Hi! On Wed, Aug 05, 2020 at 06:51:44PM +0200, Christophe Leroy wrote: > Le 05/08/2020 à 16:03, Segher Boessenkool a écrit : > >On Wed, Aug 05, 2020 at 07:09:23AM +0000, Christophe Leroy wrote: > >>+/* > >>+ * The macros sets two stack frames, one for the caller and one for the > >>callee > >>+ * because there are no requirement for the caller to set a stack frame > >>when > >>+ * calling VDSO so it may have omitted to set one, especially on PPC64 > >>+ */ > > > >If the caller follows the ABI, there always is a stack frame. So what > >is going on? > > Looks like it is not the case. See discussion at > https://patchwork.ozlabs.org/project/linuxppc-dev/patch/2a67c333893454868bbfda773ba4b01c20272a5d.1588079622.git.christophe.leroy@xxxxxx/ > > Seems like GCC uses the redzone and doesn't set a stack frame. I guess > it doesn't know that the inline assembly contains a function call so it > doesn't set the frame. Yes, that is the problem. See https://gcc.gnu.org/onlinedocs/gcc-10.2.0/gcc/Extended-Asm.html#AssemblerTemplate where this is (briefly) discussed: "Accessing data from C programs without using input/output operands (such as by using global symbols directly from the assembler template) may not work as expected. Similarly, calling functions directly from an assembler template requires a detailed understanding of the target assembler and ABI." I don't know of a good way to tell GCC some function needs a frame (that is, one that doesn't result in extra code other than to set up the frame). I'll think about it. Segher
