diff options
author | Kiran Patil <kiran.patil@intel.com> | 2011-06-20 16:59:20 -0700 |
---|---|---|
committer | James Bottomley <JBottomley@Parallels.com> | 2011-06-29 16:29:36 -0500 |
commit | 064287eee372e8a2effe77fb909a40da9e7a1fd7 (patch) | |
tree | 0cfda86145d0bae9dedc70206c1ff1f4bb773480 /drivers | |
parent | 1ff9918b625457ce20d450d00f9ed0a12ba191b7 (diff) |
[SCSI] fcoe: Round-robin based selection of CPU for post-processing of incoming commands
Problem: Earlier mechanism of selection of CPU was, to select the same CPU
which has received incoming request. Hence in case of rx_id = 0xFFFF,
request was always posted to same NetRx queue, hence only 1 CPU is utilized
for handling the command. It was also causing problem of "running out of
exchanges from per CPU pool of exchanges (in case of DDP offload)
Fix: Implemented new algo. to select CPU for post-processing of
incoming commands when rx_id is unknown. This is simple Round robin
algo. for CPU selection.
Notes/Dependencies: N/A
Signed-off-by: Kiran Patil <kiran.patil@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Diffstat (limited to 'drivers')
-rw-r--r-- | drivers/scsi/fcoe/fcoe.c | 39 |
1 files changed, 38 insertions, 1 deletions
diff --git a/drivers/scsi/fcoe/fcoe.c b/drivers/scsi/fcoe/fcoe.c index 6378c5883514..da73115ddf4d 100644 --- a/drivers/scsi/fcoe/fcoe.c +++ b/drivers/scsi/fcoe/fcoe.c @@ -1245,6 +1245,36 @@ static int fcoe_cpu_callback(struct notifier_block *nfb, } /** + * fcoe_select_cpu() - Selects CPU to handle post-processing of incoming + * command. + * @curr_cpu: CPU which received request + * + * This routine selects next CPU based on cpumask. + * + * Returns: int (CPU number). Caller to verify if returned CPU is online or not. + */ +static unsigned int fcoe_select_cpu(unsigned int curr_cpu) +{ + static unsigned int selected_cpu; + + if (num_online_cpus() == 1) + return curr_cpu; + /* + * Doing following check, to skip "curr_cpu (smp_processor_id)" + * from selection of CPU is intentional. This is to avoid same CPU + * doing post-processing of command. "curr_cpu" to just receive + * incoming request in case where rx_id is UNKNOWN and all other + * CPU to actually process the command(s) + */ + do { + selected_cpu = cpumask_next(selected_cpu, cpu_online_mask); + if (selected_cpu >= nr_cpu_ids) + selected_cpu = cpumask_first(cpu_online_mask); + } while (selected_cpu == curr_cpu); + return selected_cpu; +} + +/** * fcoe_rcv() - Receive packets from a net device * @skb: The received packet * @netdev: The net device that the packet was received on @@ -1320,9 +1350,16 @@ int fcoe_rcv(struct sk_buff *skb, struct net_device *netdev, */ if (ntoh24(fh->fh_f_ctl) & FC_FC_EX_CTX) cpu = ntohs(fh->fh_ox_id) & fc_cpu_mask; - else + else { cpu = smp_processor_id(); + if ((fh->fh_type == FC_TYPE_FCP) && + (ntohs(fh->fh_rx_id) == FC_XID_UNKNOWN)) { + do { + cpu = fcoe_select_cpu(cpu); + } while (!cpu_online(cpu)); + } + } fps = &per_cpu(fcoe_percpu, cpu); spin_lock_bh(&fps->fcoe_rx_list.lock); if (unlikely(!fps->thread)) { |