* [dpdk-users] rte_eal_init fails with --socket-mem set to more than 192MB per socket
@ 2016-08-31 19:15 Sarthak Ray
2016-08-31 19:58 ` Wiles, Keith
0 siblings, 1 reply; 3+ messages in thread
From: Sarthak Ray @ 2016-08-31 19:15 UTC (permalink / raw)
To: users
Hi,
I am using dpdk-2.1.0 and I am not able to reserve memory beyond 192MB per socket by using --socket-mem option. I see below error logs, though my system has enough free memory.
EAL: Not enough memory available on socket 0! Requested: 256MB, available: 192MB
PANIC in rte_eal_init():
# numactl -H
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
node 0 size: 65170 MB
node 0 free: 47433 MB
node 1 cpus: 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71
node 1 size: 65536 MB
node 1 free: 49999 MB
node distances:
node 0 1
0: 10 21
1: 21 10
Is there any max limit for reserving memory per socket? If yes, then please suggest me how to increase that limit.
Thanks in advance,
Sarthak
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [dpdk-users] rte_eal_init fails with --socket-mem set to more than 192MB per socket
2016-08-31 19:15 [dpdk-users] rte_eal_init fails with --socket-mem set to more than 192MB per socket Sarthak Ray
@ 2016-08-31 19:58 ` Wiles, Keith
2016-09-01 6:46 ` Sarthak Ray
0 siblings, 1 reply; 3+ messages in thread
From: Wiles, Keith @ 2016-08-31 19:58 UTC (permalink / raw)
To: Sarthak Ray; +Cc: users
Regards,
Keith
> On Aug 31, 2016, at 2:15 PM, Sarthak Ray <sarthak_ray@outlook.com> wrote:
>
> Hi,
>
> I am using dpdk-2.1.0 and I am not able to reserve memory beyond 192MB per socket by using --socket-mem option. I see below error logs, though my system has enough free memory.
>
> EAL: Not enough memory available on socket 0! Requested: 256MB, available: 192MB
> PANIC in rte_eal_init():
\Most of the time this means that contiguous memory is not available and you have fragmented huge pages. The normal fix is to make sure you allocate the huge pages early in boot up, which to me means making sure you have the /etc/sysctl.conf file setup with the number of huge pages.
vm.nr_hugepages=NNN
>
> # numactl -H
> available: 2 nodes (0-1)
> node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
> node 0 size: 65170 MB
> node 0 free: 47433 MB
> node 1 cpus: 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71
> node 1 size: 65536 MB
> node 1 free: 49999 MB
> node distances:
> node 0 1
> 0: 10 21
> 1: 21 10
>
> Is there any max limit for reserving memory per socket? If yes, then please suggest me how to increase that limit.
>
> Thanks in advance,
> Sarthak
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [dpdk-users] rte_eal_init fails with --socket-mem set to more than 192MB per socket
2016-08-31 19:58 ` Wiles, Keith
@ 2016-09-01 6:46 ` Sarthak Ray
0 siblings, 0 replies; 3+ messages in thread
From: Sarthak Ray @ 2016-09-01 6:46 UTC (permalink / raw)
To: Wiles, Keith; +Cc: users
Hi Keith,
Thanks for your response; but I am reserving enough hugepages during boot time so that user space can't fragment those memories.
System boot arguments related to hugepages
hugepagesz=1g hugepages=24 hugepagesz=2m hugepages=1024
Memory status before my application starts.
# cat /proc/meminfo
MemTotal: 131990816 kB
MemFree: 103888408 kB
Buffers: 220 kB
Cached: 326596 kB
SwapCached: 0 kB
Active: 619104 kB
Inactive: 309796 kB
Active(anon): 602108 kB
Inactive(anon): 952 kB
Active(file): 16996 kB
Inactive(file): 308844 kB
Unevictable: 0 kB
Mlocked: 0 kB
SwapTotal: 0 kB
SwapFree: 0 kB
Dirty: 2252 kB
Writeback: 0 kB
AnonPages: 601976 kB
Mapped: 41560 kB
Shmem: 976 kB
Slab: 130140 kB
SReclaimable: 69672 kB
SUnreclaim: 60468 kB
KernelStack: 6104 kB
PageTables: 10320 kB
NFS_Unstable: 0 kB
Bounce: 0 kB
WritebackTmp: 0 kB
CommitLimit: 52888208 kB
Committed_AS: 1771680 kB
VmallocTotal: 34359738367 kB
VmallocUsed: 1310168 kB
VmallocChunk: 34291387912 kB
HugePages_Total: 512
HugePages_Free: 512
HugePages_Rsvd: 0
HugePages_Surp: 0
Hugepagesize: 2048 kB
DirectMap4k: 12480 kB
DirectMap2M: 1906688 kB
DirectMap1G: 134217728 kB
Memory status after my application exits with rte_eal_init failure.
# cat /proc/meminfo
MemTotal: 131990816 kB
MemFree: 99747864 kB
Buffers: 240 kB
Cached: 440156 kB
SwapCached: 0 kB
Active: 839728 kB
Inactive: 327164 kB
Active(anon): 820448 kB
Inactive(anon): 1920 kB
Active(file): 19280 kB
Inactive(file): 325244 kB
Unevictable: 3704100 kB
Mlocked: 3704100 kB
SwapTotal: 0 kB
SwapFree: 0 kB
Dirty: 7832 kB
Writeback: 0 kB
AnonPages: 4433016 kB
Mapped: 135536 kB
Shmem: 9088 kB
Slab: 133932 kB
SReclaimable: 71568 kB
SUnreclaim: 62364 kB
KernelStack: 7344 kB
PageTables: 20876 kB
NFS_Unstable: 0 kB
Bounce: 0 kB
WritebackTmp: 0 kB
CommitLimit: 52798096 kB
Committed_AS: 5902824 kB
VmallocTotal: 34359738367 kB
VmallocUsed: 1310168 kB
VmallocChunk: 34291387912 kB
HugePages_Total: 600
HugePages_Free: 0
HugePages_Rsvd: 0
HugePages_Surp: 0
Hugepagesize: 2048 kB
DirectMap4k: 12480 kB
DirectMap2M: 1906688 kB
DirectMap1G: 134217728 kB
Thanks,
Sarthak
________________________________
From: Wiles, Keith <keith.wiles@intel.com>
Sent: Thursday, September 1, 2016 1:28:36 AM
To: Sarthak Ray
Cc: users@dpdk.org
Subject: Re: [dpdk-users] rte_eal_init fails with --socket-mem set to more than 192MB per socket
Regards,
Keith
> On Aug 31, 2016, at 2:15 PM, Sarthak Ray <sarthak_ray@outlook.com> wrote:
>
> Hi,
>
> I am using dpdk-2.1.0 and I am not able to reserve memory beyond 192MB per socket by using --socket-mem option. I see below error logs, though my system has enough free memory.
>
> EAL: Not enough memory available on socket 0! Requested: 256MB, available: 192MB
> PANIC in rte_eal_init():
\Most of the time this means that contiguous memory is not available and you have fragmented huge pages. The normal fix is to make sure you allocate the huge pages early in boot up, which to me means making sure you have the /etc/sysctl.conf file setup with the number of huge pages.
vm.nr_hugepages=NNN
>
> # numactl -H
> available: 2 nodes (0-1)
> node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
> node 0 size: 65170 MB
> node 0 free: 47433 MB
> node 1 cpus: 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71
> node 1 size: 65536 MB
> node 1 free: 49999 MB
> node distances:
> node 0 1
> 0: 10 21
> 1: 21 10
>
> Is there any max limit for reserving memory per socket? If yes, then please suggest me how to increase that limit.
>
> Thanks in advance,
> Sarthak
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2016-09-01 6:47 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-08-31 19:15 [dpdk-users] rte_eal_init fails with --socket-mem set to more than 192MB per socket Sarthak Ray
2016-08-31 19:58 ` Wiles, Keith
2016-09-01 6:46 ` Sarthak Ray
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).