| Message ID | 1374786537-10726-5-git-send-email-ben.dooks@codethink.co.uk (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Thu, Jul 25, 2013 at 10:08:57PM +0100, Ben Dooks wrote: > When testing the kprobes test code with BE8, there is either an > issue with the linker or how the code is being built. The issue > is with embedding the title text as the first part of the test. > > Change to placing a pointer to .rodata with the text title in it > for the test which seems to stop the issue of the alignment of > the data following it being changed arbitrarily by the linker. > > The proper thing to do here is to fix the linker, however this > patch also makes the output much easier to read as there are no > variable length data items here any more. Can you elaborate? I don't understand from this what problem you are trying to solve. Can you give a more concrete example? It should be impossible for the linker to change the alignment of anything _within_ an input section (bugs notwithstanding). The assembler might be doing something unexpected, particularly for Thumb kernels, but I don't know of any bug relating to this. Cheers ---Dave > > CC: Jon Medhurst <tixy@yxit.co.uk> > Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk> > --- > arch/arm/kernel/kprobes-test.c | 12 +++++------- > arch/arm/kernel/kprobes-test.h | 6 ++++-- > 2 files changed, 9 insertions(+), 9 deletions(-) > > diff --git a/arch/arm/kernel/kprobes-test.c b/arch/arm/kernel/kprobes-test.c > index 6cfa04d..a915ac3 100644 > --- a/arch/arm/kernel/kprobes-test.c > +++ b/arch/arm/kernel/kprobes-test.c > @@ -111,9 +111,7 @@ > * @ TESTCASE_START > * bl __kprobes_test_case_start > * @ start of inline data... > - * .ascii "mov r0, r7" @ text title for test case > - * .byte 0 > - * .align 2 > + * .word title_addr @ text title for test case > * > * @ TEST_ARG_REG > * .byte ARG_TYPE_REG > @@ -959,7 +957,7 @@ void __naked __kprobes_test_case_start(void) > __asm__ __volatile__ ( > "stmdb sp!, {r4-r11} \n\t" > "sub sp, sp, #"__stringify(TEST_MEMORY_SIZE)"\n\t" > - "bic r0, lr, #1 @ r0 = inline title string \n\t" > + "bic r0, lr, #3 @ r0 = inline title block \n\t" > "mov r1, sp \n\t" > "bl kprobes_test_case_start \n\t" > "bx r0 \n\t" > @@ -1336,15 +1334,15 @@ static unsigned long next_instruction(unsigned long pc) > return pc + 4; > } > > -static uintptr_t __used kprobes_test_case_start(const char *title, void *stack) > +static uintptr_t __used kprobes_test_case_start(const char **title, void *stack) > { > struct test_arg *args; > struct test_arg_end *end_arg; > unsigned long test_code; > > - args = (struct test_arg *)PTR_ALIGN(title + strlen(title) + 1, 4); > + args = (struct test_arg *)(title + 1); > > - current_title = title; > + current_title = *title; > current_args = args; > current_stack = stack; > > diff --git a/arch/arm/kernel/kprobes-test.h b/arch/arm/kernel/kprobes-test.h > index e28a869..a71db09 100644 > --- a/arch/arm/kernel/kprobes-test.h > +++ b/arch/arm/kernel/kprobes-test.h > @@ -113,9 +113,11 @@ struct test_arg_end { > "bl __kprobes_test_case_start \n\t" \ > /* don't use .asciz here as 'title' may be */ \ > /* multiple strings to be concatenated. */ \ > - ".ascii "#title" \n\t" \ > + ".pushsection .rodata \n\t" \ > + "9999: .ascii "#title" \n\t" \ > ".byte 0 \n\t" \ > - ".align 2 \n\t" > + ".popsection \n\t" \ > + ".word 9999b \n\t" > > #define TEST_ARG_REG(reg, val) \ > ".byte "__stringify(ARG_TYPE_REG)" \n\t" \ > -- > 1.7.10.4 > > > _______________________________________________ > linux-arm-kernel mailing list > linux-arm-kernel@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
On 26/07/13 15:02, Dave Martin wrote: > On Thu, Jul 25, 2013 at 10:08:57PM +0100, Ben Dooks wrote: >> When testing the kprobes test code with BE8, there is either an >> issue with the linker or how the code is being built. The issue >> is with embedding the title text as the first part of the test. >> >> Change to placing a pointer to .rodata with the text title in it >> for the test which seems to stop the issue of the alignment of >> the data following it being changed arbitrarily by the linker. >> >> The proper thing to do here is to fix the linker, however this >> patch also makes the output much easier to read as there are no >> variable length data items here any more. > > Can you elaborate? I don't understand from this what problem you are > trying to solve. Can you give a more concrete example? > > It should be impossible for the linker to change the alignment of > anything _within_ an input section (bugs notwithstanding). > > The assembler might be doing something unexpected, particularly > for Thumb kernels, but I don't know of any bug relating to this. The output in vmlinux looked ok, but there was some weird problems with some of the structures ending up un-swapped and some swapped. I do not have a concrete example of this as I didn't have time to work out exactly what was happening to the binary when it was being produced. Placing the strings in the .rodata section seemed to fix the whole issue which allowed me to turn the kprobes test on without them blowing up due to mis-reading their list of actions. > Cheers > ---Dave > >> >> CC: Jon Medhurst<tixy@yxit.co.uk> >> Signed-off-by: Ben Dooks<ben.dooks@codethink.co.uk> >> --- >> arch/arm/kernel/kprobes-test.c | 12 +++++------- >> arch/arm/kernel/kprobes-test.h | 6 ++++-- >> 2 files changed, 9 insertions(+), 9 deletions(-) >> >> diff --git a/arch/arm/kernel/kprobes-test.c b/arch/arm/kernel/kprobes-test.c >> index 6cfa04d..a915ac3 100644 >> --- a/arch/arm/kernel/kprobes-test.c >> +++ b/arch/arm/kernel/kprobes-test.c >> @@ -111,9 +111,7 @@ >> * @ TESTCASE_START >> * bl __kprobes_test_case_start >> * @ start of inline data... >> - * .ascii "mov r0, r7" @ text title for test case >> - * .byte 0 >> - * .align 2 >> + * .word title_addr @ text title for test case >> * >> * @ TEST_ARG_REG >> * .byte ARG_TYPE_REG >> @@ -959,7 +957,7 @@ void __naked __kprobes_test_case_start(void) >> __asm__ __volatile__ ( >> "stmdb sp!, {r4-r11} \n\t" >> "sub sp, sp, #"__stringify(TEST_MEMORY_SIZE)"\n\t" >> - "bic r0, lr, #1 @ r0 = inline title string \n\t" >> + "bic r0, lr, #3 @ r0 = inline title block \n\t" >> "mov r1, sp \n\t" >> "bl kprobes_test_case_start \n\t" >> "bx r0 \n\t" >> @@ -1336,15 +1334,15 @@ static unsigned long next_instruction(unsigned long pc) >> return pc + 4; >> } >> >> -static uintptr_t __used kprobes_test_case_start(const char *title, void *stack) >> +static uintptr_t __used kprobes_test_case_start(const char **title, void *stack) >> { >> struct test_arg *args; >> struct test_arg_end *end_arg; >> unsigned long test_code; >> >> - args = (struct test_arg *)PTR_ALIGN(title + strlen(title) + 1, 4); >> + args = (struct test_arg *)(title + 1); >> >> - current_title = title; >> + current_title = *title; >> current_args = args; >> current_stack = stack; >> >> diff --git a/arch/arm/kernel/kprobes-test.h b/arch/arm/kernel/kprobes-test.h >> index e28a869..a71db09 100644 >> --- a/arch/arm/kernel/kprobes-test.h >> +++ b/arch/arm/kernel/kprobes-test.h >> @@ -113,9 +113,11 @@ struct test_arg_end { >> "bl __kprobes_test_case_start \n\t" \ >> /* don't use .asciz here as 'title' may be */ \ >> /* multiple strings to be concatenated. */ \ >> - ".ascii "#title" \n\t" \ >> + ".pushsection .rodata \n\t" \ >> + "9999: .ascii "#title" \n\t" \ >> ".byte 0 \n\t" \ >> - ".align 2 \n\t" >> + ".popsection \n\t" \ >> + ".word 9999b \n\t" >> >> #define TEST_ARG_REG(reg, val) \ >> ".byte "__stringify(ARG_TYPE_REG)" \n\t" \ >> -- >> 1.7.10.4 >> >> >> _______________________________________________ >> linux-arm-kernel mailing list >> linux-arm-kernel@lists.infradead.org >> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
diff --git a/arch/arm/kernel/kprobes-test.c b/arch/arm/kernel/kprobes-test.c index 6cfa04d..a915ac3 100644 --- a/arch/arm/kernel/kprobes-test.c +++ b/arch/arm/kernel/kprobes-test.c @@ -111,9 +111,7 @@ * @ TESTCASE_START * bl __kprobes_test_case_start * @ start of inline data... - * .ascii "mov r0, r7" @ text title for test case - * .byte 0 - * .align 2 + * .word title_addr @ text title for test case * * @ TEST_ARG_REG * .byte ARG_TYPE_REG @@ -959,7 +957,7 @@ void __naked __kprobes_test_case_start(void) __asm__ __volatile__ ( "stmdb sp!, {r4-r11} \n\t" "sub sp, sp, #"__stringify(TEST_MEMORY_SIZE)"\n\t" - "bic r0, lr, #1 @ r0 = inline title string \n\t" + "bic r0, lr, #3 @ r0 = inline title block \n\t" "mov r1, sp \n\t" "bl kprobes_test_case_start \n\t" "bx r0 \n\t" @@ -1336,15 +1334,15 @@ static unsigned long next_instruction(unsigned long pc) return pc + 4; } -static uintptr_t __used kprobes_test_case_start(const char *title, void *stack) +static uintptr_t __used kprobes_test_case_start(const char **title, void *stack) { struct test_arg *args; struct test_arg_end *end_arg; unsigned long test_code; - args = (struct test_arg *)PTR_ALIGN(title + strlen(title) + 1, 4); + args = (struct test_arg *)(title + 1); - current_title = title; + current_title = *title; current_args = args; current_stack = stack; diff --git a/arch/arm/kernel/kprobes-test.h b/arch/arm/kernel/kprobes-test.h index e28a869..a71db09 100644 --- a/arch/arm/kernel/kprobes-test.h +++ b/arch/arm/kernel/kprobes-test.h @@ -113,9 +113,11 @@ struct test_arg_end { "bl __kprobes_test_case_start \n\t" \ /* don't use .asciz here as 'title' may be */ \ /* multiple strings to be concatenated. */ \ - ".ascii "#title" \n\t" \ + ".pushsection .rodata \n\t" \ + "9999: .ascii "#title" \n\t" \ ".byte 0 \n\t" \ - ".align 2 \n\t" + ".popsection \n\t" \ + ".word 9999b \n\t" #define TEST_ARG_REG(reg, val) \ ".byte "__stringify(ARG_TYPE_REG)" \n\t" \
When testing the kprobes test code with BE8, there is either an issue with the linker or how the code is being built. The issue is with embedding the title text as the first part of the test. Change to placing a pointer to .rodata with the text title in it for the test which seems to stop the issue of the alignment of the data following it being changed arbitrarily by the linker. The proper thing to do here is to fix the linker, however this patch also makes the output much easier to read as there are no variable length data items here any more. CC: Jon Medhurst <tixy@yxit.co.uk> Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk> --- arch/arm/kernel/kprobes-test.c | 12 +++++------- arch/arm/kernel/kprobes-test.h | 6 ++++-- 2 files changed, 9 insertions(+), 9 deletions(-)