Re: [PATCH] devmem2: ensure word is 32-bit, add support for 64-bit double

[Denys Dmytriyenko <denis@denix.org>]

On Tue, Jun 19, 2018 at 05:58:29PM -0700, Andre McCurdy wrote:
> On Tue, Jun 19, 2018 at 4:42 PM, Denys Dmytriyenko <denis@denix.org> wrote:
> > From: Denys Dmytriyenko <denys@ti.com>
> >
> > Since sizeof(unsigned long) can be 8-byte on 64-bit architectures, use
> > uint32_t instead for "word" access to always be 4-byte/32-bit long.
> >
> > Also introduce proper "double" 8-byte/64-bit access by using uint64_t.
> >
> > Signed-off-by: Denys Dmytriyenko <denys@ti.com>
> > ---
> >  meta-oe/recipes-support/devmem2/devmem2.bb         |  4 +-
> >  ...sure-word-is-32-bit-and-add-support-for-6.patch | 70 ++++++++++++++++++++++
> >  2 files changed, 73 insertions(+), 1 deletion(-)
> >  create mode 100644 meta-oe/recipes-support/devmem2/devmem2/0001-devmem.c-ensure-word-is-32-bit-and-add-support-for-6.patch
> >
> > diff --git a/meta-oe/recipes-support/devmem2/devmem2.bb b/meta-oe/recipes-support/devmem2/devmem2.bb
> > index c86eb2e..9bd1eb7 100644
> > --- a/meta-oe/recipes-support/devmem2/devmem2.bb
> > +++ b/meta-oe/recipes-support/devmem2/devmem2.bb
> > @@ -4,7 +4,9 @@ LIC_FILES_CHKSUM = "file://devmem2.c;endline=38;md5=a9eb9f3890384519f435aedf9862
> >  PR = "r7"
> >
> >  SRC_URI = "http://www.free-electrons.com/pub/mirror/devmem2.c;downloadfilename=devmem2-new.c \
> > -           file://devmem2-fixups-2.patch;apply=yes;striplevel=0"
> > +           file://devmem2-fixups-2.patch;apply=yes;striplevel=0 \
> > +           file://0001-devmem.c-ensure-word-is-32-bit-and-add-support-for-6.patch"
> > +
> >  S = "${WORKDIR}"
> >
> >  CFLAGS += "-DFORCE_STRICT_ALIGNMENT"
> > diff --git a/meta-oe/recipes-support/devmem2/devmem2/0001-devmem.c-ensure-word-is-32-bit-and-add-support-for-6.patch b/meta-oe/recipes-support/devmem2/devmem2/0001-devmem.c-ensure-word-is-32-bit-and-add-support-for-6.patch
> > new file mode 100644
> > index 0000000..ad8ae67
> > --- /dev/null
> > +++ b/meta-oe/recipes-support/devmem2/devmem2/0001-devmem.c-ensure-word-is-32-bit-and-add-support-for-6.patch
> > @@ -0,0 +1,70 @@
> > +From 1360a907879dd24041797a3b709d49aeac2ab444 Mon Sep 17 00:00:00 2001
> > +From: Denys Dmytriyenko <denys@ti.com>
> > +Date: Tue, 29 May 2018 16:55:42 -0400
> > +Subject: [PATCH] devmem.c: ensure word is 32-bit and add support for 64-bit
> > + double
> > +
> > +Signed-off-by: Denys Dmytriyenko <denys@ti.com>
> > +---
> > + devmem2.c | 23 +++++++++++++++++------
> > + 1 file changed, 17 insertions(+), 6 deletions(-)
> > +
> > +diff --git a/devmem2.c b/devmem2.c
> > +index 5845381..68131b2 100644
> > +--- a/devmem2.c
> > ++++ b/devmem2.c
> > +@@ -39,6 +39,7 @@
> > +
> > + #include <stdio.h>
> > + #include <stdlib.h>
> > ++#include <stdint.h>
> > + #include <unistd.h>
> > + #include <string.h>
> > + #include <errno.h>
> > +@@ -69,7 +70,7 @@ int main(int argc, char **argv) {
> > +       if(argc < 2) {
> > +               fprintf(stderr, "\nUsage:\t%s { address } [ type [ data ] ]\n"
> > +                       "\taddress : memory address to act upon\n"
> > +-                      "\ttype    : access operation type : [b]yte, [h]alfword, [w]ord\n"
> > ++                      "\ttype    : access operation type : [b]yte, [h]alfword, [w]ord, [d]ouble\n"
> 
> The busybox version of devmem (which seems to be the "upstream" for
> devmem development activity, such as it is) has already picked "l" as
> the command line option for 64bit values.

Hmm, busybox now does devmem? "who would have thunk it" :)
Still not upstream though...
I don't mind changing it to "l" (long?) to match busybox - will send v2.


> > +                       "\tdata    : data to be written\n\n",
> > +                       argv[0]);
> > +               exit(1);
> > +@@ -103,9 +104,14 @@ int main(int argc, char **argv) {
> > +                       read_result = *((unsigned short *) virt_addr);
> > +                       break;
> > +               case 'w':
> > +-                      data_size = sizeof(unsigned long);
> > ++                      data_size = sizeof(uint32_t);
> > +                       virt_addr = fixup_addr(virt_addr, data_size);
> > +-                      read_result = *((unsigned long *) virt_addr);
> > ++                      read_result = *((uint32_t *) virt_addr);
> > ++                      break;
> > ++              case 'd':
> > ++                      data_size = sizeof(uint64_t);
> > ++                      virt_addr = fixup_addr(virt_addr, data_size);
> > ++                      read_result = *((uint64_t *) virt_addr);
> > +                       break;
> > +               default:
> > +                       fprintf(stderr, "Illegal data type '%c'.\n", access_type);
> > +@@ -129,9 +135,14 @@ int main(int argc, char **argv) {
> > +                               read_result = *((unsigned short *) virt_addr);
> > +                               break;
> > +                       case 'w':
> > +-                              virt_addr = fixup_addr(virt_addr, sizeof(unsigned long));
> > +-                              *((unsigned long *) virt_addr) = write_val;
> > +-                              read_result = *((unsigned long *) virt_addr);
> > ++                              virt_addr = fixup_addr(virt_addr, sizeof(uint32_t));
> > ++                              *((uint32_t *) virt_addr) = write_val;
> > ++                              read_result = *((uint32_t *) virt_addr);
> > ++                              break;
> > ++                      case 'd':
> > ++                              virt_addr = fixup_addr(virt_addr, sizeof(uint64_t));
> > ++                              *((uint64_t *) virt_addr) = write_val;
> > ++                              read_result = *((uint64_t *) virt_addr);
> > +                               break;
> > +               }
> > +               sprintf(fmt_str, "Write at address 0x%%08lX (%%p): 0x%%0%dlX, "
> > +--
> > +2.7.4
> > +
> > --
> > 2.7.4
> >
> > --
> > _______________________________________________
> > Openembedded-devel mailing list
> > Openembedded-devel@lists.openembedded.org
> > http://lists.openembedded.org/mailman/listinfo/openembedded-devel
>