leandrof/linux
1
0
Fork 0
forked from Minki/linux
Code Issues Pull Requests Packages Projects Releases Wiki Activity
699a8c7c4b
linux/drivers/scsi/qedf/qedf_fip.c
Saurav Kashyap b29a907f10 scsi: qedf: Initiator fails to re-login to switch after link down 
Problem Statement:

 - Driver has fc_id of 0xcc0200

 - Driver gets link down (due to test) and calls fcoe_ctlr_link_down().

 - At this point, the fc_id of the initiator port is zeroed out.

 - Driver gets a link up 14 seconds later.

 - Driver performs FIP VLAN request, gets a response from the switch.

 - No change in VLAN is detected.

 - Driver then notifies libfcoe via fcoe_ctlr_link_up().

 - Libfcoe then issues a multicast discovery solicitation as expected.

 - Cisco FCF responds to that correctly.

 - Libfcoe at this point starts a 3 sec count-down to allow any other FCFs
   to be discovered. However, at this point, it has been 20 seconds since
   the last FKA from the driver (which would have been sent prior to
   backlink toggle), which causes the CVL to be issued from Cisco CVL from
   the switch is dropped by the driver as the vx_port identification
   descriptor is present and has value of 0xcc0200, which does not match
   the driver's value of 0.  Libfcoe completes the 3 sec count down and
   proceeds to issue FLOGI as per protocol.  Switch rejects FLogi request.
   All subsequent FLOGI requests from libfc are rejected by the switch
   (possibly because it is now expecting a new solicitation).  This
   situation will continue until the next link toggle.

Solution:

The Vx_port descriptor in the CVL has three fields:

MAC address
Fabric ID
Port Name

Today, the code checks for both #1 and #2 above. In the case where we went
through a link down, both these will be zero until FLOGI succeeds.

We should change our code to check if any one of these 3 is valid and if
so, handle the CVL (basically switching from AND to OR). The port name
field is definitely expected to be valid always.

Signed-off-by: Saurav Kashyap <skashyap@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
2019-08-29 18:51:19 -04:00

302 lines
8.4 KiB
C
Raw Blame History

// SPDX-License-Identifier: GPL-2.0-only
/*
* QLogic FCoE Offload Driver
* Copyright (c) 2016-2018 Cavium Inc.
*/
#include <linux/if_ether.h>
#include <linux/if_vlan.h>
#include "qedf.h"
extern const struct qed_fcoe_ops *qed_ops;
/*
* FIP VLAN functions that will eventually move to libfcoe.
*/
void qedf_fcoe_send_vlan_req(struct qedf_ctx *qedf)
{
struct sk_buff *skb;
char *eth_fr;
struct fip_vlan *vlan;
#define MY_FIP_ALL_FCF_MACS ((__u8[6]) { 1, 0x10, 0x18, 1, 0, 2 })
static u8 my_fcoe_all_fcfs[ETH_ALEN] = MY_FIP_ALL_FCF_MACS;
unsigned long flags = 0;
int rc = -1;
skb = dev_alloc_skb(sizeof(struct fip_vlan));
if (!skb) {
QEDF_ERR(&qedf->dbg_ctx,
"Failed to allocate skb.\n");
return;
}
eth_fr = (char *)skb->data;
vlan = (struct fip_vlan *)eth_fr;
memset(vlan, 0, sizeof(*vlan));
ether_addr_copy(vlan->eth.h_source, qedf->mac);
ether_addr_copy(vlan->eth.h_dest, my_fcoe_all_fcfs);
vlan->eth.h_proto = htons(ETH_P_FIP);
vlan->fip.fip_ver = FIP_VER_ENCAPS(FIP_VER);
vlan->fip.fip_op = htons(FIP_OP_VLAN);
vlan->fip.fip_subcode = FIP_SC_VL_REQ;
vlan->fip.fip_dl_len = htons(sizeof(vlan->desc) / FIP_BPW);
vlan->desc.mac.fd_desc.fip_dtype = FIP_DT_MAC;
vlan->desc.mac.fd_desc.fip_dlen = sizeof(vlan->desc.mac) / FIP_BPW;
ether_addr_copy(vlan->desc.mac.fd_mac, qedf->mac);
vlan->desc.wwnn.fd_desc.fip_dtype = FIP_DT_NAME;
vlan->desc.wwnn.fd_desc.fip_dlen = sizeof(vlan->desc.wwnn) / FIP_BPW;
put_unaligned_be64(qedf->lport->wwnn, &vlan->desc.wwnn.fd_wwn);
skb_put(skb, sizeof(*vlan));
skb->protocol = htons(ETH_P_FIP);
skb_reset_mac_header(skb);
skb_reset_network_header(skb);
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_DISC, "Sending FIP VLAN "
"request.");
if (atomic_read(&qedf->link_state) != QEDF_LINK_UP) {
QEDF_WARN(&(qedf->dbg_ctx), "Cannot send vlan request "
"because link is not up.\n");
kfree_skb(skb);
return;
}
set_bit(QED_LL2_XMIT_FLAGS_FIP_DISCOVERY, &flags);
rc = qed_ops->ll2->start_xmit(qedf->cdev, skb, flags);
if (rc) {
QEDF_ERR(&qedf->dbg_ctx, "start_xmit failed rc = %d.\n", rc);
kfree_skb(skb);
return;
}
}
static void qedf_fcoe_process_vlan_resp(struct qedf_ctx *qedf,
struct sk_buff *skb)
{
struct fip_header *fiph;
struct fip_desc *desc;
u16 vid = 0;
ssize_t rlen;
size_t dlen;
fiph = (struct fip_header *)(((void *)skb->data) + 2 * ETH_ALEN + 2);
rlen = ntohs(fiph->fip_dl_len) * 4;
desc = (struct fip_desc *)(fiph + 1);
while (rlen > 0) {
dlen = desc->fip_dlen * FIP_BPW;
switch (desc->fip_dtype) {
case FIP_DT_VLAN:
vid = ntohs(((struct fip_vlan_desc *)desc)->fd_vlan);
break;
}
desc = (struct fip_desc *)((char *)desc + dlen);
rlen -= dlen;
}
if (atomic_read(&qedf->link_state) == QEDF_LINK_DOWN) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_DISC,
"Dropping VLAN response as link is down.\n");
return;
}
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_DISC, "VLAN response, "
"vid=0x%x.\n", vid);
if (vid > 0 && qedf->vlan_id != vid) {
qedf_set_vlan_id(qedf, vid);
/* Inform waiter that it's ok to call fcoe_ctlr_link up() */
if (!completion_done(&qedf->fipvlan_compl))
complete(&qedf->fipvlan_compl);
}
}
void qedf_fip_send(struct fcoe_ctlr *fip, struct sk_buff *skb)
{
struct qedf_ctx *qedf = container_of(fip, struct qedf_ctx, ctlr);
struct ethhdr *eth_hdr;
struct fip_header *fiph;
u16 op, vlan_tci = 0;
u8 sub;
int rc = -1;
if (!test_bit(QEDF_LL2_STARTED, &qedf->flags)) {
QEDF_WARN(&(qedf->dbg_ctx), "LL2 not started\n");
kfree_skb(skb);
return;
}
fiph = (struct fip_header *) ((void *)skb->data + 2 * ETH_ALEN + 2);
eth_hdr = (struct ethhdr *)skb_mac_header(skb);
op = ntohs(fiph->fip_op);
sub = fiph->fip_subcode;
/*
* Add VLAN tag to non-offload FIP frame based on current stored VLAN
* for FIP/FCoE traffic.
*/
__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), qedf->vlan_id);
/* Get VLAN ID from skb for printing purposes */
__vlan_hwaccel_get_tag(skb, &vlan_tci);
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_LL2, "FIP frame send: "
"dest=%pM op=%x sub=%x vlan=%04x.", eth_hdr->h_dest, op, sub,
vlan_tci);
if (qedf_dump_frames)
print_hex_dump(KERN_WARNING, "fip ", DUMP_PREFIX_OFFSET, 16, 1,
skb->data, skb->len, false);
rc = qed_ops->ll2->start_xmit(qedf->cdev, skb, 0);
if (rc) {
QEDF_ERR(&qedf->dbg_ctx, "start_xmit failed rc = %d.\n", rc);
kfree_skb(skb);
return;
}
}
static u8 fcoe_all_enode[ETH_ALEN] = FIP_ALL_ENODE_MACS;
/* Process incoming FIP frames. */
void qedf_fip_recv(struct qedf_ctx *qedf, struct sk_buff *skb)
{
struct ethhdr *eth_hdr;
struct fip_header *fiph;
struct fip_desc *desc;
struct fip_mac_desc *mp;
struct fip_wwn_desc *wp;
struct fip_vn_desc *vp;
size_t rlen, dlen;
u16 op;
u8 sub;
bool fcf_valid = false;
/* Default is to handle CVL regardless of fabric id descriptor */
bool fabric_id_valid = true;
bool fc_wwpn_valid = false;
u64 switch_name;
u16 vlan = 0;
eth_hdr = (struct ethhdr *)skb_mac_header(skb);
fiph = (struct fip_header *) ((void *)skb->data + 2 * ETH_ALEN + 2);
op = ntohs(fiph->fip_op);
sub = fiph->fip_subcode;
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_LL2,
"FIP frame received: skb=%p fiph=%p source=%pM destn=%pM op=%x sub=%x vlan=%04x",
skb, fiph, eth_hdr->h_source, eth_hdr->h_dest, op,
sub, vlan);
if (qedf_dump_frames)
print_hex_dump(KERN_WARNING, "fip ", DUMP_PREFIX_OFFSET, 16, 1,
skb->data, skb->len, false);
if (!ether_addr_equal(eth_hdr->h_dest, qedf->mac) &&
!ether_addr_equal(eth_hdr->h_dest, fcoe_all_enode) &&
!ether_addr_equal(eth_hdr->h_dest, qedf->data_src_addr)) {
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_LL2,
"Dropping FIP type 0x%x pkt due to destination MAC mismatch dest_mac=%pM ctlr.dest_addr=%pM data_src_addr=%pM.\n",
op, eth_hdr->h_dest, qedf->mac,
qedf->data_src_addr);
kfree_skb(skb);
return;
}
/* Handle FIP VLAN resp in the driver */
if (op == FIP_OP_VLAN && sub == FIP_SC_VL_NOTE) {
qedf_fcoe_process_vlan_resp(qedf, skb);
kfree_skb(skb);
} else if (op == FIP_OP_CTRL && sub == FIP_SC_CLR_VLINK) {
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_DISC, "Clear virtual "
"link received.\n");
/* Check that an FCF has been selected by fcoe */
if (qedf->ctlr.sel_fcf == NULL) {
QEDF_INFO(&(qedf->dbg_ctx), QEDF_LOG_DISC,
"Dropping CVL since FCF has not been selected "
"yet.");
kfree_skb(skb);
return;
}
/*
* We need to loop through the CVL descriptors to determine
* if we want to reset the fcoe link
*/
rlen = ntohs(fiph->fip_dl_len) * FIP_BPW;
desc = (struct fip_desc *)(fiph + 1);
while (rlen >= sizeof(*desc)) {
dlen = desc->fip_dlen * FIP_BPW;
switch (desc->fip_dtype) {
case FIP_DT_MAC:
mp = (struct fip_mac_desc *)desc;
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_DISC,
"Switch fd_mac=%pM.\n", mp->fd_mac);
if (ether_addr_equal(mp->fd_mac,
qedf->ctlr.sel_fcf->fcf_mac))
fcf_valid = true;
break;
case FIP_DT_NAME:
wp = (struct fip_wwn_desc *)desc;
switch_name = get_unaligned_be64(&wp->fd_wwn);
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_DISC,
"Switch fd_wwn=%016llx fcf_switch_name=%016llx.\n",
switch_name,
qedf->ctlr.sel_fcf->switch_name);
if (switch_name ==
qedf->ctlr.sel_fcf->switch_name)
fc_wwpn_valid = true;
break;
case FIP_DT_VN_ID:
fabric_id_valid = false;
vp = (struct fip_vn_desc *)desc;
QEDF_ERR(&qedf->dbg_ctx,
"CVL vx_port fd_fc_id=0x%x fd_mac=%pM fd_wwpn=%016llx.\n",
ntoh24(vp->fd_fc_id), vp->fd_mac,
get_unaligned_be64(&vp->fd_wwpn));
/* Check for vx_port wwpn OR Check vx_port
* fabric ID OR Check vx_port MAC
*/
if ((get_unaligned_be64(&vp->fd_wwpn) ==
qedf->wwpn) ||
(ntoh24(vp->fd_fc_id) ==
qedf->lport->port_id) ||
(ether_addr_equal(vp->fd_mac,
qedf->data_src_addr))) {
fabric_id_valid = true;
}
break;
default:
/* Ignore anything else */
break;
}
desc = (struct fip_desc *)((char *)desc + dlen);
rlen -= dlen;
}
QEDF_INFO(&qedf->dbg_ctx, QEDF_LOG_DISC,
"fcf_valid=%d fabric_id_valid=%d fc_wwpn_valid=%d.\n",
fcf_valid, fabric_id_valid, fc_wwpn_valid);
if (fcf_valid && fabric_id_valid && fc_wwpn_valid)
qedf_ctx_soft_reset(qedf->lport);
kfree_skb(skb);
} else {
/* Everything else is handled by libfcoe */
__skb_pull(skb, ETH_HLEN);
fcoe_ctlr_recv(&qedf->ctlr, skb);
}
}
u8 *qedf_get_src_mac(struct fc_lport *lport)
{
struct qedf_ctx *qedf = lport_priv(lport);
return qedf->data_src_addr;
}
Reference in New Issue View Git Blame Copy Permalink