SMP with clang hangs on hifive in release config
lsf37 opened this issue · comments
The combination of clang, release, and SMP hangs on boot on the hifive board, see e.g. https://github.com/seL4/seL4/actions/runs/7508485461/job/20463803054
U-Boot 2018.09-g6f6e014-dirty (Jul 04 2019 - 12:40:44 +1000)
DRAM: 2 GiB
MMC:
In: serial
Out: serial
Err: serial
Net: gmac0
RISC-V # setenv autoload no
dhcp
setenv autoload no
RISC-V # dhcp
gmac0: PHY present at 0
gmac0: Starting autonegotiation...
gmac0: Autonegotiation complete
gmac0: link up, 1000Mbps full-duplex (lpa: 0x3800)
BOOTP broadcast 1
BOOTP broadcast 2
BOOTP broadcast 3
BOOTP broadcast 4
BOOTP broadcast 5
DHCP client bound to address 172.16.1.52 (3766 ms)
RISC-V # tftp 0x80000000 /hifive/sel4-image
tftp 0x80000000 /hifive/sel4-image
gmac0: PHY present at 0
gmac0: Starting autonegotiation...
gmac0: Autonegotiation complete
gmac0: link up, 1000Mbps full-duplex (lpa: 0x3800)
Using gmac0 device
TFTP from server 172.16.0.2; our IP address is 172.16.1.52
Filename '/hifive/sel4-image'.
Load address: 0x80000000
Loading: *�#################################################################
#################################################################
#################################################################
############################################
608.4 KiB/s
done
Bytes transferred = 3498264 (356118 hex)
RISC-V # go 0x80000000
go 0x80000000
## Starting application at 0x80000000 ...
OpenSBI v0.9
____ _____ ____ _____
/ __ \ / ____| _ \_ _|
| | | |_ __ ___ _ __ | (___ | |_) || |
| | | | '_ \ / _ \ '_ \ \___ \| _ < | |
| |__| | |_) | __/ | | |____) | |_) || |_
\____/| .__/ \___|_| |_|_____/|____/_____|
| |
|_|
Platform Name : sifive,hifive-unleashed-a00
Platform Features : timer,mfdeleg
Platform HART Count : 5
Firmware Base : 0x80000000
Firmware Size : 132 KB
Runtime SBI Version : 0.2
Domain0 Name : root
Domain0 Boot HART : 2
Domain0 HARTs : 0*,1*,2*,3*,4*
Domain0 Region00 : 0x0000000080000000-0x000000008003ffff ()
Domain0 Region01 : 0x0000000000000000-0xffffffffffffffff (R,W,X)
Domain0 Next Address : 0x0000000080200000
Domain0 Next Arg1 : 0x0000000082200000
Domain0 Next Mode : S-mode
Domain0 SysReset : yes
Boot HART ID : 2
Boot HART Domain : root
Boot HART ISA : rv64imafdcsu
Boot HART Features : scounteren,mcounteren
Boot HART PMP Count : 16
Boot HART PMP Granularity : 4
Boot HART PMP Address Bits: 36
Boot HART MHPM Count : 0
Boot HART MHPM Count : 0
Boot HART MIDELEG : 0x0000000000000222
Boot HART MEDELEG : 0x000000000000b109
ELF-loader started on (HART 1) (NODES 4)
paddr=[80200000..8035d027]
Looking for DTB in CPIO archive...found at 8021b908.
Loaded DTB from 8021b908.
paddr=[84023000..84025fff]
ELF-loading image 'kernel' to 84000000
paddr=[84000000..84022fff]
vaddr=[ffffffff84000000..ffffffff84022fff]
virt_entry=ffffffff84000000
ELF-loading image 'sel4test-driver' to 84026000
paddr=[84026000..8420bfff]
vaddr=[10000..1f5fff]
virt_entry=14474
Main entry hart_id:1
Secondary entry hart_id:2 core_id:1
Hart ID 2 core ID 1
Hart ID 1 core ID 0
[[Timeout]]
It works fine on verification and debug builds (they succeeded in the same job above). Looks like this is actually before we enter the kernel. A successful run in debug mode shows this:
ELF-loader started on (HART 1) (NODES 4)
paddr=[80200000..80bc8027]
Looking for DTB in CPIO archive...found at 8039ea88.
Loaded DTB from 8039ea88.
paddr=[84032000..84034fff]
ELF-loading image 'kernel' to 84000000
paddr=[84000000..84031fff]
vaddr=[ffffffff84000000..ffffffff84031fff]
virt_entry=ffffffff84000000
ELF-loading image 'sel4test-driver' to 84035000
paddr=[84035000..846cafff]
vaddr=[10000..6a5fff]
virt_entry=1ef2c
Main entry hart_id:1
Secondary entry hart_id:2 core_id:1
Secondary entry hart_id:4 core_id:3
Secondary entry hart_id:3 core_id:2
Hart ID 2 core ID 1
Hart ID 3 core ID 2
Hart ID 4 core ID 3
Hart ID 1 core ID 0
Enabling MMU and paging
Jumping to kernel-image entry point...
@axel-h any idea what might be going on here?
Maybe. Sometime ago I have been experimenting with RISC-V SMP a bit and noticed some issues. I will have a look at my old patches, and try running them in CI to see if this changes something.