From: Trd thg <trdthg47@gmail.com>
To: Alistair Francis <alistair23@gmail.com>
Cc: qemu-riscv@nongnu.org, qemu-devel@nongnu.org,
MingZhu Yan <yanmingzhu@iscas.ac.cn>
Subject: Fwd: Re: [PATCH] hw/char/riscv_htif: Fix htif_mm_write that causes infinite loop in ACT.
Date: Wed, 16 Oct 2024 18:32:27 +0800 [thread overview]
Message-ID: <CAPVrrNJTasrpMxbWFacFGQkvgf1LRyZoJpn1Z-9E-_iXmGfuyA@mail.gmail.com> (raw)
In-Reply-To: <CAPVrrNLf449o57ZsC-T5qZeWJcQvvfOBVd+U+m60dwzvf3RBxg@mail.gmail.com>
Sorry about the email style, I'm not familiar with it yet
> It's probably worth including this in the commit message.
Agree, I'll do it.
> It would be good to reference the sail implementation and the
justification for the change there
Sail implementation is probably here:
https://github.com/riscv/sail-riscv/blob/master/model/riscv_platform.sail#L340
- related commit:
https://github.com/riscv/sail-riscv/commit/848312ce7c59fa08c304cab4b4e8060c67d5dfc9
The following is the "infinite loop" part is exported by objdump from
the test ELF of the add instruction generated from ACT
```txt
00000000800082a0 <write_tohost>:
800082a0: 00001f17 auipc t5,0x1
800082a4: d61f2023 sw ra,-672(t5) # 80009000 <tohost>
800082a8: ff9ff06f j 800082a0 <write_tohost>
```
QEMU cannot respond to the above `sw` behavior, It makes no
distinction regarding the size written,
I guess that currently qemu will only respond to the writing to the
high 32 bits and low 32 bits of tohost twice in succession.
I think the behavior here is very strange, and you can find
aswaterman's description of HTIF(include RV32) at:
https://github.com/riscv-software-src/riscv-isa-sim/issues/364#issuecomment-607657754
So this patch try to distinguish these cases based on size and addr.
About the sail impl, there are some related discussions at:
https://github.com/riscv/sail-riscv/issues/218
I made some summaries:
- The implementation of sail as a workaround is not very ideal; we
should follow the experience of spike
I checked the impl of spike. Although spike handles tohost in a
syscall manner, I don't think this means that spike is better.
Compared to sail, it does not distinguish whether the write to
tohost is 4 bytes or 8 bytes, but at least in act, it always works
properly.
- spike code:
https://github.com/riscv-software-src/riscv-isa-sim/blob/master/fesvr/htif.cc#L265
- HTIF has been deprecated
I don't know if this is true or not, because I haven't found any
official comment.
However, since ACT is still using htif and is not expected to change
for quite a long time, and no new solutions have emerged, we should go
ahead and implement it
- What are the advantages of QEMU supporting ACT
- do cross validation with sail/spike, and contributors can use ACT
to verify the correctness when adding new instructions to QEMU. (we do
have this need.)
> If you think that the benefits of supporting ACT are not obvious,
then I think we can remove all the code related to supporting ACT
(should be the commit mentioned above)
What do you think about this? Thank you!
Alistair Francis <alistair23@gmail.com> 于2024年10月16日周三 13:27写道：
>
> On Mon, Oct 14, 2024 at 8:08 PM 阎明铸 <yanmingzhu@iscas.ac.cn> wrote:
> >
> > Thank you for your reply and I'm sorry that I didn't explain it clearly.
> >
> > - ACT is an official riscv test suite to check the riscv support of the DUT(device under test).
>
> It's probably worth including this in the commit message.
>
> > - Currently ACT support using [sail-riscv](https://github.com/riscv/sail-riscv)(default) or [spike](https://github.com/riscv-software-src/riscv-isa-sim)
> > - QEMU is not supported yet，but someone made a commit: [commit](https://github.com/qemu/qemu/commit/66247edc8b6fb36d6b905babcd795068ea989ad5)
> >
> > But there are still problems, so I'm trying to fix it. After debugging, I found that it's a htif problem, and the idea of fixing it is referenced from the sail-riscv implementation
>
> It would be good to reference the sail implementation and the
> justification for the change there
>
> Alistair
>
> >
> > "Alistair Francis" <alistair23@gmail.com>写道：
> > > On Fri, Sep 27, 2024 at 11:26 PM MingZhu Yan <trdthg47@gmail.com> wrote:
> > > >
> > > > Applications sometimes only write the lower 32-bit payload bytes, this is used
> > > > in ACT tests. As a workaround, this refers to the solution of sail-riscv.
> > >
> > > I'm not sure what ACT is, but this feels like a guest bug, not a QEMU issue.
> > >
> > > Alistair
> > >
> > > > if the payload is written a few times with the same value, we process the whole
> > > > htif command anyway.
> > > >
> > > > Signed-off-by: MingZhu Yan <yanmingzhu@iscas.ac.cn>
> > > > ---
> > > > hw/char/riscv_htif.c | 35 +++++++++++++++++++----------------
> > > > 1 file changed, 19 insertions(+), 16 deletions(-)
> > > >
> > > > diff --git a/hw/char/riscv_htif.c b/hw/char/riscv_htif.c
> > > > index 9bef60def1..d74cce3bef 100644
> > > > --- a/hw/char/riscv_htif.c
> > > > +++ b/hw/char/riscv_htif.c
> > > > @@ -65,16 +65,8 @@ void htif_symbol_callback(const char *st_name, int st_info, uint64_t st_value,
> > > > {
> > > > if (strcmp("fromhost", st_name) == 0) {
> > > > fromhost_addr = st_value;
> > > > - if (st_size != 8) {
> > > > - error_report("HTIF fromhost must be 8 bytes");
> > > > - exit(1);
> > > > - }
> > > > } else if (strcmp("tohost", st_name) == 0) {
> > > > tohost_addr = st_value;
> > > > - if (st_size != 8) {
> > > > - error_report("HTIF tohost must be 8 bytes");
> > > > - exit(1);
> > > > - }
> > > > } else if (strcmp("begin_signature", st_name) == 0) {
> > > > begin_sig_addr = st_value;
> > > > } else if (strcmp("end_signature", st_name) == 0) {
> > > > @@ -290,18 +282,26 @@ static void htif_mm_write(void *opaque, hwaddr addr,
> > > > uint64_t value, unsigned size)
> > > > {
> > > > HTIFState *s = opaque;
> > > > - if (addr == TOHOST_OFFSET1) {
> > > > - if (s->tohost == 0x0) {
> > > > - s->allow_tohost = 1;
> > > > - s->tohost = value & 0xFFFFFFFF;
> > > > + int htif_cmd_write = 0;
> > > > + if (size == 8 && addr == TOHOST_OFFSET1) {
> > > > + htif_cmd_write = 1;
> > > > + s->tohost = value;
> > > > + htif_handle_tohost_write(s, s->tohost);
> > > > + } else if (size == 4 && addr == TOHOST_OFFSET1) {
> > > > + if ((value) == (s->tohost & 0xFFFF)) {
> > > > + s->allow_tohost = s->allow_tohost + 1;
> > > > } else {
> > > > s->allow_tohost = 0;
> > > > }
> > > > - } else if (addr == TOHOST_OFFSET2) {
> > > > - if (s->allow_tohost) {
> > > > - s->tohost |= value << 32;
> > > > - htif_handle_tohost_write(s, s->tohost);
> > > > + s->tohost = deposit64(s->tohost, 0, 32, value);
> > > > + } else if (size == 4 && addr == TOHOST_OFFSET2) {
> > > > + if ((value & 0xFF) == (s->tohost & 0xFF00)) {
> > > > + s->allow_tohost = s->allow_tohost + 1;
> > > > + } else {
> > > > + s->allow_tohost = 1;
> > > > }
> > > > + htif_cmd_write = 1;
> > > > + s->tohost = deposit64(s->tohost, 32, 32, value);
> > > > } else if (addr == FROMHOST_OFFSET1) {
> > > > s->fromhost_inprogress = 1;
> > > > s->fromhost = value & 0xFFFFFFFF;
> > > > @@ -312,6 +312,9 @@ static void htif_mm_write(void *opaque, hwaddr addr,
> > > > qemu_log("Invalid htif write: address %016" PRIx64 "\n",
> > > > (uint64_t)addr);
> > > > }
> > > > + if ((s->tohost == 1 && htif_cmd_write) || s->allow_tohost > 2) {
> > > > + htif_handle_tohost_write(s, s->tohost);
> > > > + }
> > > > }
> > > >
> > > > static const MemoryRegionOps htif_mm_ops = {
> > > > --
> > > > 2.34.1
> > > >
> > > >
> > </yanmingzhu@iscas.ac.cn></trdthg47@gmail.com>
next prev parent reply other threads:[~2024-10-16 10:33 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-09-27 8:35 [PATCH] hw/char/riscv_htif: Fix htif_mm_write that causes infinite loop in ACT MingZhu Yan
2024-10-01 6:12 ` Alistair Francis
2024-10-14 10:08 ` 阎明铸
2024-10-16 5:27 ` Alistair Francis
[not found] ` <CAPVrrNLf449o57ZsC-T5qZeWJcQvvfOBVd+U+m60dwzvf3RBxg@mail.gmail.com>
2024-10-16 10:32 ` Trd thg [this message]
2024-10-17 4:22 ` Alistair Francis
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=CAPVrrNJTasrpMxbWFacFGQkvgf1LRyZoJpn1Z-9E-_iXmGfuyA@mail.gmail.com \
--to=trdthg47@gmail.com \
--cc=alistair23@gmail.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-riscv@nongnu.org \
--cc=yanmingzhu@iscas.ac.cn \
/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;
as well as URLs for NNTP newsgroup(s).