */
int32_t mapped_length;
struct spdk_sock_request sock_req;
struct iovec iov[SPDK_ISCSI_MAX_SGL_DESCRIPTORS];
TAILQ_ENTRY(spdk_iscsi_pdu) tailq;
/*
* 60 bytes of AHS should suffice for now.
* This should always be at the end of PDU data structure.
* we need to not zero this out when doing memory clear.
*/
uint8_t ahs[ISCSI_AHS_LEN];
struct {
uint16_t length; /* iSCSI SenseLength (big-endian) */
uint8_t data[32];
} sense;
};
SPDK_STATIC_ASSERT(offsetof(struct spdk_iscsi_pdu,
sock_req) + sizeof(struct spdk_sock_request) == offsetof(struct spdk_iscsi_pdu, iov),
"Compiler inserted padding between iov and sock_req");
enum iscsi_connection_state {
ISCSI_CONN_STATE_INVALID = 0,
ISCSI_CONN_STATE_RUNNING = 1,
ISCSI_CONN_STATE_EXITING = 2,
ISCSI_CONN_STATE_EXITED = 3,
};
enum iscsi_chap_phase {
ISCSI_CHAP_PHASE_NONE = 0,
ISCSI_CHAP_PHASE_WAIT_A = 1,
ISCSI_CHAP_PHASE_WAIT_NR = 2,
ISCSI_CHAP_PHASE_END = 3,
};
enum session_type {
SESSION_TYPE_INVALID = 0,
SESSION_TYPE_NORMAL = 1,
SESSION_TYPE_DISCOVERY = 2,
};
iscsi: Load CHAP secrets from file once at boot and use them in memory For secure iSCSI targets, dynamic reconfiguration of CHAP secrets is must to have. Currently CHAP secrets are loaded for every CHAP authentication operation. The current implementation will not work correctly when CHAP secrets in the file are changed dynamically. If SPDK loads CHAP secrets from the file only at boot and they can be configured by RPCs, user can change CHAP secrets safely during run time. Even if there are any users who expect dynamic reconfiguration of CHAP secrets based on the current implementation, if we provide this better alternative based on RPCs, they will be able to continue to satisfy their requirement. This patch change the current implementation so that SPDK loads CHAP secrets from the file once at boot and uses them in memory hereafter. Besides, use fixed size buffers to hold CHAP secrets. Previously dynamically allocated buffers by strdup() had been used, but it required many nomem checks and should be avoided. Other iSCSI targets/initiators have used fixed size buffers and SPDK follows others. Set the size of buffers for both user names and secrets to 256 (the last byte is for NULL termination). 256 is sufficiently large compared with others. CHANGELOG will be updated in the separate patch because new RPCs will be added instead. Change-Id: I499e792817c2ed01c3d970bbd3d34a6b1fccf65b Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com> Reviewed-on: https://review.gerrithub.io/421463 Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-20 01:32:30 +00:00
#define ISCSI_CHAP_CHALLENGE_LEN 1024
#define ISCSI_CHAP_MAX_USER_LEN 255
#define ISCSI_CHAP_MAX_SECRET_LEN 255
struct iscsi_chap_auth {
enum iscsi_chap_phase chap_phase;
iscsi: Load CHAP secrets from file once at boot and use them in memory For secure iSCSI targets, dynamic reconfiguration of CHAP secrets is must to have. Currently CHAP secrets are loaded for every CHAP authentication operation. The current implementation will not work correctly when CHAP secrets in the file are changed dynamically. If SPDK loads CHAP secrets from the file only at boot and they can be configured by RPCs, user can change CHAP secrets safely during run time. Even if there are any users who expect dynamic reconfiguration of CHAP secrets based on the current implementation, if we provide this better alternative based on RPCs, they will be able to continue to satisfy their requirement. This patch change the current implementation so that SPDK loads CHAP secrets from the file once at boot and uses them in memory hereafter. Besides, use fixed size buffers to hold CHAP secrets. Previously dynamically allocated buffers by strdup() had been used, but it required many nomem checks and should be avoided. Other iSCSI targets/initiators have used fixed size buffers and SPDK follows others. Set the size of buffers for both user names and secrets to 256 (the last byte is for NULL termination). 256 is sufficiently large compared with others. CHANGELOG will be updated in the separate patch because new RPCs will be added instead. Change-Id: I499e792817c2ed01c3d970bbd3d34a6b1fccf65b Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com> Reviewed-on: https://review.gerrithub.io/421463 Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-20 01:32:30 +00:00
char user[ISCSI_CHAP_MAX_USER_LEN + 1];
char secret[ISCSI_CHAP_MAX_SECRET_LEN + 1];
char muser[ISCSI_CHAP_MAX_USER_LEN + 1];
char msecret[ISCSI_CHAP_MAX_SECRET_LEN + 1];
uint8_t chap_id[1];
uint8_t chap_mid[1];
int chap_challenge_len;
uint8_t chap_challenge[ISCSI_CHAP_CHALLENGE_LEN];
int chap_mchallenge_len;
uint8_t chap_mchallenge[ISCSI_CHAP_CHALLENGE_LEN];
};
iscsi: Load CHAP secrets from file once at boot and use them in memory For secure iSCSI targets, dynamic reconfiguration of CHAP secrets is must to have. Currently CHAP secrets are loaded for every CHAP authentication operation. The current implementation will not work correctly when CHAP secrets in the file are changed dynamically. If SPDK loads CHAP secrets from the file only at boot and they can be configured by RPCs, user can change CHAP secrets safely during run time. Even if there are any users who expect dynamic reconfiguration of CHAP secrets based on the current implementation, if we provide this better alternative based on RPCs, they will be able to continue to satisfy their requirement. This patch change the current implementation so that SPDK loads CHAP secrets from the file once at boot and uses them in memory hereafter. Besides, use fixed size buffers to hold CHAP secrets. Previously dynamically allocated buffers by strdup() had been used, but it required many nomem checks and should be avoided. Other iSCSI targets/initiators have used fixed size buffers and SPDK follows others. Set the size of buffers for both user names and secrets to 256 (the last byte is for NULL termination). 256 is sufficiently large compared with others. CHANGELOG will be updated in the separate patch because new RPCs will be added instead. Change-Id: I499e792817c2ed01c3d970bbd3d34a6b1fccf65b Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com> Reviewed-on: https://review.gerrithub.io/421463 Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-20 01:32:30 +00:00
struct spdk_iscsi_auth_secret {
char user[ISCSI_CHAP_MAX_USER_LEN + 1];
char secret[ISCSI_CHAP_MAX_SECRET_LEN + 1];
char muser[ISCSI_CHAP_MAX_USER_LEN + 1];
char msecret[ISCSI_CHAP_MAX_SECRET_LEN + 1];
TAILQ_ENTRY(spdk_iscsi_auth_secret) tailq;
};
struct spdk_iscsi_auth_group {
int32_t tag;
TAILQ_HEAD(, spdk_iscsi_auth_secret) secret_head;
TAILQ_ENTRY(spdk_iscsi_auth_group) tailq;
};
struct spdk_iscsi_sess {
uint32_t connections;
struct spdk_iscsi_conn **conns;
struct spdk_scsi_port *initiator_port;
int tag;
uint64_t isid;
uint16_t tsih;
struct spdk_iscsi_tgt_node *target;
int queue_depth;
struct iscsi_param *params;
enum session_type session_type;
uint32_t MaxConnections;
uint32_t MaxOutstandingR2T;
uint32_t DefaultTime2Wait;
uint32_t DefaultTime2Retain;
uint32_t FirstBurstLength;
uint32_t MaxBurstLength;
bool InitialR2T;
bool ImmediateData;
bool DataPDUInOrder;
bool DataSequenceInOrder;
uint32_t ErrorRecoveryLevel;
uint32_t ExpCmdSN;
uint32_t MaxCmdSN;
uint32_t current_text_itt;
};
struct spdk_iscsi_poll_group {
struct spdk_poller *poller;
struct spdk_poller *nop_poller;
STAILQ_HEAD(connections, spdk_iscsi_conn) connections;
struct spdk_sock_group *sock_group;
iscsi: Assign connections to poll groups instead of lcores. This patch binds poll groups to SPDK thread through IO channel and assigns connections to poll groups instead of cores. iSCSI subsystem registers iSCSI global object as an IO device, and create poll groups as context of IO channels of the IO device. Each portal get and hold portal group on which the corresponding acceptor is running. When a connection is constructed, iSCSI subsystem assigns a poll group to the connection by getting it from the corresponding portal. When a connection enters full-feature phase, iSCSI subsystem schedules the connection to a poll group by round-robin. Then, each connection can know its running SPDK thread directly and can use SPDK message passing infrastructure instead of SPDK event framework. By this change, iSCSI connections are binded to SPDK thread, and not binded to processor core anymore. Some other changes in this patch are - core ID is removed from the output of get_iscsi_connections RPC. The upcoming patches will change the RPC to use spdk_for_each_channel and can access SPDK thread safely, and add SPDK thread ID instead. - utilize UT multithread framework added by the last patch to test iSCSI poll groups by UT. Change-Id: Iec73c778aa413bcabdb63141cc41d4160911ea0e Signed-off-by: Ben Walker <benjamin.walker@intel.com> Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463359 Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
2019-07-30 03:00:15 +00:00
TAILQ_ENTRY(spdk_iscsi_poll_group) link;
};
struct spdk_iscsi_opts {
char *authfile;
char *nodebase;
int32_t timeout;
int32_t nopininterval;
bool disable_chap;
bool require_chap;
bool mutual_chap;
int32_t chap_group;
uint32_t MaxSessions;
uint32_t MaxConnectionsPerSession;
uint32_t MaxConnections;
uint32_t MaxQueueDepth;
uint32_t DefaultTime2Wait;
uint32_t DefaultTime2Retain;
uint32_t FirstBurstLength;
bool ImmediateData;
uint32_t ErrorRecoveryLevel;
bool AllowDuplicateIsid;
uint32_t MaxLargeDataInPerConnection;
uint32_t MaxR2TPerConnection;
uint32_t pdu_pool_size;
uint32_t immediate_data_pool_size;
uint32_t data_out_pool_size;
};
struct spdk_iscsi_globals {
char *authfile;
char *nodebase;
pthread_mutex_t mutex;
uint32_t refcnt;
TAILQ_HEAD(, spdk_iscsi_portal) portal_head;
TAILQ_HEAD(, spdk_iscsi_portal_grp) pg_head;
TAILQ_HEAD(, spdk_iscsi_init_grp) ig_head;
TAILQ_HEAD(, spdk_iscsi_tgt_node) target_head;
iscsi: Load CHAP secrets from file once at boot and use them in memory For secure iSCSI targets, dynamic reconfiguration of CHAP secrets is must to have. Currently CHAP secrets are loaded for every CHAP authentication operation. The current implementation will not work correctly when CHAP secrets in the file are changed dynamically. If SPDK loads CHAP secrets from the file only at boot and they can be configured by RPCs, user can change CHAP secrets safely during run time. Even if there are any users who expect dynamic reconfiguration of CHAP secrets based on the current implementation, if we provide this better alternative based on RPCs, they will be able to continue to satisfy their requirement. This patch change the current implementation so that SPDK loads CHAP secrets from the file once at boot and uses them in memory hereafter. Besides, use fixed size buffers to hold CHAP secrets. Previously dynamically allocated buffers by strdup() had been used, but it required many nomem checks and should be avoided. Other iSCSI targets/initiators have used fixed size buffers and SPDK follows others. Set the size of buffers for both user names and secrets to 256 (the last byte is for NULL termination). 256 is sufficiently large compared with others. CHANGELOG will be updated in the separate patch because new RPCs will be added instead. Change-Id: I499e792817c2ed01c3d970bbd3d34a6b1fccf65b Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com> Reviewed-on: https://review.gerrithub.io/421463 Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-20 01:32:30 +00:00
TAILQ_HEAD(, spdk_iscsi_auth_group) auth_group_head;
iscsi: Assign connections to poll groups instead of lcores. This patch binds poll groups to SPDK thread through IO channel and assigns connections to poll groups instead of cores. iSCSI subsystem registers iSCSI global object as an IO device, and create poll groups as context of IO channels of the IO device. Each portal get and hold portal group on which the corresponding acceptor is running. When a connection is constructed, iSCSI subsystem assigns a poll group to the connection by getting it from the corresponding portal. When a connection enters full-feature phase, iSCSI subsystem schedules the connection to a poll group by round-robin. Then, each connection can know its running SPDK thread directly and can use SPDK message passing infrastructure instead of SPDK event framework. By this change, iSCSI connections are binded to SPDK thread, and not binded to processor core anymore. Some other changes in this patch are - core ID is removed from the output of get_iscsi_connections RPC. The upcoming patches will change the RPC to use spdk_for_each_channel and can access SPDK thread safely, and add SPDK thread ID instead. - utilize UT multithread framework added by the last patch to test iSCSI poll groups by UT. Change-Id: Iec73c778aa413bcabdb63141cc41d4160911ea0e Signed-off-by: Ben Walker <benjamin.walker@intel.com> Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463359 Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
2019-07-30 03:00:15 +00:00
TAILQ_HEAD(, spdk_iscsi_poll_group) poll_group_head;
int32_t timeout;
int32_t nopininterval;
bool disable_chap;
bool require_chap;
bool mutual_chap;
int32_t chap_group;
uint32_t MaxSessions;
uint32_t MaxConnectionsPerSession;
uint32_t MaxConnections;
uint32_t MaxQueueDepth;
uint32_t DefaultTime2Wait;
uint32_t DefaultTime2Retain;
uint32_t FirstBurstLength;
bool ImmediateData;
uint32_t ErrorRecoveryLevel;
bool AllowDuplicateIsid;
lib/iscsi: Make the max number of read subtasks for large read I/O configurable For some use case that there is heavy large read I/O, the performance bottleneck due to MAX_LARGE_DATAIN_PER_CONNECTION was reported. The following assumes that all I/Os are large read. Large read primary task whose I/O size is more than SPDK_BDEV_LARGE_BUF_MAX_SIZE (=64KB) is split into multiple read subtasks. spdk_iscsi_globals::MaxQueueDepth limits maximum number of outstanding read primary tasks, and MAX_LARGE_DATAIN_PER_CONNECTION (=64) limits maximum number of outstanding read subtasks. MAX_LARGE_DATAIN_PER_CONNECTION is also used to calculate PDU pool. To remove the performance bottleneck, change the macro constant MAX_LARGE_DATAIN_PER_CONNECTION to a global variable spdk_iscsi_globals::MaxLargeDataInPerConnection. We don't see any negative side effect if we set spdk_iscsi_globals::MaxLargeDataInPerConnection to 64. The use case that reported the performance issue will change the value of spdk_iscsi_globals::MaxLargeDataInPerConnection by its own responsibility. The next patch will add the value of spdk_iscsi_globals::MaxLargeDataInPerConnection to iSCSI options, and make it configurable by JSON RPC. Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com> Change-Id: Ifc30cdb8e00d50f4d3755ff399263cf5d0b681b6 Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3755 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2020-08-11 15:29:25 +00:00
uint32_t MaxLargeDataInPerConnection;
uint32_t MaxR2TPerConnection;
uint32_t pdu_pool_size;
uint32_t immediate_data_pool_size;
uint32_t data_out_pool_size;
struct spdk_mempool *pdu_pool;
struct spdk_mempool *pdu_immediate_data_pool;
struct spdk_mempool *pdu_data_out_pool;
struct spdk_mempool *session_pool;
struct spdk_mempool *task_pool;
struct spdk_iscsi_sess **session;
};
#define ISCSI_SECURITY_NEGOTIATION_PHASE 0
#define ISCSI_OPERATIONAL_NEGOTIATION_PHASE 1
#define ISCSI_NSG_RESERVED_CODE 2
#define ISCSI_FULL_FEATURE_PHASE 3
/* logout reason */
#define ISCSI_LOGOUT_REASON_CLOSE_SESSION 0
#define ISCSI_LOGOUT_REASON_CLOSE_CONNECTION 1
#define ISCSI_LOGOUT_REASON_REMOVE_CONN_FOR_RECOVERY 2
enum spdk_error_codes {
SPDK_ISCSI_CONNECTION_FATAL = -1,
SPDK_PDU_FATAL = -2,
};
#define DGET24(B) \
((( (uint32_t) *((uint8_t *)(B)+0)) << 16) \
| (((uint32_t) *((uint8_t *)(B)+1)) << 8) \
| (((uint32_t) *((uint8_t *)(B)+2)) << 0))
#define DSET24(B,D) \
(((*((uint8_t *)(B)+0)) = (uint8_t)((uint32_t)(D) >> 16)), \
((*((uint8_t *)(B)+1)) = (uint8_t)((uint32_t)(D) >> 8)), \
((*((uint8_t *)(B)+2)) = (uint8_t)((uint32_t)(D) >> 0)))
#define xstrdup(s) (s ? strdup(s) : (char *)NULL)
extern struct spdk_iscsi_globals g_iscsi;
extern struct spdk_iscsi_opts *g_spdk_iscsi_opts;
struct spdk_iscsi_task;
struct spdk_json_write_ctx;
typedef void (*spdk_iscsi_init_cb)(void *cb_arg, int rc);
void spdk_iscsi_init(spdk_iscsi_init_cb cb_fn, void *cb_arg);
typedef void (*spdk_iscsi_fini_cb)(void *arg);
void spdk_iscsi_fini(spdk_iscsi_fini_cb cb_fn, void *cb_arg);
void shutdown_iscsi_conns_done(void);
void spdk_iscsi_config_json(struct spdk_json_write_ctx *w);
struct spdk_iscsi_opts *iscsi_opts_alloc(void);
void iscsi_opts_free(struct spdk_iscsi_opts *opts);
struct spdk_iscsi_opts *iscsi_opts_copy(struct spdk_iscsi_opts *src);
void iscsi_opts_info_json(struct spdk_json_write_ctx *w);
int iscsi_set_discovery_auth(bool disable_chap, bool require_chap,
bool mutual_chap, int32_t chap_group);
int iscsi_chap_get_authinfo(struct iscsi_chap_auth *auth, const char *authuser,
int ag_tag);
int iscsi_add_auth_group(int32_t tag, struct spdk_iscsi_auth_group **_group);
struct spdk_iscsi_auth_group *iscsi_find_auth_group_by_tag(int32_t tag);
void iscsi_delete_auth_group(struct spdk_iscsi_auth_group *group);
int iscsi_auth_group_add_secret(struct spdk_iscsi_auth_group *group,
const char *user, const char *secret,
const char *muser, const char *msecret);
int iscsi_auth_group_delete_secret(struct spdk_iscsi_auth_group *group,
const char *user);
void iscsi_auth_groups_info_json(struct spdk_json_write_ctx *w);
void iscsi_task_response(struct spdk_iscsi_conn *conn,
struct spdk_iscsi_task *task);
int iscsi_build_iovs(struct spdk_iscsi_conn *conn, struct iovec *iovs, int iovcnt,
struct spdk_iscsi_pdu *pdu, uint32_t *mapped_length);
int iscsi_handle_incoming_pdus(struct spdk_iscsi_conn *conn);
void iscsi_task_mgmt_response(struct spdk_iscsi_conn *conn,
struct spdk_iscsi_task *task);
void iscsi_free_sess(struct spdk_iscsi_sess *sess);
void iscsi_clear_all_transfer_task(struct spdk_iscsi_conn *conn,
struct spdk_scsi_lun *lun,
struct spdk_iscsi_pdu *pdu);
bool iscsi_del_transfer_task(struct spdk_iscsi_conn *conn, uint32_t CmdSN);
uint32_t iscsi_pdu_calc_header_digest(struct spdk_iscsi_pdu *pdu);
uint32_t iscsi_pdu_calc_data_digest(struct spdk_iscsi_pdu *pdu);
/* Memory management */
void iscsi_put_pdu(struct spdk_iscsi_pdu *pdu);
struct spdk_iscsi_pdu *iscsi_get_pdu(struct spdk_iscsi_conn *conn);
void iscsi_op_abort_task_set(struct spdk_iscsi_task *task,
uint8_t function);
void iscsi_queue_task(struct spdk_iscsi_conn *conn, struct spdk_iscsi_task *task);
static inline struct spdk_mobj *
iscsi_datapool_get(struct spdk_mempool *pool)
{
return spdk_mempool_get(pool);
}
static inline void
iscsi_datapool_put(struct spdk_mobj *mobj)
{
assert(mobj != NULL);
mobj->data_len = 0;
spdk_mempool_put(mobj->mp, (void *)mobj);
}
static inline uint32_t
iscsi_get_max_immediate_data_size(void)
{
/*
* Specify enough extra space in addition to FirstBurstLength to
* account for a header digest, data digest and additional header
* segments (AHS). These are not normally used but they do not
* take up much space and we need to make sure the worst-case scenario
* can be satisified by the size returned here.
*/
return g_iscsi.FirstBurstLength +
ISCSI_DIGEST_LEN + /* data digest */
ISCSI_DIGEST_LEN + /* header digest */
8 + /* bidirectional AHS */
52; /* extended CDB AHS (for a 64-byte CDB) */
}
#endif /* SPDK_ISCSI_H */