From: Gautham R Shenoy <ego@linux.vnet.ibm.com>
To: Abhishek Goel <huntbag@linux.vnet.ibm.com>
Cc: ego@linux.vnet.ibm.com, mikey@neuling.org,
linux-kernel@vger.kernel.org, npiggin@gmail.com,
linuxppc-dev@ozlabs.org, oohall@gmail.com, psampat@linux.ibm.com
Subject: Re: [RFC 1/3] powernv/cpuidle : Support for pre-entry and post exit of stop state in firmware
Date: Mon, 27 Apr 2020 15:56:26 +0530 [thread overview]
Message-ID: <20200427102626.GC29708@in.ibm.com> (raw)
In-Reply-To: <20200427021027.114582-1-huntbag@linux.vnet.ibm.com>
Hi Abhishek,
On Sun, Apr 26, 2020 at 09:10:25PM -0500, Abhishek Goel wrote:
> This patch provides kernel framework fro opal support of save restore
> of sprs in idle stop loop. Opal support for stop states is needed to
> selectively enable stop states or to introduce a quirk quickly in case
> a buggy stop state is present.
>
> We make a opal call from kernel if firmware-stop-support for stop
> states is present and enabled. All the quirks for pre-entry of stop
> state is handled inside opal. A call from opal is made into kernel
> where we execute stop afer saving of NVGPRs.
> After waking up from 0x100 vector in kernel, we enter back into opal.
> All the quirks in post exit path, if any, are then handled in opal,
> from where we return successfully back to kernel.
> For deep stop states in which additional SPRs are lost, saving and
> restoration will be done in OPAL.
>
> This idea was first proposed by Nick here:
> https://patchwork.ozlabs.org/patch/1208159/
>
> The corresponding skiboot patch for this kernel patch is here:
> https://patchwork.ozlabs.org/project/skiboot/list/?series=172831
>
> When we callback from OPAL into kernel, r13 is clobbered. So, to
> access PACA we need to restore it from HSPRGO. In future we can
> handle this into OPAL as in here:
> https://patchwork.ozlabs.org/patch/1245275/
>
> Signed-off-by: Abhishek Goel <huntbag@linux.vnet.ibm.com>
> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> ---
>
> v1->v2 : No change in this patch.
>
> arch/powerpc/include/asm/opal-api.h | 8 ++++-
> arch/powerpc/include/asm/opal.h | 3 ++
> arch/powerpc/kernel/idle_book3s.S | 5 +++
> arch/powerpc/platforms/powernv/idle.c | 37 ++++++++++++++++++++++
> arch/powerpc/platforms/powernv/opal-call.c | 2 ++
> 5 files changed, 54 insertions(+), 1 deletion(-)
>
> diff --git a/arch/powerpc/include/asm/opal-api.h b/arch/powerpc/include/asm/opal-api.h
> index c1f25a760eb1..a2c782c99c9e 100644
> --- a/arch/powerpc/include/asm/opal-api.h
> +++ b/arch/powerpc/include/asm/opal-api.h
> @@ -214,7 +214,9 @@
> #define OPAL_SECVAR_GET 176
> #define OPAL_SECVAR_GET_NEXT 177
> #define OPAL_SECVAR_ENQUEUE_UPDATE 178
> -#define OPAL_LAST 178
> +#define OPAL_REGISTER_OS_OPS 181
> +#define OPAL_CPU_IDLE 182
> +#define OPAL_LAST 182
>
> #define QUIESCE_HOLD 1 /* Spin all calls at entry */
> #define QUIESCE_REJECT 2 /* Fail all calls with OPAL_BUSY */
> @@ -1181,6 +1183,10 @@ struct opal_mpipl_fadump {
> struct opal_mpipl_region region[];
> } __packed;
>
> +struct opal_os_ops {
> + __be64 os_idle_stop;
> +};
> +
> #endif /* __ASSEMBLY__ */
>
> #endif /* __OPAL_API_H */
> diff --git a/arch/powerpc/include/asm/opal.h b/arch/powerpc/include/asm/opal.h
> index 9986ac34b8e2..3c340bc4df8e 100644
> --- a/arch/powerpc/include/asm/opal.h
> +++ b/arch/powerpc/include/asm/opal.h
> @@ -400,6 +400,9 @@ void opal_powercap_init(void);
> void opal_psr_init(void);
> void opal_sensor_groups_init(void);
>
> +extern int64_t opal_register_os_ops(struct opal_os_ops *os_ops);
> +extern int64_t opal_cpu_idle(__be64 srr1_addr, uint64_t psscr);
> +
> #endif /* __ASSEMBLY__ */
>
> #endif /* _ASM_POWERPC_OPAL_H */
> diff --git a/arch/powerpc/kernel/idle_book3s.S b/arch/powerpc/kernel/idle_book3s.S
> index 22f249b6f58d..8d287d1d06c0 100644
> --- a/arch/powerpc/kernel/idle_book3s.S
> +++ b/arch/powerpc/kernel/idle_book3s.S
> @@ -49,6 +49,8 @@ _GLOBAL(isa300_idle_stop_noloss)
> */
> _GLOBAL(isa300_idle_stop_mayloss)
> mtspr SPRN_PSSCR,r3
> + mr r6, r13
> + mfspr r13, SPRN_HSPRG0
> std r1,PACAR1(r13)
> mflr r4
> mfcr r5
> @@ -74,6 +76,7 @@ _GLOBAL(isa300_idle_stop_mayloss)
> std r31,-8*18(r1)
> std r4,-8*19(r1)
> std r5,-8*20(r1)
> + std r6,-8*21(r1)
> /* 168 bytes */
> PPC_STOP
> b . /* catch bugs */
> @@ -91,8 +94,10 @@ _GLOBAL(idle_return_gpr_loss)
> ld r1,PACAR1(r13)
> ld r4,-8*19(r1)
> ld r5,-8*20(r1)
> + ld r6,-8*21(r1)
> mtlr r4
> mtcr r5
> + mr r13,r6
> /*
> * KVM nap requires r2 to be saved, rather than just restoring it
> * from PACATOC. This could be avoided for that less common case
> diff --git a/arch/powerpc/platforms/powernv/idle.c b/arch/powerpc/platforms/powernv/idle.c
> index 78599bca66c2..1841027b25c5 100644
> --- a/arch/powerpc/platforms/powernv/idle.c
> +++ b/arch/powerpc/platforms/powernv/idle.c
> @@ -35,6 +35,7 @@
> static u32 supported_cpuidle_states;
> struct pnv_idle_states_t *pnv_idle_states;
> int nr_pnv_idle_states;
> +static bool firmware_stop_supported;
>
> /*
> * The default stop state that will be used by ppc_md.power_save
> @@ -602,6 +603,25 @@ struct p9_sprs {
> u64 uamor;
> };
>
> +/*
> + * This function is called from OPAL if firmware support for stop
> + * states is present and enabled. It provides a fallback for idle
> + * stop states via OPAL.
> + */
> +static uint64_t os_idle_stop(uint64_t psscr, bool save_gprs)
> +{
> + /*
> + * For lite state which does not lose even GPRS we call
> + * idle_stop_noloss while for all other states we call
> + * idle_stop_mayloss. Saving and restoration of other additional
> + * SPRs if required is handled in OPAL. All the quirks are also
> + * handled in OPAL.
> + */
> + if (!save_gprs)
> + return isa300_idle_stop_noloss(psscr);
I think PSSCR[ESL|EC] = 0 case is an overkill to go into OPAL and come
back via a callback. That can be handled in the kernel itself.
> + return isa300_idle_stop_mayloss(psscr);
> +}
> +
> static unsigned long power9_idle_stop(unsigned long psscr, bool mmu_on)
> {
> int cpu = raw_smp_processor_id();
> @@ -613,6 +633,16 @@ static unsigned long power9_idle_stop(unsigned long psscr, bool mmu_on)
> unsigned long mmcr0 = 0;
> struct p9_sprs sprs = {}; /* avoid false used-uninitialised */
> bool sprs_saved = false;
> + int rc = 0;
> +
> + /*
> + * Kernel takes decision whether to make OPAL call or not. This logic
> + * will be combined with the logic for BE opal to take decision.
> + */
> + if (firmware_stop_supported) {
> + rc = opal_cpu_idle(cpu_to_be64(__pa(&srr1)), (uint64_t) psscr);
Couple of comments here.
1) If PSSCR[ESL|EC] = 0, the current code expects mmu_on=true. When
we make an OPAL call and come back into the kernel via the
callback today, we will be in real-mode, with mmu turned off.
2) You seem to be choosing the opal cpuidle support as the default
case, and not as a fallback. Thus, with this patch you will miss
out on the deep stop-state support.
> + goto out;
> + }
>
> if (!(psscr & (PSSCR_EC|PSSCR_ESL))) {
> /* EC=ESL=0 case */
> @@ -1232,6 +1262,10 @@ static int pnv_parse_cpuidle_dt(void)
> pr_warn("opal: PowerMgmt Node not found\n");
> return -ENODEV;
> }
> +
> + if (of_device_is_compatible(np, "firmware-stop-supported"))
> + firmware_stop_supported = true;
IMO, at least for POWER9 generation, you need to do this only when
"idle-stop" device-tree cpu-feature is unavailable.
> +
> nr_idle_states = of_property_count_u32_elems(np,
> "ibm,cpu-idle-state-flags");
>
> @@ -1326,6 +1360,7 @@ static int pnv_parse_cpuidle_dt(void)
>
> static int __init pnv_init_idle_states(void)
> {
> + struct opal_os_ops os_ops;
> int cpu;
> int rc = 0;
>
> @@ -1349,6 +1384,8 @@ static int __init pnv_init_idle_states(void)
> }
> }
>
> + os_ops.os_idle_stop = be64_to_cpu(os_idle_stop);
> + rc = opal_register_os_ops((struct opal_os_ops *)(&os_ops));
> /* In case we error out nr_pnv_idle_states will be zero */
> nr_pnv_idle_states = 0;
> supported_cpuidle_states = 0;
> diff --git a/arch/powerpc/platforms/powernv/opal-call.c b/arch/powerpc/platforms/powernv/opal-call.c
> index 5cd0f52d258f..c885e607ba62 100644
> --- a/arch/powerpc/platforms/powernv/opal-call.c
> +++ b/arch/powerpc/platforms/powernv/opal-call.c
> @@ -293,3 +293,5 @@ OPAL_CALL(opal_mpipl_query_tag, OPAL_MPIPL_QUERY_TAG);
> OPAL_CALL(opal_secvar_get, OPAL_SECVAR_GET);
> OPAL_CALL(opal_secvar_get_next, OPAL_SECVAR_GET_NEXT);
> OPAL_CALL(opal_secvar_enqueue_update, OPAL_SECVAR_ENQUEUE_UPDATE);
> +OPAL_CALL(opal_register_os_ops, OPAL_REGISTER_OS_OPS);
> +OPAL_CALL(opal_cpu_idle, OPAL_CPU_IDLE);
> --
> 2.17.1
>
next prev parent reply other threads:[~2020-04-27 10:28 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-04-27 2:10 [RFC 1/3] powernv/cpuidle : Support for pre-entry and post exit of stop state in firmware Abhishek Goel
2020-04-27 2:10 ` [RFC 2/3] powernv/cpuidle : Interface for an idle-stop dependency structure Abhishek Goel
2020-04-27 13:47 ` Gautham R Shenoy
2020-04-27 2:10 ` [RFC 3/3] powernv/cpuidle : Introduce capability for firmware-enabled-stop Abhishek Goel
2020-04-27 13:48 ` Gautham R Shenoy
2020-04-27 10:26 ` Gautham R Shenoy [this message]
2020-04-28 1:08 ` [RFC 1/3] powernv/cpuidle : Support for pre-entry and post exit of stop state in firmware Nicholas Piggin
2020-04-30 5:52 ` Abhishek
2020-05-02 11:40 ` Nicholas Piggin
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=20200427102626.GC29708@in.ibm.com \
--to=ego@linux.vnet.ibm.com \
--cc=huntbag@linux.vnet.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linuxppc-dev@ozlabs.org \
--cc=mikey@neuling.org \
--cc=npiggin@gmail.com \
--cc=oohall@gmail.com \
--cc=psampat@linux.ibm.com \
/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).