Message ID | 1706431786e7ce832b6fc22f90630a4a39a001d5.1529466573.git-series.andrew.donnellan@au1.ibm.com |
---|---|
State | Accepted |
Headers | show |
Series | occ: GPU presence detection | expand |
On 06/20/2018 09:19 AM, Andrew Donnellan wrote: > We need to use the OCC to obtain presence data for the SXM2 slots on > Witherspoon systems. This is needed to determine device type for NVLink > GPUs and OpenCAPI devices which can be plugged into the same slot. Support > for this will be implemented in a future patch. > > Currently, OCC initialisation is done just before handing over to Linux, > which is well after NPU probe. On FSP systems, OCC boot starts very late, > so we wait until the last possible moment to initialise the skiboot side in > order to give it the maximum time to boot. On BMC systems, OCC boot starts > earlier, so there aren't any issues in moving it earlier in the skiboot > init sequence. > > When running on a BMC machine, call occ_pstates_init() as early as > possible in the init sequence. On FSP machines, continue to call it from > its current location. > > Signed-off-by: Andrew Donnellan <andrew.donnellan@au1.ibm.com> > > --- > > v1->v2: > > - Change check to rely on BMC vs FSP rather than POWER8 vs POWER9 (thanks > Shilpa) Thanks for making the change. Reviewed-by: Shilpasri G Bhat <shilpa.bhat@linux.vnet.ibm.com> > --- > core/init.c | 14 +++++++++++++- > 1 file changed, 13 insertions(+), 1 deletion(-) > > diff --git a/core/init.c b/core/init.c > index 1ad747e77e4a..272a57a145ee 100644 > --- a/core/init.c > +++ b/core/init.c > @@ -505,7 +505,8 @@ void __noreturn load_and_boot_kernel(bool is_reboot) > > ipmi_set_fw_progress_sensor(IPMI_FW_OS_BOOT); > > - occ_pstates_init(); > + if (fsp_present()) > + occ_pstates_init(); > > if (!is_reboot) { > /* We wait for the nvram read to complete here so we can > @@ -1079,6 +1080,17 @@ void __noreturn __nomcount main_cpu_entry(const void *fdt) > > op_display(OP_LOG, OP_MOD_INIT, 0x0002); > > + /* > + * On some POWER9 BMC systems, we need to initialise the OCC > + * before the NPU to facilitate NVLink/OpenCAPI presence > + * detection, so we set it up as early as possible. On FSP > + * systems, Hostboot starts booting the OCC later, so we delay > + * OCC initialisation as late as possible to give it the > + * maximum time to boot up. > + */ > + if (!fsp_present()) > + occ_pstates_init(); > + > pci_nvram_init(); > > preload_io_vpd(); >
diff --git a/core/init.c b/core/init.c index 1ad747e77e4a..272a57a145ee 100644 --- a/core/init.c +++ b/core/init.c @@ -505,7 +505,8 @@ void __noreturn load_and_boot_kernel(bool is_reboot) ipmi_set_fw_progress_sensor(IPMI_FW_OS_BOOT); - occ_pstates_init(); + if (fsp_present()) + occ_pstates_init(); if (!is_reboot) { /* We wait for the nvram read to complete here so we can @@ -1079,6 +1080,17 @@ void __noreturn __nomcount main_cpu_entry(const void *fdt) op_display(OP_LOG, OP_MOD_INIT, 0x0002); + /* + * On some POWER9 BMC systems, we need to initialise the OCC + * before the NPU to facilitate NVLink/OpenCAPI presence + * detection, so we set it up as early as possible. On FSP + * systems, Hostboot starts booting the OCC later, so we delay + * OCC initialisation as late as possible to give it the + * maximum time to boot up. + */ + if (!fsp_present()) + occ_pstates_init(); + pci_nvram_init(); preload_io_vpd();
We need to use the OCC to obtain presence data for the SXM2 slots on Witherspoon systems. This is needed to determine device type for NVLink GPUs and OpenCAPI devices which can be plugged into the same slot. Support for this will be implemented in a future patch. Currently, OCC initialisation is done just before handing over to Linux, which is well after NPU probe. On FSP systems, OCC boot starts very late, so we wait until the last possible moment to initialise the skiboot side in order to give it the maximum time to boot. On BMC systems, OCC boot starts earlier, so there aren't any issues in moving it earlier in the skiboot init sequence. When running on a BMC machine, call occ_pstates_init() as early as possible in the init sequence. On FSP machines, continue to call it from its current location. Signed-off-by: Andrew Donnellan <andrew.donnellan@au1.ibm.com> --- v1->v2: - Change check to rely on BMC vs FSP rather than POWER8 vs POWER9 (thanks Shilpa) --- core/init.c | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-)