On Mon, Mar 21, 2022 at 10:41 AM Huacai Chen <chenhuacai@xxxxxxxxxx> wrote:
> On Mon, Mar 21, 2022 at 5:01 PM Arnd Bergmann <arnd@xxxxxxxx> wrote:
> >
> > On Sat, Mar 19, 2022 at 3:38 PM Huacai Chen <chenhuacai@xxxxxxxxxx> wrote:
> > >
> > > This patch adds system call support and related uaccess.h for LoongArch.
> > >
> > > Q: Why keep __ARCH_WANT_NEW_STAT definition while there is statx:
> > > A: Until the latest glibc release (2.34), statx is only used for 32-bit
> > > platforms, or 64-bit platforms with 32-bit timestamp. I.e., Most 64-
> > > bit platforms still use newstat now.
> > >
> > > Q: Why keep _ARCH_WANT_SYS_CLONE definition while there is clone3:
> > > A: The latest glibc release (2.34) has some basic support for clone3 but
> > > it isn't complete. E.g., pthread_create() and spawni() have converted
> > > to use clone3 but fork() will still use clone. Moreover, some seccomp
> > > related applications can still not work perfectly with clone3.
> >
> > Please leave those out of the mainline kernel support though: Any users
> > of existing glibc binaries can keep using patched kernels for the moment,
> > and then later drop those pages when the proper glibc support gets
> > merged.
> The glibc commit d8ea0d0168b190bdf138a20358293c939509367f ("Add an
> internal wrapper for clone, clone2 and clone3") modified nearly
> everything in order to move to clone3(), except arch_fork() which used
> by fork(). And I cannot find any submitted patches to solve it. So I
> don't think this is just a forget, maybe there are other fundamental
> problems?
I don't think there are fundamental issues, they probably did not consider
it necessary because so far all architectures supported clone().
Adding Christian Brauner and H.J. Lu for clarificatoin.
> > > +#define __get_user(x, ptr) \
> > > +({ \
> > > + int __gu_err = 0; \
> > > + \
> > > + __chk_user_ptr(ptr); \
> > > + __get_user_common((x), sizeof(*(ptr)), ptr); \
> > > + __gu_err; \
> > > +})
> >
> > It would be good to also provide a
> > __kernel_kernel_nofault()/__put_kernel_nofault()
> > implementation, as the default based on __get_user()/__put_user is not
> > ideal.
> They are provided in this file below.
Ok, I see them now, not sure what I did wrong when I looked earlier.
Arnd
