| Message ID | CAN8C2CoYZcb1VHS1JmZ_5Bv5Q0=LUMBBE-ax7=Twvy=1qEL+0A@mail.gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | [Arm] Remaining difference in crt0 code | expand |
On 12/04/2019 21:42, Alexander Fedotov wrote: > Hello Richard > > There is still some difference in crt0.S code in libgloss and libc > respectively. Could you help me figure out what version is correct so > we can get rid of this. > I don't think we can clean this up file at a time. Trying to do so will likely create inconsistencies in the sources as some files will be patched and others not. "git blame" and "git show" will be your friends here... For example, taking the first hunk: > Differences are following: > > --- newlib/libc/sys/arm/crt0.S 2019-04-12 23:28:32.237488800 +0300 > +++ libgloss/arm/crt0.S 2019-04-12 23:28:32.237488800 +0300 > @@ -79,6 +79,19 @@ > .fnstart > #endif > > + /* __ARM_ARCH_PROFILE is defined from GCC 4.8 onwards, however > __ARM_ARCH_7A > + has been defined since 4.2 onwards, which is when v7-a support was added > + and hence 'A' profile support was added in the compiler. Allow for this > + file to be built with older compilers. We only call this for A profile > + cores. */ > +#if defined (__ARM_ARCH_7A__) || (__ARM_ARCH_PROFILE == 'A') > +/* The init hook does not use the stack and is called before the > stack has been set up. */ > +#ifdef ARM_RDI_MONITOR > + bl _rdimon_hw_init_hook > + .weak FUNCTION (_rdimon_hw_init_hook) > +#endif > +#endif > + "git blame" shows that this hunk is derived from two commits 99be2bc4ff and then shortly afterwards 639951dda7. "git show" can then be used to view these individual patches and you can also use the date information to identify the post to the newlib mailing list, which may contain additional information about the patch. Unwinding all of this is going to be a bit tedious, I fear, so I really appreciate any assistance you can give. R.
Ok, thanks. On Mon, Apr 15, 2019 at 12:14 PM Richard Earnshaw (lists) <Richard.Earnshaw@arm.com> wrote: > > On 12/04/2019 21:42, Alexander Fedotov wrote: > > Hello Richard > > > > There is still some difference in crt0.S code in libgloss and libc > > respectively. Could you help me figure out what version is correct so > > we can get rid of this. > > > > I don't think we can clean this up file at a time. Trying to do so will > likely create inconsistencies in the sources as some files will be > patched and others not. > > "git blame" and "git show" will be your friends here... For example, > taking the first hunk: > > > Differences are following: > > > > --- newlib/libc/sys/arm/crt0.S 2019-04-12 23:28:32.237488800 +0300 > > +++ libgloss/arm/crt0.S 2019-04-12 23:28:32.237488800 +0300 > > @@ -79,6 +79,19 @@ > > .fnstart > > #endif > > > > + /* __ARM_ARCH_PROFILE is defined from GCC 4.8 onwards, however > > __ARM_ARCH_7A > > + has been defined since 4.2 onwards, which is when v7-a support was added > > + and hence 'A' profile support was added in the compiler. Allow for this > > + file to be built with older compilers. We only call this for A profile > > + cores. */ > > +#if defined (__ARM_ARCH_7A__) || (__ARM_ARCH_PROFILE == 'A') > > +/* The init hook does not use the stack and is called before the > > stack has been set up. */ > > +#ifdef ARM_RDI_MONITOR > > + bl _rdimon_hw_init_hook > > + .weak FUNCTION (_rdimon_hw_init_hook) > > +#endif > > +#endif > > + > > "git blame" shows that this hunk is derived from two commits 99be2bc4ff > and then shortly afterwards 639951dda7. "git show" can then be used to > view these individual patches and you can also use the date information > to identify the post to the newlib mailing list, which may contain > additional information about the patch. > > Unwinding all of this is going to be a bit tedious, I fear, so I really > appreciate any assistance you can give. > > R.
Well, at first glance all the differences are related to semihosting v2. I think Tamar can help me with that (if he doesn't mind). Especially I do see his comment in https://cygwin.com/git/gitweb.cgi?p=newlib-cygwin.git;a=commit;h=6d6a623e7d8eb9e521bdbd73a7eafdd482678cea "This patch does this, I have not updated newlib/libc/* as that is quite out of date already. A proper sync is needed in order to get things back in sync. A different patch for this would be best." I can say time " The time has come" :) Moreover Tamar has a commit rights so it can be done much faster. Alex On Mon, Apr 15, 2019 at 1:17 PM Alexander Fedotov <alfedotov@gmail.com> wrote: > > Ok, thanks. > > On Mon, Apr 15, 2019 at 12:14 PM Richard Earnshaw (lists) > <Richard.Earnshaw@arm.com> wrote: > > > > On 12/04/2019 21:42, Alexander Fedotov wrote: > > > Hello Richard > > > > > > There is still some difference in crt0.S code in libgloss and libc > > > respectively. Could you help me figure out what version is correct so > > > we can get rid of this. > > > > > > > I don't think we can clean this up file at a time. Trying to do so will > > likely create inconsistencies in the sources as some files will be > > patched and others not. > > > > "git blame" and "git show" will be your friends here... For example, > > taking the first hunk: > > > > > Differences are following: > > > > > > --- newlib/libc/sys/arm/crt0.S 2019-04-12 23:28:32.237488800 +0300 > > > +++ libgloss/arm/crt0.S 2019-04-12 23:28:32.237488800 +0300 > > > @@ -79,6 +79,19 @@ > > > .fnstart > > > #endif > > > > > > + /* __ARM_ARCH_PROFILE is defined from GCC 4.8 onwards, however > > > __ARM_ARCH_7A > > > + has been defined since 4.2 onwards, which is when v7-a support was added > > > + and hence 'A' profile support was added in the compiler. Allow for this > > > + file to be built with older compilers. We only call this for A profile > > > + cores. */ > > > +#if defined (__ARM_ARCH_7A__) || (__ARM_ARCH_PROFILE == 'A') > > > +/* The init hook does not use the stack and is called before the > > > stack has been set up. */ > > > +#ifdef ARM_RDI_MONITOR > > > + bl _rdimon_hw_init_hook > > > + .weak FUNCTION (_rdimon_hw_init_hook) > > > +#endif > > > +#endif > > > + > > > > "git blame" shows that this hunk is derived from two commits 99be2bc4ff > > and then shortly afterwards 639951dda7. "git show" can then be used to > > view these individual patches and you can also use the date information > > to identify the post to the newlib mailing list, which may contain > > additional information about the patch. > > > > Unwinding all of this is going to be a bit tedious, I fear, so I really > > appreciate any assistance you can give. > > > > R.
--- newlib/libc/sys/arm/crt0.S 2019-04-12 23:28:32.237488800 +0300 +++ libgloss/arm/crt0.S 2019-04-12 23:28:32.237488800 +0300 @@ -79,6 +79,19 @@ .fnstart #endif + /* __ARM_ARCH_PROFILE is defined from GCC 4.8 onwards, however __ARM_ARCH_7A + has been defined since 4.2 onwards, which is when v7-a support was added + and hence 'A' profile support was added in the compiler. Allow for this + file to be built with older compilers. We only call this for A profile + cores. */ +#if defined (__ARM_ARCH_7A__) || (__ARM_ARCH_PROFILE == 'A') +/* The init hook does not use the stack and is called before the stack has been set up. */ +#ifdef ARM_RDI_MONITOR + bl _rdimon_hw_init_hook + .weak FUNCTION (_rdimon_hw_init_hook) +#endif +#endif + /* Start by setting up a stack. */
Hello Richard There is still some difference in crt0.S code in libgloss and libc respectively. Could you help me figure out what version is correct so we can get rid of this. Differences are following: #ifdef ARM_RDP_MONITOR /* Issue Demon SWI to read stack info. */ @@ -105,13 +118,20 @@ bkpt AngelSWI #elif defined(__thumb2__) /* We are in thumb mode for startup on armv7 architectures. */ - AngelSWIAsm AngelSWI + AngelSWIAsm (AngelSWI) #else /* We are always in ARM mode for startup on pre armv7 archs. */ - AngelSWIAsm AngelSWI_ARM + AngelSWIAsm (AngelSWI_ARM) #endif ldr r0, .LC0 /* Point at values read. */ + /* Set __heap_limit. */ + ldr r1, [r0, #4] + cmp r1, #0 + beq .LC33 + ldr r2, =__heap_limit + str r1, [r2] +.LC33: ldr r1, [r0, #0] cmp r1, #0 bne .LC32 @@ -279,13 +299,10 @@ movs r1, r0 #else movs r0, #AngelSWI_Reason_GetCmdLine - adr r1, .LC30 /* Space for command line. */ -#ifdef THUMB_VXM - bkpt AngelSWI -#else - AngelSWIAsm AngelSWI -#endif + ldr r1, .LC30 /* Space for command line. */ + AngelSWIAsm (AngelSWI) ldr r1, .LC30 + ldr r1, [r1] #endif /* Parse string at r1. */ movs r0, #0 /* Count of arguments so far. */ @@ -309,7 +326,7 @@ beq .LC10 /* See whether we are scanning a string. */ - cmp r3, #'\"' + cmp r3, #'"' #ifdef __thumb__ beq .LC20 cmp r3, #'\'' @@ -396,8 +413,17 @@ for _fini to be called at program exit. */ movs r4, r0 movs r5, r1 +#ifdef _LITE_EXIT + /* Make reference to atexit weak to avoid unconditionally pulling in + support code. Refer to comments in __atexit.c for more details. */ + .weak FUNCTION(atexit) + ldr r0, .Latexit + cmp r0, #0 + beq .Lweak_atexit +#endif ldr r0, .Lfini bl FUNCTION (atexit) +.Lweak_atexit: bl FUNCTION (_init) movs r0, r4 movs r1, r5 @@ -469,13 +495,19 @@ .LC2: .word __bss_end__ #ifdef __USES_INITFINI__ +#ifdef _LITE_EXIT +.Latexit: + .word FUNCTION(atexit) + + /* Weak reference _fini in case of lite exit. */ + .weak FUNCTION(_fini) +#endif .Lfini: .word FUNCTION(_fini) #endif #ifdef ARM_RDI_MONITOR .LC30: - .word CommandLine - .word 255 + .word AngelSWIArgs .LC31: .word __end__ @@ -488,6 +520,9 @@ __stack_base__: .word 0 StackLimit: .word 0 CommandLine: .space 256,0 /* Maximum length of 255 chars handled. */ +AngelSWIArgs: + .word CommandLine + .word 255 #endif #ifdef __pe__ Best regards, Alex