OpenSBI Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Yao Zi <ziyao@disroot.org>
To: Anup Patel <anup@brainfault.org>
Cc: opensbi@lists.infradead.org
Subject: Re: [PATCH v3 1/2] lib: sbi: hart: Detect existence of cycle and instret CSRs for Zicntr
Date: Wed, 7 May 2025 07:17:33 +0000	[thread overview]
Message-ID: <aBsJDVTeJN5sPOXA@pie.lan> (raw)
In-Reply-To: <CAAhSdy0vWwu3RPkT4d64nowK_EAz7NpX+K4A8X8iPHSAtB=Hpg@mail.gmail.com>

On Wed, May 07, 2025 at 11:53:55AM +0530, Anup Patel wrote:
> On Sat, May 3, 2025 at 4:28 PM Yao Zi <ziyao@disroot.org> wrote:
> >
> > Zicntr extension specifies three read-only CSRs, time, cycle and
> > instret. It isn't sufficient to report Zicntr is fully supported with
> > only time CSR detected.
> >
> > This patch introduces a bitmap to sbi_hart_features to record
> > availability of these CSRs, which are detected using traps. Zicntr is
> > reported as present if and only if three CSRs are all available on the
> > HARTs.
> >
> > Suggested-by: Anup Patel <anup@brainfault.org>
> > Signed-off-by: Yao Zi <ziyao@disroot.org>
> > ---
> >  include/sbi/sbi_hart.h | 10 ++++++++++
> >  lib/sbi/sbi_hart.c     | 35 ++++++++++++++++++++++++++++-------
> >  2 files changed, 38 insertions(+), 7 deletions(-)
> 
> Why are the changes in lib/sbi/sbi_timer.c dropped ?

Oops, I just forgot about it when writing v2. sbi_timer.c should be
changed to check for only timer csr with sbi_hart_has_csr(). I'll add
it back as a separate patch in v4.

> Regards,
> Anup

Thanks,
Yao Zi

> >
> > diff --git a/include/sbi/sbi_hart.h b/include/sbi/sbi_hart.h
> > index bce88c5..dfbe8e4 100644
> > --- a/include/sbi/sbi_hart.h
> > +++ b/include/sbi/sbi_hart.h
> > @@ -93,6 +93,14 @@ struct sbi_hart_ext_data {
> >
> >  extern const struct sbi_hart_ext_data sbi_hart_ext[];
> >
> > +/** CSRs should be detected by access and trapping */
> > +enum sbi_hart_csrs {
> > +       SBI_HART_CSR_CYCLE = 0,
> > +       SBI_HART_CSR_TIME,
> > +       SBI_HART_CSR_INSTRET,
> > +       SBI_HART_CSR_MAX,
> > +};
> > +
> >  /*
> >   * Smepmp enforces access boundaries between M-mode and
> >   * S/U-mode. When it is enabled, the PMPs are programmed
> > @@ -112,6 +120,7 @@ struct sbi_hart_features {
> >         bool detected;
> >         int priv_version;
> >         unsigned long extensions[BITS_TO_LONGS(SBI_HART_EXT_MAX)];
> > +       unsigned long csrs[BITS_TO_LONGS(SBI_HART_CSR_MAX)];
> >         unsigned int pmp_count;
> >         unsigned int pmp_addr_bits;
> >         unsigned int pmp_log2gran;
> > @@ -150,6 +159,7 @@ bool sbi_hart_has_extension(struct sbi_scratch *scratch,
> >                             enum sbi_hart_extensions ext);
> >  void sbi_hart_get_extensions_str(struct sbi_scratch *scratch,
> >                                  char *extension_str, int nestr);
> > +bool sbi_hart_has_csr(struct sbi_scratch *scratch, enum sbi_hart_csrs csr);
> >
> >  void __attribute__((noreturn)) sbi_hart_hang(void);
> >
> > diff --git a/lib/sbi/sbi_hart.c b/lib/sbi/sbi_hart.c
> > index fc4925b..74ccdd8 100644
> > --- a/lib/sbi/sbi_hart.c
> > +++ b/lib/sbi/sbi_hart.c
> > @@ -747,6 +747,20 @@ void sbi_hart_get_extensions_str(struct sbi_scratch *scratch,
> >                 sbi_strncpy(extensions_str, "none", nestr);
> >  }
> >
> > +/**
> > + * Check whether a particular CSR is present on the HART
> > + *
> > + * @param scratch pointer to the HART scratch space
> > + * @param csr the CSR number to check
> > + */
> > +bool sbi_hart_has_csr(struct sbi_scratch *scratch, enum sbi_hart_csrs csr)
> > +{
> > +       struct sbi_hart_features *hfeatures =
> > +                       sbi_scratch_offset_ptr(scratch, hart_features_offset);
> > +
> > +       return __test_bit(csr, hfeatures->csrs);
> > +}
> > +
> >  static unsigned long hart_pmp_get_allowed_addr(void)
> >  {
> >         unsigned long val = 0;
> > @@ -803,7 +817,6 @@ static int hart_detect_features(struct sbi_scratch *scratch)
> >         struct sbi_hart_features *hfeatures =
> >                 sbi_scratch_offset_ptr(scratch, hart_features_offset);
> >         unsigned long val, oldval;
> > -       bool has_zicntr = false;
> >         int rc;
> >
> >         /* If hart features already detected then do nothing */
> > @@ -812,6 +825,7 @@ static int hart_detect_features(struct sbi_scratch *scratch)
> >
> >         /* Clear hart features */
> >         sbi_memset(hfeatures->extensions, 0, sizeof(hfeatures->extensions));
> > +       sbi_memset(hfeatures->csrs, 0, sizeof(hfeatures->csrs));
> >         hfeatures->pmp_count = 0;
> >         hfeatures->mhpm_mask = 0;
> >         hfeatures->priv_version = SBI_HART_PRIV_VER_UNKNOWN;
> > @@ -938,9 +952,6 @@ __pmp_skip:
> >         /* Detect if hart supports sscofpmf */
> >         __check_ext_csr(SBI_HART_PRIV_VER_1_11,
> >                         CSR_SCOUNTOVF, SBI_HART_EXT_SSCOFPMF);
> > -       /* Detect if hart supports time CSR */
> > -       __check_ext_csr(SBI_HART_PRIV_VER_UNKNOWN,
> > -                       CSR_TIME, SBI_HART_EXT_ZICNTR);
> >         /* Detect if hart has AIA local interrupt CSRs */
> >         __check_ext_csr(SBI_HART_PRIV_VER_UNKNOWN,
> >                         CSR_MTOPI, SBI_HART_EXT_SMAIA);
> > @@ -962,8 +973,16 @@ __pmp_skip:
> >
> >  #undef __check_ext_csr
> >
> > -       /* Save trap based detection of Zicntr */
> > -       has_zicntr = sbi_hart_has_extension(scratch, SBI_HART_EXT_ZICNTR);
> > +#define __check_csr_existence(__csr, __csr_id)                         \
> > +       csr_read_allowed(__csr, &trap);                                 \
> > +       if (!trap.cause)                                                \
> > +               __set_bit(__csr_id, hfeatures->csrs);
> > +
> > +       __check_csr_existence(CSR_CYCLE, SBI_HART_CSR_CYCLE);
> > +       __check_csr_existence(CSR_TIME, SBI_HART_CSR_TIME);
> > +       __check_csr_existence(CSR_INSTRET, SBI_HART_CSR_INSTRET);
> > +
> > +#undef __check_csr_existence
> >
> >         /* Let platform populate extensions */
> >         rc = sbi_platform_extensions_init(sbi_platform_thishart_ptr(),
> > @@ -973,7 +992,9 @@ __pmp_skip:
> >
> >         /* Zicntr should only be detected using traps */
> >         __sbi_hart_update_extension(hfeatures, SBI_HART_EXT_ZICNTR,
> > -                                   has_zicntr);
> > +                           sbi_hart_has_csr(scratch, SBI_HART_CSR_CYCLE) &&
> > +                           sbi_hart_has_csr(scratch, SBI_HART_CSR_TIME)  &&
> > +                           sbi_hart_has_csr(scratch, SBI_HART_CSR_INSTRET));
> >
> >         /* Extensions implied by other extensions and features */
> >         if (hfeatures->mhpm_mask)
> > --
> > 2.49.0
> >
> 
> -- 
> opensbi mailing list
> opensbi@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/opensbi

-- 
opensbi mailing list
opensbi@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/opensbi

  reply	other threads:[~2025-05-07  8:16 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-05-03 10:57 [PATCH v3 0/2] Register Zicntr in FDT when emulating is possible Yao Zi
2025-05-03 10:57 ` [PATCH v3 1/2] lib: sbi: hart: Detect existence of cycle and instret CSRs for Zicntr Yao Zi
2025-05-07  6:23   ` Anup Patel
2025-05-07  7:17     ` Yao Zi [this message]
2025-05-03 10:57 ` [PATCH v3 2/2] lib: utils: fdt: Claim Zicntr if time CSR emulation is possible Yao Zi

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=aBsJDVTeJN5sPOXA@pie.lan \
    --to=ziyao@disroot.org \
    --cc=anup@brainfault.org \
    --cc=opensbi@lists.infradead.org \
    /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