On 12/19/19 2:16 PM, Qian Cai wrote:
On Dec 19, 2019, at 8:06 AM, Vlastimil Babka vbabka@suse.cz wrote:
Commit 96a2b03f281d ("mm, debug_pagelloc: use static keys to enable debugging") has introduced a static key to reduce overhead when debug_pagealloc is compiled in but not enabled. It relied on the assumption that jump_label_init() is called before parse_early_param() as in start_kernel(), so when the "debug_pagealloc=on" option is parsed, it is safe to enable the static key.
However, it turns out multiple architectures call parse_early_param() earlier from their setup_arch(). x86 also calls jump_label_init() even earlier, so no issue was found while testing the commit, but same is not true for e.g. ppc64 and s390 where the kernel would not boot with debug_pagealloc=on as found by our QA.
This was daily tested on linux-next here for those arches and never saw an issue.
Well I assume nobody has booted the kernel specifically with debug_pagealloc=on. In case a randconfig run enabled CONFIG_DEBUG_PAGEALLOC_ENABLE_DEFAULT then the problem wouldn't manifest as the static key would be initialized with DEFINE_STATIC_KEY_TRUE by the compiler itself, and the enabling-by-param would be a no-op.
Are you able to reproduce it on mainline or linux-next?
I didn't try, but our kernel is 5.3-based so that's quite recent. The offending commit was also introduced in 5.3-rc1 so there's that.