kernel_optimize_test/drivers/scsi/iscsi_tcp.h

171 lines
5.1 KiB
C
Raw Normal View History

/*
* iSCSI Initiator TCP Transport
* Copyright (C) 2004 Dmitry Yusupov
* Copyright (C) 2004 Alex Aizman
* Copyright (C) 2005 - 2006 Mike Christie
* Copyright (C) 2006 Red Hat, Inc. All rights reserved.
* maintained by open-iscsi@googlegroups.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; either version 2 of the License, or
* (at your option) any later version.
*
* 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.
*
* See the file COPYING included with this distribution for more details.
*/
#ifndef ISCSI_TCP_H
#define ISCSI_TCP_H
#include <scsi/libiscsi.h>
/* Socket's Receive state machine */
#define IN_PROGRESS_WAIT_HEADER 0x0
#define IN_PROGRESS_HEADER_GATHER 0x1
#define IN_PROGRESS_DATA_RECV 0x2
#define IN_PROGRESS_DDIGEST_RECV 0x3
#define IN_PROGRESS_PAD_RECV 0x4
/* xmit state machine */
[SCSI] iscsi_tcp: fix potential lockup with write commands There is a race condition in iscsi_tcp.c that may cause it to forget that it received a R2T from the target. This race may cause a data-out command (such as a write) to lock up. The race occurs here: static int iscsi_send_unsol_pdu(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) { struct iscsi_tcp_cmd_task *tcp_ctask = ctask->dd_data; int rc; if (tcp_ctask->xmstate & XMSTATE_UNS_HDR) { BUG_ON(!ctask->unsol_count); tcp_ctask->xmstate &= ~XMSTATE_UNS_HDR; <---- RACE ... static int iscsi_r2t_rsp(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) { ... tcp_ctask->xmstate |= XMSTATE_SOL_HDR_INIT; <---- RACE ... While iscsi_xmitworker() (called from scsi_queue_work()) is preparing to send unsolicited data, iscsi_tcp_data_recv() (called from tcp_read_sock()) interrupts it upon receipt of a R2T from the target. Both contexts do read-modify-write of tcp_ctask->xmstate. Usually, gcc on x86 will make &= and |= atomic on UP (not guaranteed of course), but in this case iscsi_send_unsol_pdu() reads the value of xmstate before clearing the bit, which causes gcc to read xmstate into a CPU register, test it, clear the bit, and then store it back to memory. If the recv interrupt happens during this sequence, then the XMSTATE_SOL_HDR_INIT bit set by the recv interrupt will be lost, and the R2T will be forgotten. The patch below (against 2.6.24-rc1) converts accesses of xmstate to use set_bit, clear_bit, and test_bit instead of |= and &=. I have tested this patch and verified that it fixes the problem. Another possible approach would be to hold a lock during most of the rx/tx setup and post-processing, and drop the lock only for the actual rx/tx. Signed-off-by: Tony Battersby <tonyb@cybernetics.com> Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2007-11-15 04:38:42 +08:00
#define XMSTATE_VALUE_IDLE 0
#define XMSTATE_BIT_CMD_HDR_INIT 0
#define XMSTATE_BIT_CMD_HDR_XMIT 1
#define XMSTATE_BIT_IMM_HDR 2
#define XMSTATE_BIT_IMM_DATA 3
#define XMSTATE_BIT_UNS_INIT 4
#define XMSTATE_BIT_UNS_HDR 5
#define XMSTATE_BIT_UNS_DATA 6
#define XMSTATE_BIT_SOL_HDR 7
#define XMSTATE_BIT_SOL_DATA 8
#define XMSTATE_BIT_W_PAD 9
#define XMSTATE_BIT_W_RESEND_PAD 10
#define XMSTATE_BIT_W_RESEND_DATA_DIGEST 11
#define XMSTATE_BIT_IMM_HDR_INIT 12
#define XMSTATE_BIT_SOL_HDR_INIT 13
2006-09-01 06:09:27 +08:00
#define ISCSI_PAD_LEN 4
#define ISCSI_SG_TABLESIZE SG_ALL
#define ISCSI_TCP_MAX_CMD_LEN 16
struct crypto_hash;
struct socket;
/* Socket connection recieve helper */
struct iscsi_tcp_recv {
struct iscsi_hdr *hdr;
struct sk_buff *skb;
int offset;
int len;
int hdr_offset;
int copy;
int copied;
int padding;
struct iscsi_cmd_task *ctask; /* current cmd in progress */
/* copied and flipped values */
int datalen;
int datadgst;
char zero_copy_hdr;
};
struct iscsi_tcp_conn {
struct iscsi_conn *iscsi_conn;
struct socket *sock;
struct iscsi_hdr hdr; /* header placeholder */
char hdrext[4*sizeof(__u16) +
sizeof(__u32)];
int data_copied;
int stop_stage; /* conn_stop() flag: *
* stop to recover, *
* stop to terminate */
/* iSCSI connection-wide sequencing */
int hdr_size; /* PDU header size */
/* control data */
struct iscsi_tcp_recv in; /* TCP receive context */
int in_progress; /* connection state machine */
/* old values for socket callbacks */
void (*old_data_ready)(struct sock *, int);
void (*old_state_change)(struct sock *);
void (*old_write_space)(struct sock *);
/* data and header digests */
struct hash_desc tx_hash; /* CRC32C (Tx) */
struct hash_desc rx_hash; /* CRC32C (Rx) */
/* MIB custom statistics */
uint32_t sendpage_failures_cnt;
uint32_t discontiguous_hdr_cnt;
ssize_t (*sendpage)(struct socket *, struct page *, int, size_t, int);
};
struct iscsi_buf {
struct scatterlist sg;
unsigned int sent;
char use_sendmsg;
};
struct iscsi_data_task {
struct iscsi_data hdr; /* PDU */
char hdrext[sizeof(__u32)]; /* Header-Digest */
struct iscsi_buf digestbuf; /* digest buffer */
uint32_t digest; /* data digest */
};
struct iscsi_tcp_mgmt_task {
struct iscsi_hdr hdr;
char hdrext[sizeof(__u32)]; /* Header-Digest */
[SCSI] iscsi_tcp: fix potential lockup with write commands There is a race condition in iscsi_tcp.c that may cause it to forget that it received a R2T from the target. This race may cause a data-out command (such as a write) to lock up. The race occurs here: static int iscsi_send_unsol_pdu(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) { struct iscsi_tcp_cmd_task *tcp_ctask = ctask->dd_data; int rc; if (tcp_ctask->xmstate & XMSTATE_UNS_HDR) { BUG_ON(!ctask->unsol_count); tcp_ctask->xmstate &= ~XMSTATE_UNS_HDR; <---- RACE ... static int iscsi_r2t_rsp(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) { ... tcp_ctask->xmstate |= XMSTATE_SOL_HDR_INIT; <---- RACE ... While iscsi_xmitworker() (called from scsi_queue_work()) is preparing to send unsolicited data, iscsi_tcp_data_recv() (called from tcp_read_sock()) interrupts it upon receipt of a R2T from the target. Both contexts do read-modify-write of tcp_ctask->xmstate. Usually, gcc on x86 will make &= and |= atomic on UP (not guaranteed of course), but in this case iscsi_send_unsol_pdu() reads the value of xmstate before clearing the bit, which causes gcc to read xmstate into a CPU register, test it, clear the bit, and then store it back to memory. If the recv interrupt happens during this sequence, then the XMSTATE_SOL_HDR_INIT bit set by the recv interrupt will be lost, and the R2T will be forgotten. The patch below (against 2.6.24-rc1) converts accesses of xmstate to use set_bit, clear_bit, and test_bit instead of |= and &=. I have tested this patch and verified that it fixes the problem. Another possible approach would be to hold a lock during most of the rx/tx setup and post-processing, and drop the lock only for the actual rx/tx. Signed-off-by: Tony Battersby <tonyb@cybernetics.com> Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2007-11-15 04:38:42 +08:00
unsigned long xmstate; /* mgmt xmit progress */
struct iscsi_buf headbuf; /* header buffer */
struct iscsi_buf sendbuf; /* in progress buffer */
int sent;
};
struct iscsi_r2t_info {
__be32 ttt; /* copied from R2T */
__be32 exp_statsn; /* copied from R2T */
uint32_t data_length; /* copied from R2T */
uint32_t data_offset; /* copied from R2T */
struct iscsi_buf headbuf; /* Data-Out Header Buffer */
struct iscsi_buf sendbuf; /* Data-Out in progress buffer*/
int sent; /* R2T sequence progress */
int data_count; /* DATA-Out payload progress */
struct scatterlist *sg; /* per-R2T SG list */
int solicit_datasn;
struct iscsi_data_task dtask; /* which data task */
};
struct iscsi_tcp_cmd_task {
struct iscsi_cmd hdr;
char hdrext[4*sizeof(__u16)+ /* AHS */
sizeof(__u32)]; /* HeaderDigest */
char pad[ISCSI_PAD_LEN];
int pad_count; /* padded bytes */
struct iscsi_buf headbuf; /* header buf (xmit) */
struct iscsi_buf sendbuf; /* in progress buffer*/
[SCSI] iscsi_tcp: fix potential lockup with write commands There is a race condition in iscsi_tcp.c that may cause it to forget that it received a R2T from the target. This race may cause a data-out command (such as a write) to lock up. The race occurs here: static int iscsi_send_unsol_pdu(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) { struct iscsi_tcp_cmd_task *tcp_ctask = ctask->dd_data; int rc; if (tcp_ctask->xmstate & XMSTATE_UNS_HDR) { BUG_ON(!ctask->unsol_count); tcp_ctask->xmstate &= ~XMSTATE_UNS_HDR; <---- RACE ... static int iscsi_r2t_rsp(struct iscsi_conn *conn, struct iscsi_cmd_task *ctask) { ... tcp_ctask->xmstate |= XMSTATE_SOL_HDR_INIT; <---- RACE ... While iscsi_xmitworker() (called from scsi_queue_work()) is preparing to send unsolicited data, iscsi_tcp_data_recv() (called from tcp_read_sock()) interrupts it upon receipt of a R2T from the target. Both contexts do read-modify-write of tcp_ctask->xmstate. Usually, gcc on x86 will make &= and |= atomic on UP (not guaranteed of course), but in this case iscsi_send_unsol_pdu() reads the value of xmstate before clearing the bit, which causes gcc to read xmstate into a CPU register, test it, clear the bit, and then store it back to memory. If the recv interrupt happens during this sequence, then the XMSTATE_SOL_HDR_INIT bit set by the recv interrupt will be lost, and the R2T will be forgotten. The patch below (against 2.6.24-rc1) converts accesses of xmstate to use set_bit, clear_bit, and test_bit instead of |= and &=. I have tested this patch and verified that it fixes the problem. Another possible approach would be to hold a lock during most of the rx/tx setup and post-processing, and drop the lock only for the actual rx/tx. Signed-off-by: Tony Battersby <tonyb@cybernetics.com> Signed-off-by: Mike Christie <michaelc@cs.wisc.edu> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
2007-11-15 04:38:42 +08:00
unsigned long xmstate; /* xmit xtate machine */
int sent;
struct scatterlist *sg; /* per-cmd SG list */
struct scatterlist *bad_sg; /* assert statement */
int sg_count; /* SG's to process */
uint32_t exp_datasn; /* expected target's R2TSN/DataSN */
int data_offset;
struct iscsi_r2t_info *r2t; /* in progress R2T */
struct iscsi_queue r2tpool;
struct kfifo *r2tqueue;
struct iscsi_r2t_info **r2ts;
int digest_count;
uint32_t immdigest; /* for imm data */
struct iscsi_buf immbuf; /* for imm data digest */
struct iscsi_data_task unsol_dtask; /* unsol data task */
};
#endif /* ISCSI_H */