On 9/10/2019 11:04 PM, Ingo Molnar wrote:
* Mike Travis <mike.tra...@hpe.com> wrote:
@@ -1596,7 +1687,7 @@ static void __init uv_system_init_hub(vo
uv_nmi_setup();
uv_cpu_init();
uv_scir_register_cpu_notifier();
- proc_mkdir("sgi_uv", NULL);
+ uv_setup_proc_files(0);
This slipped through previously: platform drivers have absolutely no
business mucking in /proc.
Please describe the hardware via sysfs as pretty much everyone else does.
Thanks,
Ingo
If I was doing it now I definitely would put it in the sysfs realm. The
problem is Jack did it back in (I think) 2007. The earliest commit I
could find:
commit a3d732f93785da17e0137210deadb4616f5536fc
Author: Cliff Wickman <c...@sgi.com>
Date: Mon Nov 10 16:16:31 2008 -0600
x86, UV: fix redundant creation of sgi_uv
Impact: fix double entry creation in /proc
And in the past 12 years probably a hundred user programs are now keying
of the presence of /proc/sgi_uv to signal this is indeed a UV system.
Changing the location of this node also affects all the UV utilities
including those not written by us.
