Re: [PATCH mach v4 1/1] Integrate HPET so the functions used for getting time can have a higher accuracy

[Michael Banck]

Hi,

On Fri, Jun 13, 2025 at 12:42:53AM +0000, Damien Zammit wrote:
> Are you sure you are giving it enough memory?

For those hangs I saw, I am pretty sure. However, I had to move to the
pae-enabled kernel with 4GB memory in order to make the full testsuite
run reliably, it does take slightly more than 2GB at some point and then
crashes as it does not seem to swap out.

Also, I noticed that there's about 600M of free memory difference
between before running the testsuite and afterwards:

|               total        used        free      shared  buff/cache   
available                                                    
|Mem:         4129164      412932     3716232           0           0     
3716232                                                    

|               total        used        free      shared  buff/cache   
available                                                    
|Mem:         4098876     1015220     3083656           0           0     
3083656                                                    

That seems to be taken up mostly by the rumpdisk task:

|PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND     
                                                   
|  7 root       7 -13  343552 211800      0 R  15.8   5.2   5:15.41 rumpdisk    
                                                   

The good news is that this does not increase on consecutive runs, I was
able to run the buildfarm testsuite several times now without crashes or
hangs or having to reboot.


Michael