[Bug 204807] Hardware monitoring sensor nct6798d doesn't work unless acpi_enforce_resources=lax is enabled

[bugzilla-daemon@bugzilla.kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=204807

--- Comment #170 from Eugene Shalygin (eugene.shalygin@gmail.com) ---
(In reply to Denis Pauk from comment #169)

> Could you look to nct6775_data:{un}lock? Do you mean something like it?

Yes, something like that. I think if we hide that under a formal interface and
put in a module, it can be used in both hwmon drivers. And yes, I believe we
should query EC directly, because the WMI functions are simply wrappers that
end up at the same calls to ec_read(). Since we going to lock the ACPI mutex
ourselves, there is no reason to use WMI functions to access EC registers.

Here is what I thought of:

union access_lock_data {
         struct mutex lock;
         struct {
                acpi_handle aml_mutex[MAX_ACPI_MUTEX_COUNT]; // maybe there is
a firmware where there is no single mutex to guard both chips?
                int acpi_mutex_count;
         } acpi_lock;
};

struct access_lock_funcs {
        int (*lock)(access_lock_funcs* funcs);
        int (*unlock)(access_lock_funcs* funcs);
};

struct access_lock_state {
        union access_lock_data data;
        struct access_lock_funcs funcs;
};

int get_access_lock(int flags, struct access_lock_funcs* res); // flags to tell
what needs to be guarded (HWMON chip, EC, or both)
void free_access_lock(struct access_lock_funcs* funcs);

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching the assignee of the bug.