> This patch implements a checker to detect instructions which > should be paravirtualized instead of direct writing raw instruction. > This patch does rough check so that it doesn't fully cover all cases, > but it can detects most cases of paravirtualization breakage of hand > written assembly codes.
There are still some "itc.d" instructions in ivt.S (in the #ifndef CONFIG_SMP code). This checker caught them ... but the error messages from the build were not as elegant as they might be AS arch/ia64/kernel/pvchk-ivt.o arch/ia64/kernel/ivt.S: Assembler messages: arch/ia64/kernel/ivt.S:583: Warning: itc.d should not be used directly. arch/ia64/kernel/ivt.S:583: Error: junk at end of line, first unrecognized character is `r' arch/ia64/kernel/ivt.S:649: Warning: itc.i should not be used directly. arch/ia64/kernel/ivt.S:649: Error: junk at end of line, first unrecognized character is `r' arch/ia64/kernel/ivt.S:701: Warning: itc.d should not be used directly. arch/ia64/kernel/ivt.S:701: Error: junk at end of line, first unrecognized character is `r' make[1]: *** [arch/ia64/kernel/pvchk-ivt.o] Error 1 make: *** [arch/ia64/kernel] Error 2 I'll hold off on applying part32/32 until the CONFIG_SMP=n case is fixed. -Tony _______________________________________________ Virtualization mailing list [email protected] https://lists.linux-foundation.org/mailman/listinfo/virtualization
