HypX Internal Logging Mechanism

HypX has a fairly simple logging mechanism.

The HypX logging mechanism starts at a function called dprintf which works similar to how dprintf works in libc.

In HypX’s version of dprintf, there are 2 valid fds: 3 and 4. 3 indicates a regular log while 4 indicates a panic log. If 4 is run then the code will infinite loop.

Do note that this logging mechanism is only used for hypervisor internal logging. There is a seperate logging mechanism for the RKP log. This logging mechanism writes data to a known physical address so that the kernel can expose it in /proc/uh_log.

Initializing the Log

The log is initialized in a function called initialize_log. This sets up a few key variables for the logging mechanism.

The first important variable is log_buf which is where the log starts. This is always 0x80209200. The second is max_log_size which indicates the size of the log before it resets to the beginning.

A look at dprintf

dprintf is a fairly complicated function but not much of it actually matters. The main magic happens at __dprintf -> write_to_log. write_to_log is where the format string stuff happens. To actually write the formatted string to the log, a function called putchar is used to write each character individually. putchar calls another subroutine called __putchar which has the following implementation:

__putchar

void __putchar(char c)

{
uint next_log_cursor;

if (log_initialized? != 0) {
    *(char *)(log_buf + (ulong)log_cursor) = c;
    next_log_cursor = 0;
    if (max_log_size != 0) {
    next_log_cursor = (log_cursor + 1) / max_log_size;
    }
    next_log_cursor = (log_cursor + 1) - next_log_cursor * max_log_size;
    log_cursor = (ushort)next_log_cursor;
    if ((next_log_cursor & 0xffff) == 0) {
    DAT_8020900c = DAT_8020900c + 1;
    return;
    }
}
return;
}

The above implementation is fairly simple. All it does is write the character to log_buf + log_cursor. If log_cursor+1 (The next place to write) is equal to max_log_size. Then, we reset the log_cursor to 0. This ensures that the log doesn’t overflow data.

The format of a log entry is as follows:

[cpu] [timestamp] [log text]