On 5/21/23 23:37, Hans Petter Selasky wrote:
DEFINE_MUTEX is Linux’s API, implemented in LinuxKPI. So Linux’s
behaviour is absolutely relevant. Any deviation from Linux’s semantics
is an incompatibility that requires patching any sources that are built
for FreeBSD using LinuxKPI. It is generally best to minimise that.
Jess, listen:
Probably at some point, there could be an initialization macro for the
SX locks used in the LinuxKPI, being of type SX_NOWITNESS, simply
plotting in the constants needed in the sx structure, exactly like they
do under Linux.
Then neither SYSINITs, nor my patch will be needed for the sake of
DEFINE_MUTEX(). Can we wait for that?
The reason for this is Linux has a bad habit of not destroying locks, so
FreeBSD cannot add any WITNESS objects for locks that are not destroyed
anyway. Most of the work done by sx_init() is putting a few constants
into various SX structure fields. Only the WITNESS part needs execution,
if you get what I mean.
This change is made so I can let others easily test another patchset.
Also this change is harmless as such.
Looking at the whole Linux tree, the most common way is to use
DEFINE_MUTEX() outside macros and functions:
# grep -r DEFINE_MUTEX * | grep -vE ":static " | grep -vE
":DEFINE_MUTEX" | wc -l
57
# grep -r DEFINE_MUTEX * | wc -l
1293
--HPS
