[BUG] um: initramfs doesn't work on uml

From: Masami Hiramatsu
Date: Wed Apr 26 2017 - 23:15:19 EST


Hello,

I found that user-mode-linux kernel doesn't boot with initramfs
(or initrd) as below. I've investigated the reason and found that
the read_initrd@arch/um/kernel/initrd.c tried to allocate memory
for initrd by using alloc_bootmem, but since init_bootmem will
be called from setup_arch afterwards, that alloc_bootmem always
failed.

$ ./vmlinux.um initrd=./initramfs.um
Core dump limits :
soft - 0
hard - NONE
Checking that ptrace can change system call numbers...OK
Checking syscall emulation patch for ptrace...OK
Checking advanced syscall emulation patch for ptrace...OK
Checking environment variables for a tempdir...none found
Checking if /dev/shm is on tmpfs...OK
Checking PROT_EXEC mmap in /dev/shm...OK
Adding 28401664 bytes to physical memory to account for exec-shield gap
kmsg_dump:
<1>[ 0.000000] bootmem alloc of 1654667 bytes failed!
<0>[ 0.000000] Kernel panic - not syncing: Out of memory
<4>[ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 4.11.0-rc6-00200-gd5ff081 #1
<6>[ 0.000000] Stack:
<4>[ 0.000000] 60385e20 6006d399 00000000 6031d32b
<4>[ 0.000000] 6009b81e 601c710b 60385e30 601bcb75
<4>[ 0.000000] 60385f50 6009b5ec 0000002e 00000000
<6>[ 0.000000] Call Trace:
<6>[ 0.000000] [<6009b81e>] ?
<4>[ 0.000000] printk+0x0/0x94
<6>[ 0.000000] [<60020fe0>]
<4>[ 0.000000] show_stack+0x108/0x15e
<6>[ 0.000000] [<6006d399>] ?
<4>[ 0.000000] dump_stack_print_info+0xe4/0xed
<6>[ 0.000000] [<6009b81e>] ?
<4>[ 0.000000] printk+0x0/0x94
<6>[ 0.000000] [<601c710b>] ?
<4>[ 0.000000] bust_spinlocks+0x0/0x3b
<6>[ 0.000000] [<601bcb75>]
<4>[ 0.000000] dump_stack+0x2a/0x2c
<6>[ 0.000000] [<6009b5ec>]
<4>[ 0.000000] panic+0x170/0x321
<6>[ 0.000000] [<6009b47c>] ?
<4>[ 0.000000] panic+0x0/0x321
<6>[ 0.000000] [<6009b81e>] ?
<4>[ 0.000000] printk+0x0/0x94
<6>[ 0.000000] [<60030aec>] ?
<4>[ 0.000000] os_file_size+0x27/0x12d
<6>[ 0.000000] [<6000de0c>] ?
<4>[ 0.000000] alloc_bootmem_core+0x2a/0x10b
<6>[ 0.000000] [<6000dde2>] ?
<4>[ 0.000000] alloc_bootmem_core+0x0/0x10b
<6>[ 0.000000] [<6000daab>]
<4>[ 0.000000] alloc_bootmem_bdata+0x0/0x337
<6>[ 0.000000] [<6000e456>]
<4>[ 0.000000] __alloc_bootmem+0x49/0x52
<6>[ 0.000000] [<600048d5>]
<4>[ 0.000000] read_initrd+0x77/0x132
<6>[ 0.000000] [<60022ada>]
<4>[ 0.000000] uml_finishsetup+0x42/0x59
<6>[ 0.000000]
Aborted

I've tried to make a patch to fix this issue by calling
read_initrd after init_bootmem in setup_arch, but not sure
that follows correctly the design of uml boot process.

Could anyone review that?

Thank you,

---

Masami Hiramatsu (1):
um: Fix to call read_initrd after init_bootmem


arch/um/kernel/initrd.c | 4 +---
arch/um/kernel/um_arch.c | 6 ++++++
2 files changed, 7 insertions(+), 3 deletions(-)

--
Masami Hiramatsu <mhiramat@xxxxxxxxxx>