suspicious compiler warning in vdso2c.h
Andy
following e6577a7ce9 ("x86, vdso: Move the vvar area before the vdso text") gcc 4.3.4 tells me
.../arch/x86/vdso/vdso2c.c: In function ‘main’:.../arch/x86/vdso/vdso2c.h:118: warning: assuming signed overflow does not occur when assuming that (X + c) < X is always false.../arch/x86/vdso/vdso2c.h:118: warning: assuming signed overflow does not occur when assuming that (X + c) < X is always false
apparently because it unrolls the containing loop and finds i == sym_vvar_start on the first iteration. I think there was a hidden dependency before your change on required_syms[] first two entries matching up with special_pages[]. If that's correct, all syms[i] within said loop would seem to need replacing by syms[special_pages[i]] (similarly for required_syms[i] afaict, see below draft patch), which also is in line with the loop iterating over the array elements of special_pages[].
Question then of course is - are there any further hidden dependencies that already got or may eventually get broken?
Jan
arch/x86/vdso/vdso2c.h +7 -7
@@ -109,18 +109,18 @@ static void BITSFUNC(go)(void *raw_addr,/* Validate mapping addresses. */ for (i =0 ; i <sizeof (special_pages) /sizeof (special_pages[0 ]); i++) { -if (!syms[i]) +if (!syms[special_pages[i]])continue ;/* The mapping isn't used; ignore it. */ -if (syms[i] %4096 ) +if (syms[special_pages[i]] %4096 )fail ("%s must be a multiple of 4096\n" , - required_syms[i].name); -if (syms[sym_vvar_start] > syms[i] +4096 ) + required_syms[special_pages[i]].name); +if (syms[sym_vvar_start] > syms[special_pages[i]] +4096 )fail ("%s underruns begin_vvar\n" , - required_syms[i].name); -if (syms[i] +4096 >0 ) + required_syms[special_pages[i]].name); +if (syms[special_pages[i]] +4096 >0 )fail ("%s is on the wrong side of the vdso text\n" , - required_syms[i].name); + required_syms[special_pages[i]].name); }if (syms[sym_vvar_start] %4096 )fail ("vvar_begin must be a multiple of 4096\n" );
Quoted 11 lines · Jan Beulich
Andy
following
e6577a7ce9("x86, vdso: Move the vvar area before the vdso text") gcc 4.3.4 tells me
.../arch/x86/vdso/vdso2c.c: In function ‘main’:.../arch/x86/vdso/vdso2c.h:118: warning: assuming signed overflow does not occur when assuming that (X + c) < X is always false.../arch/x86/vdso/vdso2c.h:118: warning: assuming signed overflow does not occur when assuming that (X + c) < X is always false
Yes. There's a more or less equivalent fix sitting in -mm.
--Andy
Quoted 44 lines
apparently because it unrolls the containing loop and finds i ==
sym_vvar_starton the first iteration. I think there was a hidden dependency before your change on required_syms[] first two entries matching up with special_pages[]. If that's correct, all syms[i] within said loop would seem to need replacing by syms[special_pages[i]] (similarly for required_syms[i] afaict, see below draft patch), which also is in line with the loop iterating over the array elements of special_pages[].Question then of course is - are there any further hidden dependencies that already got or may eventually get broken?
Jan
arch/x86/vdso/vdso2c.h +7
-7@@ -109,18 +109,18 @@ static void BITSFUNC(go)(void *raw_addr,/* Validate mapping addresses. */ for (i =0 ; i <sizeof (special_pages) /sizeof (special_pages[0 ]); i++) { -if (!syms[i]) +if (!syms[special_pages[i]])continue ;/* The mapping isn't used; ignore it. */ -if (syms[i] %4096 ) +if (syms[special_pages[i]] %4096 )fail ("%s must be a multiple of 4096\n" , - required_syms[i].name); -if (syms[sym_vvar_start] > syms[i] +4096 ) + required_syms[special_pages[i]].name); +if (syms[sym_vvar_start] > syms[special_pages[i]] +4096 )fail ("%s underruns begin_vvar\n" , - required_syms[i].name); -if (syms[i] +4096 >0 ) + required_syms[special_pages[i]].name); +if (syms[special_pages[i]] +4096 >0 )fail ("%s is on the wrong side of the vdso text\n" , - required_syms[i].name); + required_syms[special_pages[i]].name); }if (syms[sym_vvar_start] %4096 )fail ("vvar_begin must be a multiple of 4096\n" );