User Details
- User Since
- Mar 11 2014, 8:46 PM (527 w, 1 d)
Yesterday
Move nvmf_transport_type to libnvmf
Move nvmf_transport_type to libnvmf
Move nvmf_transport_type to libnvmf
Rename #if 0
Add reference to spec
This is indeed a large change, but it's not really feasible to split it up further. :(
Hmm, I could split the ramdisk/block backends into one to use the accessors and generic error handlers and a second to add the NVME opcode handling.
I could indeed split this up a bit further if that is helpful. Adding wrapper routines is certainly easy to split out. I slightly worry that adding the assertions is kind of odd without the context for seeing which places were updated to handle both and which were instead changed to assert SCSI-only.
I realize it's rather large, and I don't really expect super detailed review as it takes quite a bit of context to page NVMeoF in.
To be clear, this is the stock file from SPDK without any changes.
Mon, Apr 15
Sun, Apr 14
Sat, Apr 13
Fri, Apr 12
Sorry, s/lf/file/
Oh, if it is just a global symbol that isn't a ton of work to handle. You just need to recognize that symbol name when resolving symbols and resolve the pointer to the right thing instead. Probably what we would want to do is having __this_module store the pointer to the current linker_file_t when you are resolving symbols and require it to be the same size as a pointer.
LGTM. If the file was using normal style it wouldn't need the indentation fixup (which will add a lot of noise). I would maybe suggest instead to fix the indentation to not be unusual in lieu of the one char fixup.
I do think that GCC and clang (and really ld.bfd and lld) will collapse duplicate copies of the same anonymous string down to the same storage.
@markj worked on the avx512 bits for ossl for AES-GCM so probably should look as well
The commit log though can say "Remove pre-C99 support" though since the blocks are all for pre-C99.
This looks ok to me now.
I wonder if we could add new SI_SUB_DEVCTL_NOTIFY and SI_SUB_GEOM top level orders after SI_SUB_DRIVERS. I don't really like embedding lots of complexity among various other things within a single SYSINIT subsystem.
A few small suggestions and nits, but looks great to me overall.
Thu, Apr 11
Wed, Apr 10
Tue, Apr 9
I used this to test my initial libnvmf code against a Linux target before then implementing the in-kernel host. I also used this as the first client to test the userspace and later kernel controllers.
Note: I did not try to create a loopback NVMe sim to expose CTL LUNs as NVMe namespaces on the local host. Presumably that wouldn't be too hard to add. I have not tried to update any of the HA bits to support NVMe.