Hi All,
We are getting "warn_alloc" warning during boot of kdump kernel. This
warning is observed with latest upstream tag (v5.7-rc5).
Primary/1st Kernel
----------------------------
# dmesg | grep crash
[ 0.000000] crashkernel reserved: 0x00000000d6000000 -
0x00000000f6000000 (512 MB)
[ 0.000000] Kernel command line:
BOOT_IMAGE=(hd8,gpt2)/vmlinuz-5.7.0-rc5
root=UUID=c4050f17-526f-48a8-9804-c6b35cbb584c ro crashkernel=512M
earlycon console=ttyAMA0
# cat /proc/iomem | grep -i crash
d6000000 - f6000000 : Crash kernel
Logs from Kdump/crash kernel with warnings & dump_stack
------------------------------------------------------------------------
[ 0.239360] swapper/0: page allocation failure: order:2,
mode:0x1(GFP_DMA), nodemask=(null),cpuset=/,mems_allowed=0
[ 0.249917] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.7.0-rc5 #44
[ 0.256246] Hardware name: To be filled by O.E.M. Saber/Saber, BIOS
0ACKL027 07/01/2019
[ 0.264333] Call trace:
[ 0.266797] dump_backtrace+0x0/0x1f8
[ 0.270490] show_stack+0x20/0x30
[ 0.273833] dump_stack+0xc0/0x10c
[ 0.277263] warn_alloc+0x10c/0x178
[ 0.280781] __alloc_pages_slowpath.constprop.112+0xaec/0xb28
[ 0.286584] __alloc_pages_nodemask+0x2b4/0x300
[ 0.291156] alloc_page_interleave+0x24/0xa0
[ 0.295464] alloc_pages_current+0xe4/0x108
[ 0.299686] dma_atomic_pool_init+0x44/0x1a4
[ 0.303995] do_one_initcall+0x54/0x228
[ 0.307864] kernel_init_freeable+0x228/0x2cc
[ 0.312263] kernel_init+0x1c/0x110
[ 0.315781] ret_from_fork+0x10/0x18
We did some debugging.
As per commit 1a8e1cef7603 ("arm64: use both ZONE_DMA and ZONE_DMA32")
. DMA zone has been re-defined.
here, ZONE_DMA has a fixed range of 0x802f0000 - 0xbfffffff and
ZONE_DMA32 has range from 0xc0000000-0xfffffffff.
When bootargs is defined with "crashkernel= X" for 1st/primary kernel.
Than X amount of memory is reserved in First kernel. This reserved
memory is used to boot kdump/crash kernel and represented as "Crash
kernel" in cat /prom/iomem.
If some region of reserved memory(Crash kernel) **does not** fall in
ZONE_DMA region i.e. 0x802f0000 - 0xbfffffff, this warning is
observed.
Other drivers like scsi_register_driver [1] also fail. We also see
other kinds of error [2].
Considering DMA_ZONE has requirement of 0x802f0000 - 0xbfffffff.
Can we enforce "Crash kernel" to always reserved between 0x0000_0000
to 0xc000_0000 in reserve_crashkernel() -->memblock_find_in_range()
or
what could be best possible solution.
--pk
[1]
------------------------------------------------------------
[ 21.509239] dump_backtrace+0x0/0x1f8
[ 21.516592] show_stack+0x20/0x30
[ 21.523248] dump_stack+0xc0/0x10c
[ 21.530087] warn_alloc+0x10c/0x178
[ 21.537090] __alloc_pages_slowpath.constprop.112+0xaec/0xb28
[ 21.548644] __alloc_pages_nodemask+0x2b4/0x300
[ 21.557750] alloc_pages_current+0x90/0x108
[ 21.566155] alloc_slab_page+0x184/0x340
[ 21.574030] new_slab+0x420/0x4c8
[ 21.580681] ___slab_alloc+0x354/0x4e8
[ 21.588207] __slab_alloc+0x28/0x58
[ 21.595210] kmem_cache_alloc_trace+0x230/0x250
[ 21.604316] sr_probe+0x250/0x618 [sr_mod]
[ 21.612555] really_probe+0xe4/0x448
[ 21.619733] driver_probe_device+0xe8/0x140
[ 21.628136] device_driver_attach+0x7c/0x88
[ 21.636536] __driver_attach+0xac/0x178
[ 21.644239] bus_for_each_dev+0x7c/0xd0
[ 21.651943] driver_attach+0x2c/0x38
[ 21.659119] bus_add_driver+0x1a8/0x240
[ 21.666823] driver_register+0x6c/0x128
[ 21.674533] scsi_register_driver+0x28/0x38
[ 21.682939] init_sr+0x40/0x10000 [sr_mod]
[2]
-------------------------------------------------------------------
[ 21.450571] systemd-udevd: page allocation failure: order:0,
mode:0xcc1(GFP_KERNEL|GFP_DMA),
nodemask=(null),cpuset=/,mems_allowed=0^M
[ 21.450571] systemd-udevd: page allocation failure: order:0,
mode:0xcc1(GFP_KERNEL|GFP_DMA),
nodemask=(null),cpuset=/,mems_allowed=0^M
_______________________________________________
kexec mailing list
kexec@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/kexec
