Re: Commit 911af505 introduced a bootmem warning.

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Wed, May 15, 2013 at 02:57:42AM -0500, Robin Holt wrote:
> Paul,
> 
> When we boot Linus' current kernel we get the following warning early
> in boot:
> 
> ----------------------------------------------------------------------
> [    0.000000] Memory: 63081268k/99598336k available (4832k kernel code, 34651396k absent, 1865672k reserved, 6269k data,
> 1672k init)
> [    0.000000] Hierarchical RCU implementation.
> [    0.000000]  RCU dyntick-idle grace-period acceleration is enabled.
> [    0.000000]  RCU restricting CPUs from NR_CPUS=4096 to nr_cpu_ids=32.
> [    0.000000] ------------[ cut here ]------------
> [    0.000000] WARNING: at mm/nobootmem.c:215 ___alloc_bootmem_nopanic+0x79/0x82()
> [    0.000000] Modules linked in:
> [    0.000000] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 3.10.0-rc1-uv-hz100-rja+ #3
> [    0.000000] Hardware name: SGI UV2000/ROMLEY, BIOS SGI UV 2000/3000 series BIOS 01/15/2013
> [    0.000000]  00000000000000d7 ffffffff81a01e38 ffffffff814a873c ffffffff81a01e78
> [    0.000000]  ffffffff81045a7b ffffffff81ced5a2 0000000000000000 ffffffffffffffff
> [    0.000000]  0000000000000040 0000000000000200 0000000067da2bfc ffffffff81a01e88
> [    0.000000] Call Trace:
> [    0.000000]  [<ffffffff814a873c>] dump_stack+0x19/0x1d
> [    0.000000]  [<ffffffff81045a7b>] warn_slowpath_common+0x6b/0xa0
> [    0.000000]  [<ffffffff81045ac5>] warn_slowpath_null+0x15/0x20
> [    0.000000]  [<ffffffff81b1793a>] ___alloc_bootmem_nopanic+0x79/0x82
> [    0.000000]  [<ffffffff81b17a59>] ___alloc_bootmem+0x11/0x3c
> [    0.000000]  [<ffffffff81b17aa4>] __alloc_bootmem+0x10/0x12
> [    0.000000]  [<ffffffff81b233a7>] alloc_bootmem_cpumask_var+0x1d/0x27
> [    0.000000]  [<ffffffff81b1119b>] rcu_bootup_announce_oddness+0xd0/0x153
> [    0.000000]  [<ffffffff81b11735>] rcu_init+0x1e/0x1e6
> [    0.000000]  [<ffffffff81aedf2f>] start_kernel+0x1e6/0x43c
> [    0.000000]  [<ffffffff81aedb3b>] ? repair_env_string+0x58/0x58
> [    0.000000]  [<ffffffff81aed4d1>] x86_64_start_reservations+0x1b/0x32
> [    0.000000]  [<ffffffff81aed612>] x86_64_start_kernel+0x12a/0x131
> [    0.000000] ---[ end trace c8b13488e92fad65 ]---
> [    0.000000]  Experimental no-CBs for all CPUs
> [    0.000000]  Experimental no-CBs CPUs: 0-31.
> [    0.000000] NO_HZ: Full dynticks CPUs: 1-31.

Could you please try the following patch and let me know if it helps?

								Thanx, Paul

------------------------------------------------------------------------

rcu: Don't allocate bootmem from rcu_init()

When rcu_init() is called we already have slab working, allocating
bootmem at that point results in warnings and an allocation from
slab.  This commit therefore changes alloc_bootmem_cpumask_var() to
alloc_cpumask_var() in rcu_bootup_announce_oddness(), which is called
from rcu_init().

Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>

[paulmck: convert to zalloc_cpumask_var(), as suggested by Yinghai Lu.]

diff --git a/kernel/rcutree_plugin.h b/kernel/rcutree_plugin.h
index a8c39c7..207844e 100644
--- a/kernel/rcutree_plugin.h
+++ b/kernel/rcutree_plugin.h
@@ -87,7 +87,7 @@ static void __init rcu_bootup_announce_oddness(void)
 #ifdef CONFIG_RCU_NOCB_CPU
 #ifndef CONFIG_RCU_NOCB_CPU_NONE
 	if (!have_rcu_nocb_mask) {
-		alloc_bootmem_cpumask_var(&rcu_nocb_mask);
+		zalloc_cpumask_var(&rcu_nocb_mask, GFP_KERNEL);
 		have_rcu_nocb_mask = true;
 	}
 #ifdef CONFIG_RCU_NOCB_CPU_ZERO