Fix kmalloc overflow in LPFC driver at large core count

This patch allows the LPFC to start up without a fatal kernel bug based
on an exceeded KMALLOC_MAX_SIZE and a too large NR_CPU-based maskbits
field. The bug was based on the number of CPU cores in a system.
Using the get_cpu_mask() function declared in kernel/cpu.c allows the
driver to load on the community kernel 4.2 RC1.

Below is the kernel bug reproduced:

8<--------------------------------------------------------------------
2199382.828437 (    0.005216)| lpfc 0003:02:00.0: enabling device (0140 -> 0142)
2199382.999272 (    0.170835)| ------------[ cut here ]------------
2199382.999337 (    0.000065)| WARNING: CPU: 84 PID: 404 at mm/slab_common.c:653 kmalloc_slab+0x2f/0x89()
2199383.004534 (    0.005197)| Modules linked in: lpfc(+) usbcore(+) mptctl scsi_transport_fc sg lpc_ich i2c_i801 usb_common tpm_tis mfd_core tpm acpi_cpufreq button scsi_dh_alua scsi_dh_rdacusbcore: registered new device driver usb
2199383.020568 (    0.016034)|
2199383.020581 (    0.000013)|  scsi_dh_hp_sw scsi_dh_emc scsi_dh gru thermal sata_nv processor piix fan thermal_sysehci_hcd: USB 2.0 'Enhanced' Host Controller (EHCI) Driver
2199383.035288 (    0.014707)|
2199383.035306 (    0.000018)|  hwmon ata_piix
2199383.035336 (    0.000030)| CPU: 84 PID: 404 Comm: kworker/84:0 Not tainted 3.18.0-rc2-gat-00106-ga7ca10f-dirty #178
2199383.047077 (    0.011741)| ehci-pci: EHCI PCI platform driver
2199383.047134 (    0.000057)| Hardware name: SGI UV2000/ROMLEY, BIOS SGI UV 2000/3000 series BIOS 01/15/2013
2199383.056245 (    0.009111)| Workqueue: events work_for_cpu_fn
2199383.066174 (    0.009929)|  000000000000028d ffff88eef827bbe8 ffffffff815a542f 000000000000028d
2199383.069545 (    0.003371)|  ffffffff810ea142 ffff88eef827bc28 ffffffff8104365c ffff88eefe4006c8
2199383.076214 (    0.006669)|  0000000000000000 00000000000080d0 0000000000000000 0000000000000004
2199383.079213 (    0.002999)| Call Trace:
2199383.084084 (    0.004871)|  [<ffffffff815a542f>] dump_stack+0x49/0x62
2199383.087283 (    0.003199)|  [<ffffffff810ea142>] ? kmalloc_slab+0x2f/0x89
2199383.091415 (    0.004132)|  [<ffffffff8104365c>] warn_slowpath_common+0x77/0x92
2199383.095197 (    0.003782)|  [<ffffffff8104368c>] warn_slowpath_null+0x15/0x17
2199383.103336 (    0.008139)|  [<ffffffff810ea142>] kmalloc_slab+0x2f/0x89
2199383.107082 (    0.003746)|  [<ffffffff8110fd9e>] __kmalloc+0x13/0x16a
2199383.112531 (    0.005449)|  [<ffffffffa01a8ed9>] lpfc_pci_probe_one_s4+0x105b/0x1644 [lpfc]
2199383.115316 (    0.002785)|  [<ffffffff81302b92>] ? pci_bus_read_config_dword+0x75/0x87
2199383.123431 (    0.008115)|  [<ffffffffa01a951f>] lpfc_pci_probe_one+0x5d/0xcb5 [lpfc]
2199383.127364 (    0.003933)|  [<ffffffff81497119>] ? dbs_check_cpu+0x168/0x177
2199383.136438 (    0.009074)|  [<ffffffff81496fa5>] ? gov_queue_work+0xb4/0xc0
2199383.140407 (    0.003969)|  [<ffffffff8130b2a1>] local_pci_probe+0x1e/0x52
2199383.143105 (    0.002698)|  [<ffffffff81052c47>] work_for_cpu_fn+0x13/0x1b
2199383.147315 (    0.004210)|  [<ffffffff81054965>] process_one_work+0x222/0x35e
2199383.151379 (    0.004064)|  [<ffffffff81054e76>] worker_thread+0x3d5/0x46e
2199383.159402 (    0.008023)|  [<ffffffff81054aa1>] ? process_one_work+0x35e/0x35e
2199383.163097 (    0.003695)|  [<ffffffff810599c6>] kthread+0xc8/0xd2
2199383.167476 (    0.004379)|  [<ffffffff810598fe>] ? kthread_freezable_should_stop+0x5b/0x5b
2199383.176434 (    0.008958)|  [<ffffffff815a8cac>] ret_from_fork+0x7c/0xb0
2199383.180086 (    0.003652)|  [<ffffffff810598fe>] ? kthread_freezable_should_stop+0x5b/0x5b
2199383.192333 (    0.012247)| ehci-pci 0000:00:1a.0: EHCI Host Controller
-------------------------------------------------------------------->8

The proposed solution was approved by James Smart at Emulex and tested
on a UV2 machine with 6144 cores. With the fix, the LPFC module loads
with no unwanted effects on the system.

Signed-off-by: Ian Mitchell <imitchell@sgi.com>
Signed-off-by: Alex Thorlton <athorlton@sgi.com>
Suggested-by: Robert Elliot <elliott@hp.com>
[james.smart: resolve unused variable warning]
Signed-off-by: James Smart <james.smart@avagotech.com>
Signed-off-by: James Bottomley <JBottomley@Odin.com>

drivers/scsi/lpfc/lpfc_init.c

@@ -8678,7 +8678,6 @@ lpfc_sli4_set_affinity(struct lpfc_hba *phba, int vectors)
 #ifdef CONFIG_X86
 	struct cpuinfo_x86 *cpuinfo;
 #endif
-	struct cpumask *mask;
 	uint8_t chann[LPFC_FCP_IO_CHAN_MAX+1];
 
 	/* If there is no mapping, just return */
@@ -8772,11 +8771,8 @@ lpfc_sli4_set_affinity(struct lpfc_hba *phba, int vectors)
 			first_cpu = cpu;
 
 		/* Now affinitize to the selected CPU */
-		mask = &cpup->maskbits;
-		cpumask_clear(mask);
-		cpumask_set_cpu(cpu, mask);
 		i = irq_set_affinity_hint(phba->sli4_hba.msix_entries[idx].
-					  vector, mask);
+					  vector, get_cpu_mask(cpu));
 
 		lpfc_printf_log(phba, KERN_INFO, LOG_INIT,
 				"3330 Set Affinity: CPU %d channel %d "

drivers/scsi/lpfc/lpfc_sli4.h

@@ -454,7 +454,6 @@ struct lpfc_vector_map_info {
 	uint16_t	core_id;
 	uint16_t	irq;
 	uint16_t	channel_id;
-	struct cpumask	maskbits;
 };
 #define LPFC_VECTOR_MAP_EMPTY	0xffff