From: Leo Liang <ycliang@andestech.com>
To: u-boot@lists.denx.de
Subject: [PATCH 1/1] cmd: exception: unaligned data access on RISC-V
Date: Thu, 6 Aug 2020 10:56:48 +0800 [thread overview]
Message-ID: <20200806025648.GA18142@andestech.com> (raw)
In-Reply-To: <CAN5B=e+Zs_KEWuMBpvmQpLQrB3_zYn6Lsy5=ncJCqKgHHqXAHg@mail.gmail.com>
Hi Heinrich,
On Thu, Aug 06, 2020 at 10:15:28AM +0800, Rick Chen wrote:
> Hi Heinrich
>
> > >> From: Heinrich Schuchardt [mailto:xypron.glpk at gmx.de]
> > >> Sent: Tuesday, August 04, 2020 7:10 PM
> > >> To: Rick Jian-Zhi Chen(???)
> > >> Cc: u-boot at lists.denx.de; Heinrich Schuchardt
> > >> Subject: [PATCH 1/1] cmd: exception: unaligned data access on RISC-V
> > >>
> > >> The command 'exception' can be used to test the handling of exceptions.
> > >>
> > >> Currently the exception command only allows to create an illegal instruction exception on RISC-V.
> > >>
> > >> Provide a sub-command 'exception unaligned' to cause a misaligned load address exception.
> > >>
> > >> Adjust the online help for 'exception undefined'.
> > >>
> > >> Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
> > >> ---
> > >> cmd/riscv/exception.c | 16 +++++++++++++++-
> > >> 1 file changed, 15 insertions(+), 1 deletion(-)
> > >>
> > >> diff --git a/cmd/riscv/exception.c b/cmd/riscv/exception.c index 3c8dbbec0e..53159531d9 100644
> > >> --- a/cmd/riscv/exception.c
> > >> +++ b/cmd/riscv/exception.c
> > >> @@ -8,6 +8,17 @@
> > >> #include <common.h>
> > >> #include <command.h>
> > >>
> > >> +static int do_unaligned(struct cmd_tbl *cmdtp, int flag, int argc,
> > >> + char *const argv[])
> > >> +{
> > >> + asm volatile (
> > >> + "auipc a1, 0\n"
> > >> + "ori a1, a1, 3\n"
> > >> + "lw a2, (0)(a1)\n"
> > >> + );
> > >> + return CMD_RET_FAILURE;
> > >
> > > I suggest that we can modify it as below to print the unaligned
> > > address and data for more debug information.
> > >
> > > int ret = 0;
> > > int addr = 0;
> > >
> > > asm volatile (
> > > "auipc a1, 0\n"
> > > "ori %0, a1, 3\n"
> > > "lw %1, (0)(a1)\n"
> > > : "=r" (addr), "=r" (ret)
> > > :
> > > : "memory"
> > > );
> > > printf("unaligned addr 0x%x , ret 0x%x\n",addr,ret);
> >
> > Thanks for reviewing.
> >
> > The printf statement will never be reached if the system does not
> > support unaligned access. Why should anybody care about the actual
> > addresses?
> >
> > A reasonable message here might be:
> >
> > printf("The system supports unaligned access.\n");
>
> Yes, this print message without printing the address and data is good for me.
> My intention is that print something better than print nothing.
> For some junior users they maybe expect that the shell will echo
> something when input some commands.
>
> >
> > >
> > > return CMD_RET_SUCCESS;
> > > ====================================
> > > So if run in S-Mode, it will work as below:
> > >
> > > RISC-V # exception unaligned
> >
> > Do we ever run the U-Boot shell in S-mode? Shouldn't we always drop to
> > M-mode before reaching the shell?
> >
In addition to Rick's reponse, I'd like to add that to my knowledge,
U-Boot-SPL will bring up U-Boot-proper in S-mode, so you will get a shell in S-mode.
> > If I am in the U-Boot shell, why does S-mode not print out the registers
> > like M-Mode does? What is missing?
>
> In S-mode, Opensbi will take care this unaligned exception.
> Even in M-mode if the HW support this unaligned access, it will not
> trigger this exception. So it shall return CMD_RET_SUCCESS here.
And as explained, this S-mode exception will be taken care of by M-mode runtime (OpenSBI),
so the info of the registers will not be printed out.
Concluded that I think nothing is missing here.
Best regards,
Leo
>
> Thanks,
> Rick
>
> >
> > Best regards
> >
> > Heinrich
> >
> > > unaligned addr 0x3ff92fd7 , ret 0x35e59300
> > > RISC-V #
> > >
> > >
> > > (gdb) x/4x 0x3ff92fd0
> > > 0x3ff92fd0: 0x7ac362ef 0x00000597 0x0035e593 0xc5174190
> > > (gdb)
> > >
> > > ====================================
> > > If run in M-Mode, it will work as below:
> > >
> > > RISC-V # exception unaligned
> > > Unhandled exception: Load address misaligned
> > > EPC: 000000003ff92fdc RA: 000000003ff93032 TVAL: 000000003ff92fd7
> > > EPC: 0000000000009fdc RA: 000000000000a032 reloc adjusted
> > >
> > > ### ERROR ### Please RESET the board ###
> > >
> > > Thanks,
> > > Rick
> > >
> > >> +}
> > >> +
> > >> static int do_undefined(struct cmd_tbl *cmdtp, int flag, int argc,
> > >> char *const argv[])
> > >> {
> > >> @@ -16,6 +27,8 @@ static int do_undefined(struct cmd_tbl *cmdtp, int flag, int argc, }
> > >>
> > >> static struct cmd_tbl cmd_sub[] = {
> > >> + U_BOOT_CMD_MKENT(unaligned, CONFIG_SYS_MAXARGS, 1, do_unaligned,
> > >> + "", ""),
> > >> U_BOOT_CMD_MKENT(undefined, CONFIG_SYS_MAXARGS, 1, do_undefined,
> > >> "", ""),
> > >> };
> > >> @@ -23,7 +36,8 @@ static struct cmd_tbl cmd_sub[] = { static char exception_help_text[] =
> > >> "<ex>\n"
> > >> " The following exceptions are available:\n"
> > >> - " undefined - undefined instruction\n"
> > >> + " undefined - illegal instruction\n"
> > >> + " unaligned - load address misaligned\n"
> > >> ;
> > >>
> > >> #include <exception.h>
> > >> --
> > >> 2.27.0
> >
prev parent reply other threads:[~2020-08-06 2:56 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-08-04 11:09 [PATCH 1/1] cmd: exception: unaligned data access on RISC-V Heinrich Schuchardt
[not found] ` <752D002CFF5D0F4FA35C0100F1D73F3FA47309C8@ATCPCS16.andestech.com>
2020-08-05 8:47 ` Rick Chen
2020-08-05 10:26 ` Heinrich Schuchardt
2020-08-06 2:15 ` Rick Chen
2020-08-06 2:56 ` Leo Liang [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20200806025648.GA18142@andestech.com \
--to=ycliang@andestech.com \
--cc=u-boot@lists.denx.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox