Re: [PATCH v3 2/6] futex2: Implement vectorized wait

André Almeida [off-list ref] writes:

quoted

quoted

+/**
+ * struct futex_waitv - A waiter for vectorized wait
+ * @val:	Expected value at uaddr
+ * @uaddr:	User address to wait on
+ * @flags:	Flags for this waiter
+ * @__reserved:	Reserved member to preserve data alignment. Should be 0.
+ */
+struct futex_waitv {
+	__u64 val;
+	__u64 uaddr;
+	__u32 flags;
+	__u32 __reserved;
+};

why force uaddr  to be __u64, even for 32-bit?  uaddr could be a (void*) for
all we care, no?  Also, by adding a reserved field, you are wasting 32
bits even on 32-bit architectures.

We do that to make the structure layout compatible with both entry
points, remove the need for special cast and duplicated code, as
suggested by Thomas and Arnd:

https://lore.kernel.org/lkml/87v94310gm.ffs@tglx/

https://lore.kernel.org/lkml/CAK8P3a0MO1qJLRkCH8KrZ3+=L66KOsMRmcbrUvYdMoKykdKoyQ@mail.gmail.com/

I find this weird.  I'm not even juts talking about compat, but even on
native 32-bit. But also, 32 applications on 64, which is a big use
case for games.

The structure is mandating a 64 bit uaddr field and has an unnecessary
pad.  You are wasting 20% of the space, which is gonna be elements of a
vector coming from user space.  Worst case, you are doing copy_from_user
of an extra 1k bytes in the critical path of futex_waitv for no good
reason.

Also, if I understand correctly, Arnd suggestion, at least, was to have
two parser functions and a single syscall entry point, that would do the
translation:

if (in_compat_syscall())
   futex_parse_waitv_compat(futexv, waiters, nr_futexes);
else
   futex_parse_waitv(futexv, waiters, nr_futexes);

-- 
Gabriel Krisman Bertazi

On Thu, Sep 16, 2021 at 12:10:25AM -0400, Gabriel Krisman Bertazi wrote:

I find this weird.  I'm not even juts talking about compat, but even on
native 32-bit. But also, 32 applications on 64, which is a big use
case for games.

Seriously, people still make 32bit applications today? And for legacy
games, I would think the speed increase of modern CPUs would far offset
this little inefficiency.

On Thu, Sep 16, 2021 at 1:22 PM Peter Zijlstra [off-list ref] wrote:

On Thu, Sep 16, 2021 at 12:10:25AM -0400, Gabriel Krisman Bertazi wrote:

quoted

I find this weird.  I'm not even juts talking about compat, but even on
native 32-bit. But also, 32 applications on 64, which is a big use
case for games.

Seriously, people still make 32bit applications today? And for legacy
games, I would think the speed increase of modern CPUs would far offset
this little inefficiency.

There are 32-bit Windows games apparently, because it's easier to build it
that way than having both 32-bit and 64-bit versions.
There may be native 32-bit games built for Linux from the same sources when
that is not written portably, not sure if that's a thing.

One important reason to use compat mode is for cost savings when you can
ship an embedded system with slightly less RAM by running 32-bit user space
on it. We even still see people running 32-bit kernels on Arm boxes that have
entry-level 64-bit chips, though I hope that those will migrate the
kernel to arm64
even when they ship 32-bit user space.

Similar logic applies to cloud instances or containers. Running a 32-bit
Alpine Linux in a container means you can often go to a lower memory
instance on the host compared to a full 64-bit distro.

        Arnd

On Thu, 16 Sep 2021 13:50:14 +0200
Arnd Bergmann [off-list ref] wrote:

Similar logic applies to cloud instances or containers. Running a 32-bit
Alpine Linux in a container means you can often go to a lower memory
instance on the host compared to a full 64-bit distro.

I also found that running a 32 bit version of Chrome or FireFox keeps them
from taking up all the memory in your system ;-)  The most they can use is
4 gigs.

-- Steve