On 11/21/23 23:05, Rafael J. Wysocki wrote:
So wouldn't it be better to avoid modifying *value at all if
raw_pci_read() returns an error?
Avoiding of modification of *value at all seems better idea to me than
setting it to arbitrary initializing value, indeed.
In that case, buffer initialization can be ditched, and *value set only
in case of success.
And if it returns a success, why wouldn't it be trusted?
My concern is, that raw_pci_read() wraps platform-specific handlers,
that should conform to the next rules:
1) in case of success, they must set value32 (or else, uninitialized
data would leak to acpi_os_read_pci_configuration caller);
2) they should use passed &value32 only to set it (or else,
uninitialized data would be used/passed somewhere, is it safe?);
Is there any way to be sure, that all the existing and future
platform-specific pci-read handlers conform?
