[00:54:29] *** Quits: ziyeyang__ (~ziyeyang@134.134.139.73) (Remote host closed the connection) [00:54:51] *** Joins: ziyeyang__ (~ziyeyang@192.55.54.39) [02:18:20] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Quit: My Mac Pro has gone to sleep. ZZZzzz…) [02:21:27] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [02:26:59] *** Quits: ziyeyang__ (~ziyeyang@192.55.54.39) (Remote host closed the connection) [02:27:17] *** Joins: ziyeyang__ (ziyeyang@nat/intel/x-vvkqskyxqnnnikbb) [04:12:29] *** Quits: ziyeyang__ (ziyeyang@nat/intel/x-vvkqskyxqnnnikbb) (Remote host closed the connection) [04:14:21] *** Joins: ziyeyang_ (~ziyeyang@192.55.54.39) [04:38:10] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Quit: My Mac Pro has gone to sleep. ZZZzzz…) [04:58:17] *** Joins: ziyeyang__ (ziyeyang@nat/intel/x-mmrvsazsklfpvpwt) [04:58:18] *** Quits: ziyeyang_ (~ziyeyang@192.55.54.39) (Remote host closed the connection) [05:00:20] *** Joins: ziyeyang_ (~ziyeyang@192.55.54.39) [05:00:21] *** Quits: ziyeyang__ (ziyeyang@nat/intel/x-mmrvsazsklfpvpwt) (Remote host closed the connection) [05:35:08] *** Quits: ziyeyang_ (~ziyeyang@192.55.54.39) (Remote host closed the connection) [05:36:01] *** Joins: ziyeyang_ (ziyeyang@nat/intel/x-nlnwkeciligkktwp) [05:43:40] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [07:08:25] *** Quits: ziyeyang_ (ziyeyang@nat/intel/x-nlnwkeciligkktwp) (Quit: Leaving) [07:38:26] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Quit: My Mac Pro has gone to sleep. ZZZzzz…) [07:38:59] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [08:35:07] *** Joins: Tomasz (~tzawadzk@134.134.139.77) [08:38:58] *** Quits: tomzawadzki (~tzawadzk@192.55.54.42) (Ping timeout: 240 seconds) [09:00:03] *** Quits: Tomasz (~tzawadzk@134.134.139.77) (Ping timeout: 260 seconds) [09:33:02] *** Joins: dwaddington (Daniel@nat/ibm/x-hdxiwhiisrdxivhi) [09:33:58] About the spdk_nvme_ns_cmd_read_with_md() call. If the read is across multiple blocks, does it "collect" together the metadata from the blocks and append them? [09:34:11] I've not used the API yet. [09:49:18] *** Quits: sethhowe_ (~sethhowe@192.55.54.42) (Remote host closed the connection) [09:55:54] *** Joins: sethhowe (~sethhowe@134.134.139.73) [09:59:10] dwaddington: there are two options, depending on what the device supports - at the end of each block in the data buffer, or all of the metadata in a separate buffer [09:59:25] the NVMe specification, section 8.2 (Metadata Handling) describes them [11:13:44] great, thanks for the pointer [11:22:27] dwaddington: you can use spdk_nvme_ns_supports_extended_lba() to determine which way the namespace is formatted ("extended LBA" means metadata is at the end of each block) [11:28:21] background: simply calling spdk_bs_init() with what I believe are valid parms for a malloc dev that seems to init fine. The issue I’m dealing with is a little further downstream but this question is still valid up to here anyway [11:28:22] question: in bdev_blob_unmap() there’s a conditional check on rc return from spdk_bdev_unmap() thatcalls the provided cb however the condition is “if (rc)” but it looks to me like rc will always be <=0 so the success case (0) never calls the cb. Am I missing something? [11:29:51] (not an urgent question, working around it for now) [11:33:32] negative numbers are true [11:33:48] plus, calling the callback immediately there is the error case [11:34:01] the success path is calling the callback passed to spdk_bdev_unmap [11:34:10] success is rc == 0 [11:35:29] the callback when the unmap finishes is bdev_blob_io_complete [11:35:32] which then calls the cb [11:38:20] ahh, OK walking through the UT code somethign has be confused - in test blob_init() I could have sworn that I was walking through a success case and saw that CB getting executed there. I'll take another look, thanks [11:38:55] the tests may be failing unmaps because we didn't implement them - I don't remember [11:39:21] yeah, it might not be that I was loking at a positive test too [14:25:22] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Quit: My Mac Pro has gone to sleep. ZZZzzz…) [14:28:27] *** Joins: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) [14:29:37] *** Quits: gila (~gila@5ED4FE92.cm-7-5d.dynamic.ziggo.nl) (Client Quit) [15:50:49] bwalker: yeah, UT was a bad thing to look at to figure out the big picture, switched to an NVMe backend and figure out where my misunderstandings were... so far anyways! [16:04:05] going back to spdk_nvme_ns_supports_extended_lba() and nvme metadata support; is there a way to select which if the modes 8.2 (i.e. extended LBA or not) to use? [16:05:34] oh, I just read ..The [16:05:34] mec [16:05:34] hanism used is selected as part of the Format NVM command. [16:07:40] its the .ms field of the format structure. my Intel SSD must not support both forms then? [16:42:59] *** Parts: dwaddington (Daniel@nat/ibm/x-hdxiwhiisrdxivhi) ("Leaving")