On Fri, 17 Mar 2017 09:25:56 -0400 Dennis Dalessandro <dennis.dalessandro@xxxxxxxxx> wrote: > On 03/13/2017 07:51 AM, Thomas Bogendoerfer wrote: > > Hi, > > > > I'm tyring to connect a HFI card to an Omni-Path Edge Switch 100. The > > switch detects the QSFP just fine: > > > > Cable05 OEM Copper 3 meter 20.87 ??? MCP1600-E003-C 6 M9AL0003 Valid > > > > But on the host side, I only get: > > > > [ 53.060613] hfi1 0000:01:00.0: hfi1_0: wait_for_qsfp_init: No IntN detected, reset complete > > [ 53.082820] hfi1 0000:01:00.0: hfi1_0: i2c_bus_write: bus 1, i2c slave 0x50, offset 0x7f, len 0x1; write failed, ret -6 > > [ 53.114661] hfi1 0000:01:00.0: hfi1_0: port 1: QSFP chain 1 can't write QSFP_PAGE_SELECT_BYTE: -6 > > [ 53.137353] hfi1 0000:01:00.0: hfi1_0: QSFP not responding, waiting and retrying 0 > > > > Any hint why the card doesn't find the plugged in QSFPs ? > > Please refer to the Intel Omni Path cable support matrix for a list of > supported cables: > http://www.intel.com/content/www/us/en/high-performance-computing-fabrics/omni-path-cables.html. > > If you need help with cables not on this list, please contact the cable > manufacturer or supplier for assistance. so why does the _INTEL_ switch accept the cable while the _INTEL_ HFI card doesn't ? And the I2C interface to the QSFPs looks like bitbang on GPIO lines so even if there the quality of the cable isn't good enough, I would have expected to see at least the QSFP register via I2C (I tried via i2c-cdev/i2c-detect, but there is simply nothing). Anyway thank you for your help. Thomas. -- SUSE Linux GmbH GF: Felix Imendörffer, Jane Smithard, Graham Norton HRB 21284 (AG Nürnberg) -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
