From mboxrd@z Thu Jan 1 00:00:00 1970 From: Greg Kurz Date: Thu, 01 Apr 2021 05:51:55 +0000 Subject: Re: [PATCH v2] ppc/spapr: Add support for implement support for H_SCM_HEALTH Message-Id: <20210401075155.44b9267d@bahia.lan> MIME-Version: 1 Content-Type: multipart/mixed; boundary="Sig_/7qo0K+59=sl2FU82aBhc9E8" List-Id: References: <20210401010519.7225-1-vaibhav@linux.ibm.com> In-Reply-To: To: David Gibson Cc: xiaoguangrong.eric@gmail.com, mst@redhat.com, aneesh.kumar@linux.ibm.com, bharata@linux.vnet.ibm.com, qemu-devel@nongnu.org, kvm-ppc@vger.kernel.org, shivaprasadbhat@gmail.com, qemu-ppc@nongnu.org, imammedo@redhat.com, Vaibhav Jain , ehabkost@redhat.com --Sig_/7qo0K+59=sl2FU82aBhc9E8 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: quoted-printable On Thu, 1 Apr 2021 13:26:11 +1100 David Gibson wrote: > On Thu, Apr 01, 2021 at 06:35:19AM +0530, Vaibhav Jain wrote: > > Add support for H_SCM_HEALTH hcall described at [1] for spapr > > nvdimms. This enables guest to detect the 'unarmed' status of a > > specific spapr nvdimm identified by its DRC and if its unarmed, mark > > the region backed by the nvdimm as read-only. > >=20 > > The patch adds h_scm_health() to handle the H_SCM_HEALTH hcall which > > returns two 64-bit bitmaps (health bitmap, health bitmap mask) derived > > from 'struct nvdimm->unarmed' member. > >=20 > > Linux kernel side changes to enable handling of 'unarmed' nvdimms for > > ppc64 are proposed at [2]. > >=20 > > References: > > [1] "Hypercall Op-codes (hcalls)" > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/= tree/Documentation/powerpc/papr_hcalls.rst#n220 > > [2] "powerpc/papr_scm: Mark nvdimm as unarmed if needed during probe" > > https://lore.kernel.org/linux-nvdimm/20210329113103.476760-1-vaibha= v@linux.ibm.com/ > >=20 > > Signed-off-by: Vaibhav Jain >=20 > As well as the handful of comments below, this will definitely need to > wait for ppc-6.1 at this point. >=20 > > --- > > Changelog > >=20 > > v2: > > * Added a check for drc->dev to ensure that the dimm is plugged in > > when servicing H_SCM_HEALTH. [ Shiva ] > > * Instead of accessing the 'nvdimm->unarmed' member directly use the > > object_property_get_bool accessor to fetch it. [ Shiva ] > > * Update the usage of PAPR_PMEM_UNARMED* macros [ Greg ] > > * Updated patch description reference#1 to point appropriate section > > in the documentation. [ Greg ] > > --- > > hw/ppc/spapr_nvdimm.c | 38 ++++++++++++++++++++++++++++++++++++++ > > include/hw/ppc/spapr.h | 3 ++- > > 2 files changed, 40 insertions(+), 1 deletion(-) > >=20 > > diff --git a/hw/ppc/spapr_nvdimm.c b/hw/ppc/spapr_nvdimm.c > > index b46c36917c..34096e4718 100644 > > --- a/hw/ppc/spapr_nvdimm.c > > +++ b/hw/ppc/spapr_nvdimm.c > > @@ -31,6 +31,13 @@ > > #include "qemu/range.h" > > #include "hw/ppc/spapr_numa.h" > > =20 > > +/* DIMM health bitmap bitmap indicators. Taken from kernel's papr_scm.= c */ > > +/* SCM device is unable to persist memory contents */ > > +#define PAPR_PMEM_UNARMED (1ULL << (63 - 0)) >=20 > You can use PPC_BIT() for more clarity here. >=20 I had already suggested PPC_BIT(0) but since this macro was copied from the kernel source, I've let Vaibhav decide whether to use PPC_BIT() or keep the macro and comment it comes from the kernel. I agree I prefer PPC_BIT(0) :-) > > +/* Bits status indicators for health bitmap indicating unarmed dimm */ > > +#define PAPR_PMEM_UNARMED_MASK (PAPR_PMEM_UNARMED) >=20 > I'm not sure why you want two equal #defines here. >=20 Especially, this define doesn't make sense for the hypervisor IMHO. It is _just_ the mask of bits for the "unarmed" state in the kernel. > > + > > bool spapr_nvdimm_validate(HotplugHandler *hotplug_dev, NVDIMMDevice *= nvdimm, > > uint64_t size, Error **errp) > > { > > @@ -467,6 +474,36 @@ static target_ulong h_scm_unbind_all(PowerPCCPU *c= pu, SpaprMachineState *spapr, > > return H_SUCCESS; > > } > > =20 > > +static target_ulong h_scm_health(PowerPCCPU *cpu, SpaprMachineState *s= papr, > > + target_ulong opcode, target_ulong *ar= gs) > > +{ > > + uint32_t drc_index =3D args[0]; > > + SpaprDrc *drc =3D spapr_drc_by_index(drc_index); > > + NVDIMMDevice *nvdimm; > > + > > + if (drc && spapr_drc_type(drc) !=3D SPAPR_DR_CONNECTOR_TYPE_PMEM) { >=20 > This will fail badly if !drc (given index is way out of bounds). I'm > pretty sure you want > if (!drc || spapr_drc_type(drc) !=3D SPAPR_DR_CONNECTOR_TYPE_PMEM) { >=20 >=20 > > + return H_PARAMETER; > > + } > > + > > + /* Ensure that the dimm is plugged in */ > > + if (!drc->dev) { > > + return H_HARDWARE; >=20 > H_HARDWARE doesn't seem right - it's the guest that has chosen to > attempt this on an unplugged LMB, not the (virtual) hardware's fault. >=20 Yes. As already suggested, simply do the same as in other hcall implementations in this file, e.g. from h_scm_bind_mem() : if (!drc || !drc->dev || spapr_drc_type(drc) !=3D SPAPR_DR_CONNECTOR_TYPE_PMEM) { return H_PARAMETER; } > > + } > > + > > + nvdimm =3D NVDIMM(drc->dev); > > + > > + args[0] =3D 0; > > + /* Check if the nvdimm is unarmed and send its status via health b= itmaps */ Not sure this comment is super useful. > > + if (object_property_get_bool(OBJECT(nvdimm), NVDIMM_UNARMED_PROP, = NULL)) { > > + args[0] |=3D PAPR_PMEM_UNARMED; > > + } > > + > > + /* Update the health bitmap with the applicable mask */ > > + args[1] =3D PAPR_PMEM_UNARMED_MASK; I still think this is a misuse of PAPR_PMEM_UNARMED_MASK... The meaning of args[1] is "health-bit-valid-bitmap indicate which bits in health-bitmap are valid" according to the documentation. https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Doc= umentation/powerpc/papr_hcalls.rst#n228 Without any further detail, I tend to consider this as a hint to the guest on the bits supported by the hypervisor. Since we can only set PAPR_PMEM_UNARMED, for now, args[1] should be set to just that bit PAPR_PMEM_UNARMED. Other bits can be added later if QEMU supports more of them. > > + > > + return H_SUCCESS; > > +} > > + > > static void spapr_scm_register_types(void) > > { > > /* qemu/scm specific hcalls */ > > @@ -475,6 +512,7 @@ static void spapr_scm_register_types(void) > > spapr_register_hypercall(H_SCM_BIND_MEM, h_scm_bind_mem); > > spapr_register_hypercall(H_SCM_UNBIND_MEM, h_scm_unbind_mem); > > spapr_register_hypercall(H_SCM_UNBIND_ALL, h_scm_unbind_all); > > + spapr_register_hypercall(H_SCM_HEALTH, h_scm_health); > > } > > =20 > > type_init(spapr_scm_register_types) > > diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h > > index 47cebaf3ac..6e1eafb05d 100644 > > --- a/include/hw/ppc/spapr.h > > +++ b/include/hw/ppc/spapr.h > > @@ -538,8 +538,9 @@ struct SpaprMachineState { > > #define H_SCM_BIND_MEM 0x3EC > > #define H_SCM_UNBIND_MEM 0x3F0 > > #define H_SCM_UNBIND_ALL 0x3FC > > +#define H_SCM_HEALTH 0x400 > > =20 > > -#define MAX_HCALL_OPCODE H_SCM_UNBIND_ALL > > +#define MAX_HCALL_OPCODE H_SCM_HEALTH > > =20 > > /* The hcalls above are standardized in PAPR and implemented by pHyp > > * as well. >=20 --Sig_/7qo0K+59=sl2FU82aBhc9E8 Content-Type: application/pgp-signature Content-Description: OpenPGP digital signature -----BEGIN PGP SIGNATURE----- iQIzBAEBCAAdFiEEtIKLr5QxQM7yo0kQcdTV5YIvc9YFAmBlX3sACgkQcdTV5YIv c9YBlA//S0kZnzRddcnbSPO/BIdgPu/FpvEbVCzApps8NIrR/lNT59nRz2+B62Lq vKwq+ueubX/Hj3ipYRMeIQouk7FsJNi+ZbDvvlO4htEeNrswm7BOMUIMZRJt2+Cz NDNfkQwfCPIdKPWiUUGIv6ZOSESMcaHBs6VrVeqiqPG+YOmXcrhwDtjEk+fw1NCf iX6inkhCeBcK+C7L+YcwXIypyAL28PWUwAyEHjk6M+TQGa+WXo4QiUSYyJnXr+no um8RqSdSejrcmM4DEtEi8k+Rg9Yr3nf2fc4vGXiT1eqbkHSATg9gy6lO7fZOp0p8 nQPWY95Q7ghm7CSUpwTJv2bjL0bDVGIJ0/kD4m4+MhrBSnAapM/Hr8FtY18DvKng 5qSApuVDlY0XwUEPXtKBahIHlwT1o0O/EozsA8ybVx7TsXJBZ1gLbfqNOQn1UTDi 01QviqIKCAROXt4tBRqr2l359XoD1So/e5BoEPP71Dfi1LipOoMM9chuRB58cD+D qI+sDH9kcxiQMUEhjUSqXo53ADHFFexK8LIYL0h51fDBT4P0/4mYT9zwlWEIXRCo 0AHdhgYWpby09pS8N8A8BQStTnQC306i5C34bPnYFbs8t25QFL4BY7B44qzb3aF+ 52yfDjcBtBy85NuNPlQitbwIkG5KyQUUZ5HBu2y7UCrsx7S/L5w= =Qqjn -----END PGP SIGNATURE----- --Sig_/7qo0K+59=sl2FU82aBhc9E8--