[PATCH] (2 commits squashed into one) Beagle BSP for review
Joel Sherrill
joel.sherrill at oarcorp.com
Wed Aug 20 16:46:48 UTC 2014
On 8/20/2014 11:23 AM, Ritesh Harjani wrote:
> Hi Ben,
>
> Great work I must admit.
>
> I just have few question/suggestion. I am new to the community so high
> chances that I might be wrong here, but I noticed something so I
> thought I should tell/ask.
>
>
> diff --git a/c/src/lib/libbsp/arm/beagle/include/bsp.h
> b/c/src/lib/libbsp/arm/beagle/include/bsp.h
> new file mode 100644
> index 0000000..fc001dd1
> --- /dev/null
> +++ b/c/src/lib/libbsp/arm/beagle/include/bsp.h
> ...
> ...
> +/* Data synchronization barrier */
> +static inline void dsb(void)
> +{
> + asm volatile("dsb" : : : "memory");
> +}
> +
> +/* Instruction synchronization barrier */
> +static inline void isb(void)
> +{
> + asm volatile("isb" : : : "memory");
> +}
> ...
> ...
> I guess these similar barrier functions are already present
> in cpukit/score/cpu/arm/rtems/score/cpu.h.
> So, do we need to add this again ?
>
Good catch. Ben they are in cpukit/score/cpu/arm/rtems/score/cpu.h which
is implicitly included.
I can't there there is a CPU independent wrapper for synchronizations
like this.
>
> +/* flush data cache */
> +static inline void flush_data_cache(void)
> +{
> + asm volatile("mov r0, #0; mcr p15, #0, r0, c7, c10, #4" : : :
> "memory");
> +}
>
> the name of this function is confusing to me. This is not a flush
> operation right ? Following link says it is a data synchronization
> operation.
> http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ddi0344c/Babhejba.html
>
> Please correct me if I am wrong at any place.
>
> Thanks
> Ritesh
>
>
>
> On Wed, Aug 20, 2014 at 7:50 PM, Gedare Bloom <gedare at rtems.org
> <mailto:gedare at rtems.org>> wrote:
>
> Ben, As far as getting this merged, all of my comments can be done as
> a follow-on commit. -Gedare
>
> On Thu, Jul 24, 2014 at 4:28 PM, Ben Gras <beng at shrike-systems.com
> <mailto:beng at shrike-systems.com>> wrote:
> > Thanks for the fast & detailed review. Let me get back to it/you.
> >
> > In the meantime, any other feedback? From anyone I mean.
> >
> >
> >
> > On Thu, Jul 24, 2014 at 4:45 PM, Gedare Bloom <gedare at rtems.org
> <mailto:gedare at rtems.org>> wrote:
> >>
> >> Hi Ben,
> >> Great work. I have a few comments. I skipped the i2c.h and i2c.c
> >> files. Most of my comments are about style and a few requests to
> >> refactor some of the larger files. The refactoring can be added to
> >> your TODO if you like. Please fix the style issues if it is not a
> >> burden.
> >>
> >> +++ b/c/src/lib/libbsp/arm/beagle/README
> >> +$ ../claas-rtems/configure --target=arm-rtems4.11
> >> --enable-rtemsbsp="beaglebonewhite beagleboardxm"
> >> Replace claas-rtems with rtems. If RSB support is available, make a
> >> note about it.
> >>
> >> +++ b/c/src/lib/libbsp/arm/beagle/TODO
> >> [...]
> >> open:
> >> + . how to handle the interrupt?
> >>
> >> What does this mean?
> >>
> >> +++ b/c/src/lib/libbsp/arm/beagle/clock.c
> >> Why is the entire file ifdef'd on ARM_MULTILIB_ARCH_V4?
> >>
> >> It might be sensible to put the struct definitions in a .h file if
> >> these omap registers might be re-usable.
> >>
> >> +static struct omap_timer_registers regs_v2 = {
> >> This might be better to put behind an #if IS_AM335X since it is not
> >> used otherwise?
> >>
> >> +
> >> +
> >> +
> >> Avoid more than one blank line in a row.
> >>
> >> +static int done = 0;
> >> It would be nice if you got rid of this, but otherwise give it
> a more
> >> useful name like "mmio_init_done"
> >>
> >> +static void
> beagle_clock_handler_install(rtems_interrupt_handler isr)
> >> + if (sc != RTEMS_SUCCESSFUL) {
> >> + rtems_fatal_error_occurred(0xdeadbeef);
> >> I think there is some capabilities in ARM for bsp fatal error codes
> >> now. They should be preferred to be used to help debug these fatal
> >> conditions.
> >>
> >> +static inline uint32_t
> beagle_clock_nanoseconds_since_last_tick(void)
> >> + return (read_frc() - (uint64_t) last_tick_nanoseconds) *
> 1000000000
> >> / FRCLOCK_HZ;
> >> This line is > 80 characters, please break it or shrink it.
> >>
> >> +++ b/c/src/lib/libbsp/arm/beagle/console/console-config.c
> >> +#define CONSOLE_UART_LSR (*(volatile unsigned int
> >> *)(BSP_CONSOLE_UART_BASE+0x14))
> >> Line > 80 characters, even with the spacing modified.
> >>
> >> +static void beagle_console_init(void)
> >>
> >> + while ((CONSOLE_SYSS & 1) == 0)
> >> + ;
> >> Is this a fatal loop or is it waiting for hardware to clear
> something?
> >>
> >> + if ((CONSOLE_LSR & (CONSOLE_LSR_THRE | CONSOLE_LSR_TEMT)) ==
> >> CONSOLE_LSR_THRE) {
> >> Again > 80 characters. Is the test logically equivalent to: if (
> >> (CONSOLE_LSR & CONSOLE_LSR_THRE) == CONSOLE_LSR_THRE)
> >>
> >> + while ((CONSOLE_LSR & CONSOLE_LSR_TEMT) == 0)
> >> + ;
> >> Is this a fatal loop or is it waiting for hardware?
> >>
> >> +++ b/c/src/lib/libbsp/arm/beagle/include/bsp.h
> >> This bsp.h is really long. Probably it should be refactored
> into other
> >> headers, including non-public ones.
> >>
> >> +static inline void dsb(void)
> >> +{
> >> + asm volatile("dsb" : : : "memory");
> >> Fix the indentation.
> >>
> >> +static inline void flush_data_cache(void)
> >> Perhaps this should be using _CPU_cache_flush_entire_data()?
> Perhaps
> >> there is a difference in that the cache manager code flushes and
> >> "cleans" the cache...
> >>
> >> +
> >> +
> >> +
> >> +
> >> Excess newlines. Done a few places in this file.
> >>
> >> The comments following the defines for various AM33X_INT_ values go
> >> off the end of the 80 column character width. Same for some other
> >> comments following defines for OMAP3_TIMER, AM33X_DMTIMER1, and
> >> AM335X_TIMER_. And further below for the CM_ WKUP and CM_PER_TIMER7
> >> defines, and CLKSEL_TIMER1MS_CLK_SEL_SEL5.
> >>
> >> +#define OMAP3_TCLR_AR (1 << 1) /* Autoreload or
> one-shot mode */
> >> +#define OMAP3_TCLR_PRE (1 << 5) /* Prescaler on */
> >> +#define OMAP3_TCLR_PTV 2
> >> This PTV is odd compared to the other defines here. Is it 2 ==
> (1<<1),
> >> or is there a typo here?
> >>
> >> Tabs are used in the OMAP3_CM_ defines, it should be space
> characters.
> >> Also tabs are used in the read/write actlr, ttbcr, dacr, rrbr0
> >> functions and the refresh_tlb function.
> >>
> >> +/* i2c stuff */
> >> +typedef struct {
> >> ...
> >> +} beagle_i2c;
> >> Shouldn't this go in beagle/include/i2c.h?
> >>
> >> All of this mmu handling code should be refactored. Where
> possible, it
> >> should use the existing code in arm-cp15.h
> >>
> >> +++ b/c/src/lib/libbsp/arm/beagle/include/i2c.h
> >> This header defines static, non-inline functions. This doesn't
> make sense.
> >>
> >> +++ b/c/src/lib/libbsp/arm/beagle/irq.c
> >> +static int irqs_enabled[BSP_INTERRUPT_VECTOR_MAX+1];
> >> This is an array of 512 bytes. You could use a bit vector
> comprising 4
> >> unsigned ints for the same purpose.
> >>
> >> +volatile static int level = 0;
> >> Unused variable?
> >>
> >> +static uint32_t get_mir_reg(int vector, uint32_t *mask)
> >> + if(vector < 0) while(1) ;
> >> Make this a fatal error?
> >>
> >> + if(vector < 32) return OMAP3_INTCPS_MIR0;
> >> + if(vector < 32) return OMAP3_INTCPS_MIR0;
> >> duplicate code.
> >>
> >> + while(1) ;
> >> Make this a fatal error?
> >>
> >> +rtems_status_code bsp_interrupt_facility_initialize(void)
> >> + mmio_write(omap_intr.base + OMAP3_INTCPS_SYSCONFIG,
> >> OMAP3_SYSCONFIG_AUTOIDLE);
> >> Line length > 80.
> >>
> >> +++ b/c/src/lib/libbsp/arm/beagle/startup/bspstartmmu.c
> >>
> >> +//static uint32_t pagetable[ARM_SECTIONS] __attribute__((aligned
> >> (1024*16)));
> >> commented-out.. delete it?
> >>
> >> +BSP_START_TEXT_SECTION void beagle_setup_mmu_and_cache(void)
> >> __attribute__ ((weak));
> >> More than 80 characters.
> >>
> >> diff --git a/c/src/lib/libbsp/bfin/acinclude.m4
> >> b/c/src/lib/libbsp/bfin/acinclude.m4
> >> index ab6082e..828fd89 100644
> >> --- a/c/src/lib/libbsp/bfin/acinclude.m4
> >> +++ b/c/src/lib/libbsp/bfin/acinclude.m4
> >> diff --git a/c/src/lib/libbsp/powerpc/acinclude.m4
> >> b/c/src/lib/libbsp/powerpc/acinclude.m4
> >> index 6442399..e46fa2b 100644
> >> --- a/c/src/lib/libbsp/powerpc/acinclude.m4
> >> +++ b/c/src/lib/libbsp/powerpc/acinclude.m4
> >> Don't include these changes. Check your tool versions, and if the
> >> correct version of tools does this, provide a separate patch for
> >> generated files.
> >>
> >> -Gedare
> >>
> >> On Wed, Jul 23, 2014 at 9:00 PM, Ben Gras
> <beng at shrike-systems.com <mailto:beng at shrike-systems.com>> wrote:
> >> > All,
> >> >
> >> > Full details on how to reproduce all the work from source
> repositories
> >> > to
> >> > scripts & utilities to write a complete sd card booting RTEMS
> and test
> >> > the
> >> > whole thing:
> >> >
> >> >
> >> >
> http://www.shrike-systems.com/beagleboard-xm-beaglebone-black-and-everything-else-rtems-on-the-beagles.html
> >> >
> >> > I am submitting the attached patch for review for merging. If
> accepted
> >> > for
> >> > merging, please use the top two commits on
> >> >
> >> > https://github.com/bengras/rtems/tree/beaglebone-wip
> >> >
> >> > which have the same net effect but preserve Claas' work
> because of the
> >> > earlier commit. The squashed version attached is for more
> convenient
> >> > review.
> >> >
> >> > I was ironing out more wrinkles but given recent interest it
> seems
> >> > smarter
> >> > to merge sooner and keep polishing from mainline.
> Nevertheless I have
> >> > put a
> >> > lot of work into getting it into good shape already.
> >> >
> >> > I have rebased everything on the very latest master and verified
> >> >
> >> > That building all the tools and utilities from scratch work,
> using the
> >> > RTEMS
> >> > Source Builder repository (Ubuntu + FreeBSD).
> >> > That building the beaglebone and bbxm BSPs and linking them
> with all the
> >> > testsuite programs works (Ubuntu + FreeBSD).
> >> > That the beaglexm-emulating linaro qemu executes all of those
> tests
> >> > properly, invoked using a single command line with the
> scripts in the
> >> > RTEMS
> >> > tools repository, even though not all pass currently (Ubuntu
> + FreeBSD).
> >> > That loading & running over JTAG works, both interactively
> with gdb and
> >> > in a
> >> > batch using gdb and the test runner.
> >> > That running RTEMS executables using u-boot on the
> beaglebones from sd
> >> > card
> >> > work; both with and without MMU enabled at RTEMS start time.
> >> > That Claas' earlier commit builds.
> >> >
> >> > Thanks so far to Chris and Brandon for help, support,
> instructions and
> >> > advice in various forms :)
> >> >
> >> > Test results on qemu:
> >> > Passed: 497 Failed: 3 Timeouts: 1 Invalid: 0
> >> >
> >> > The test results on bbxm over jtag (older):
> >> > Passed: 475 Failed: 7 Timeouts: 10 Invalid: 0
> >> >
> >> > I want to iron out more wrinkles and build support (ethernet)
> but giving
> >> > the
> >> > bsp more exposure and having it in mainline so i don't have
> to keep
> >> > rebasing
> >> > & testing would be nice at this point.
> >> >
> >> >
> >> >
> >> >
> >> > _______________________________________________
> >> > devel mailing list
> >> > devel at rtems.org <mailto:devel at rtems.org>
> >> > http://lists.rtems.org/mailman/listinfo/devel
> >
> >
> _______________________________________________
> devel mailing list
> devel at rtems.org <mailto:devel at rtems.org>
> http://lists.rtems.org/mailman/listinfo/devel
>
>
--
Joel Sherrill, Ph.D. Director of Research & Development
joel.sherrill at OARcorp.com On-Line Applications Research
Ask me about RTEMS: a free RTOS Huntsville AL 35805
Support Available (256) 722-9985
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/devel/attachments/20140820/82140217/attachment-0001.html>
More information about the devel
mailing list