| Message ID | 20190619124510.25182-7-fbarrat@linux.ibm.com |
|---|---|
| State | RFC |
| Headers | show |
| Series | opencapi: enable card reset and link retraining | expand |
| Context | Check | Description |
|---|---|---|
| snowpatch_ozlabs/apply_patch | success | Successfully applied on branch master (b904cb733750de1bb0e04e5012c391a9c3094d11) |
| snowpatch_ozlabs/snowpatch_job_snowpatch-skiboot | fail | Test snowpatch/job/snowpatch-skiboot on branch master |
| snowpatch_ozlabs/snowpatch_job_snowpatch-skiboot-dco | success | Signed-off-by present |
On 19/6/19 10:45 pm, Frederic Barrat wrote: > Let's get rid of one transitional state, since there's no need to > pause in between releasing the reset signals of the ODL and the > adapter. > > Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com> > - /* > - * Minimal delay before taking adapter out of > - * reset. Could be useless, but doesn't hurt > - */ > - return pci_slot_set_sm_timeout(slot, msecs_to_tb(1)); > - So it turns out this is indeed useless? Reviewed-by: Andrew Donnellan <ajd@linux.ibm.com> > - case OCAPI_SLOT_FRESET_DEASSERT_DELAY: > deassert_adapter_reset(dev); > pci_slot_set_state(slot, > - OCAPI_SLOT_FRESET_DEASSERT_DELAY2); > + OCAPI_SLOT_FRESET_DEASSERT_DELAY); > /* give 250ms to device to be ready */ > return pci_slot_set_sm_timeout(slot, msecs_to_tb(250)); > > - case OCAPI_SLOT_FRESET_DEASSERT_DELAY2: > + case OCAPI_SLOT_FRESET_DEASSERT_DELAY: > unfence_brick(dev); > set_init_pattern(chip_id, dev); > pci_slot_set_state(slot, >
Le 21/06/2019 à 02:27, Andrew Donnellan a écrit : > On 19/6/19 10:45 pm, Frederic Barrat wrote: >> Let's get rid of one transitional state, since there's no need to >> pause in between releasing the reset signals of the ODL and the >> adapter. >> >> Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com> >> - /* >> - * Minimal delay before taking adapter out of >> - * reset. Could be useless, but doesn't hurt >> - */ >> - return pci_slot_set_sm_timeout(slot, msecs_to_tb(1)); >> - > > So it turns out this is indeed useless? Yes. The DLx designer had always stated so. I added the delay as a conservative measure a while back, as it took some time for the link retrain to become stable (needed a patch in the DLx/TLx + the new PHY settings + me understanding what was going on). But I've removed it in my setup for quite a while now, and my boot loop and reset loops have been working well, so I'm confident to remove it. Fred > Reviewed-by: Andrew Donnellan <ajd@linux.ibm.com> > >> - case OCAPI_SLOT_FRESET_DEASSERT_DELAY: >> deassert_adapter_reset(dev); >> pci_slot_set_state(slot, >> - OCAPI_SLOT_FRESET_DEASSERT_DELAY2); >> + OCAPI_SLOT_FRESET_DEASSERT_DELAY); >> /* give 250ms to device to be ready */ >> return pci_slot_set_sm_timeout(slot, msecs_to_tb(250)); >> - case OCAPI_SLOT_FRESET_DEASSERT_DELAY2: >> + case OCAPI_SLOT_FRESET_DEASSERT_DELAY: >> unfence_brick(dev); >> set_init_pattern(chip_id, dev); >> pci_slot_set_state(slot, >> >
diff --git a/hw/npu2-opencapi.c b/hw/npu2-opencapi.c index f9cb1d26..1328ee93 100644 --- a/hw/npu2-opencapi.c +++ b/hw/npu2-opencapi.c @@ -66,8 +66,7 @@ #define OCAPI_SLOT_FRESET_INIT (OCAPI_SLOT_FRESET + 2) #define OCAPI_SLOT_FRESET_ASSERT_DELAY (OCAPI_SLOT_FRESET + 3) #define OCAPI_SLOT_FRESET_DEASSERT_DELAY (OCAPI_SLOT_FRESET + 4) -#define OCAPI_SLOT_FRESET_DEASSERT_DELAY2 (OCAPI_SLOT_FRESET + 5) -#define OCAPI_SLOT_FRESET_INIT_DELAY (OCAPI_SLOT_FRESET + 6) +#define OCAPI_SLOT_FRESET_INIT_DELAY (OCAPI_SLOT_FRESET + 5) #define OCAPI_LINK_TRAINING_RETRIES 2 #define OCAPI_LINK_TRAINING_TIMEOUT 3000 /* ms */ @@ -1232,22 +1231,13 @@ static int64_t npu2_opencapi_freset(struct pci_slot *slot) case OCAPI_SLOT_FRESET_ASSERT_DELAY: npu2_opencapi_phy_reset(dev); deassert_odl_reset(chip_id, dev->brick_index); - pci_slot_set_state(slot, - OCAPI_SLOT_FRESET_DEASSERT_DELAY); - /* - * Minimal delay before taking adapter out of - * reset. Could be useless, but doesn't hurt - */ - return pci_slot_set_sm_timeout(slot, msecs_to_tb(1)); - - case OCAPI_SLOT_FRESET_DEASSERT_DELAY: deassert_adapter_reset(dev); pci_slot_set_state(slot, - OCAPI_SLOT_FRESET_DEASSERT_DELAY2); + OCAPI_SLOT_FRESET_DEASSERT_DELAY); /* give 250ms to device to be ready */ return pci_slot_set_sm_timeout(slot, msecs_to_tb(250)); - case OCAPI_SLOT_FRESET_DEASSERT_DELAY2: + case OCAPI_SLOT_FRESET_DEASSERT_DELAY: unfence_brick(dev); set_init_pattern(chip_id, dev); pci_slot_set_state(slot,
Let's get rid of one transitional state, since there's no need to pause in between releasing the reset signals of the ODL and the adapter. Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com> --- hw/npu2-opencapi.c | 16 +++------------- 1 file changed, 3 insertions(+), 13 deletions(-)