[PATCH v2 11/11] powerpc/64s/slb: Add no_slb_preload early cmdline param

[Ritesh Harjani (IBM)]

pavrampu <pavrampu at imap.linux.ibm.com> writes:

>
> Thanks for the patches, I have verified patches with below tests and no 
> issues are observed.
>
> Run below tests in hash mode
> 1. slb multihit error injection module, test from 
> https://github.ibm.com/mahesh-salgaonkar/mce_err_inject.git
>     [ insmod mcetest_slb.ko ]
> 2. perf stat -e context-switches hackbench -l 1000 -g 50
> 3. perf stat -e context-switches stress-ng --cpu 0 --timeout 60s
> 4. Regression tests:
> LTP mm tests
> HTX mdt.mem
> stress-ng Memory
> memtester
> kselftest
> 5. Boot with dyndbg="file arch/powerpc/mm/book3s64/hash_utils.c +p" and 
> check there are no errors
> 6. echo "file arch/powerpc/mm/book3s64/hash_utils.c +p" > 
> /sys/kernel/debug/dynamic_debug/control
> 7. Verify hash mmu messages in dmesg
> # dmesg | grep -Eai "Page orders"
> # dmesg | grep -Eai "Using 1TB"
> # dmesg | grep -Eai "Huge page(16GB)"
> 8. hpt_order is updated properly based on "Maximum Memory" value and it 
> is no affected by dlpar operations.
> 9. Boot lpar in radix mode and verify 
> /sys/kernel/debug/powerpc/hpt_order is not available
> 10.Boot lpar in hash mode and check DirectMap16M has valid value,
> 11. Verify DirectMap16M is updated properly with memory dlpar 
> operations.
> 12. Create devdax device and check DirectMap16M is updated, Convert it 
> to system RAM and back to devdax and verify DirectMap16M
> 13. Boot kernel in radix mode and /sys/kernel/debug/kernel_page_tables 
> has PXX level info.
> 14. Boot lpar with no_slb_preload cmd line argument systemd-analyze with 
> and without no_slb_preload
> 15. Check slb_faults with and without no_slb_preload
> 16. kdump/fadump on 1TB system with above patches in radix and hash 
> mode.
> 17. 31TB P8 system boot with above patches
>
>
> Feel free to add:
> Tested-by: Pavithra Prakash (IBM) <pavrampu at linux.ibm.com>

Thanks a lot Pavithra for an extensive testing of this patch series!

-ritesh