> -----Original Message----- > From: Niklas Cassel <niklas.cas...@wdc.com> > Sent: Wednesday, October 14, 2020 7:59 AM > To: Dmitry Fomichev <dmitry.fomic...@wdc.com> > Cc: Keith Busch <kbu...@kernel.org>; Klaus Jensen > <k.jen...@samsung.com>; Kevin Wolf <kw...@redhat.com>; Philippe > Mathieu-Daudé <phi...@redhat.com>; Maxim Levitsky > <mlevi...@redhat.com>; Fam Zheng <f...@euphon.net>; Alistair Francis > <alistair.fran...@wdc.com>; Matias Bjorling <matias.bjorl...@wdc.com>; > Damien Le Moal <damien.lem...@wdc.com>; qemu-bl...@nongnu.org; > qemu-devel@nongnu.org > Subject: Re: [PATCH v6 05/11] hw/block/nvme: Support Zoned Namespace > Command Set > > On Wed, Oct 14, 2020 at 06:42:06AM +0900, Dmitry Fomichev wrote: > > The emulation code has been changed to advertise NVM Command Set > when > > "zoned" device property is not set (default) and Zoned Namespace > > Command Set otherwise. > > > > Define values and structures that are needed to support Zoned > > Namespace Command Set (NVMe TP 4053) in PCI NVMe controller > emulator. > > Define trace events where needed in newly introduced code. > > > > In order to improve scalability, all open, closed and full zones > > are organized in separate linked lists. Consequently, almost all > > zone operations don't require scanning of the entire zone array > > (which potentially can be quite large) - it is only necessary to > > enumerate one or more zone lists. > > > > Handlers for three new NVMe commands introduced in Zoned Namespace > > Command Set specification are added, namely for Zone Management > > Receive, Zone Management Send and Zone Append. > > > > Device initialization code has been extended to create a proper > > configuration for zoned operation using device properties. > > > > Read/Write command handler is modified to only allow writes at the > > write pointer if the namespace is zoned. For Zone Append command, > > writes implicitly happen at the write pointer and the starting write > > pointer value is returned as the result of the command. Write Zeroes > > handler is modified to add zoned checks that are identical to those > > done as a part of Write flow. > > > > Subsequent commits in this series add ZDE support and checks for > > active and open zone limits. > > > > Signed-off-by: Niklas Cassel <niklas.cas...@wdc.com> > > Signed-off-by: Hans Holmberg <hans.holmb...@wdc.com> > > Signed-off-by: Ajay Joshi <ajay.jo...@wdc.com> > > Signed-off-by: Chaitanya Kulkarni <chaitanya.kulka...@wdc.com> > > Signed-off-by: Matias Bjorling <matias.bjorl...@wdc.com> > > Signed-off-by: Aravind Ramesh <aravind.ram...@wdc.com> > > Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawas...@wdc.com> > > Signed-off-by: Adam Manzanares <adam.manzana...@wdc.com> > > Signed-off-by: Dmitry Fomichev <dmitry.fomic...@wdc.com> > > (snip) > > > @@ -2260,6 +3155,11 @@ static void nvme_select_ns_iocs(NvmeCtrl *n) > > ns->iocs = nvme_cse_iocs_nvm; > > } > > break; > > + case NVME_CSI_ZONED: > > + if (NVME_CC_CSS(n->bar.cc) == NVME_CC_CSS_CSI) { > > + ns->iocs = nvme_cse_iocs_zoned; > > + } > > + break; > > } > > } > > } > > Who knows how this whole command set mess is supposed to work, > since e.g. the Key Value Command Set assigns opcodes for new commands > (Delete, Exist) with a opcode values (0x10,0x14) smaller than the > current highest opcode value (0x15) in the NVM Command Set, > while those opcodes (0x10,0x14) are reserved in the NVM Command Set. > > At least for Zoned Command Set, they defined the new commands > (Zone Mgmt Send, Zone Mgmt Recv) to opcode values (0x79,0x7a) > that are higher than the current highest opcode value in the > NVM Command Set. > > So since we know that the Zoned Command Set is a strict superset of > the NVM Command Set, I guess it might be nice to do something like: > > case NVME_CSI_ZONED: > if (NVME_CC_CSS(n->bar.cc) == NVME_CC_CSS_CSI) { > ns->iocs = nvme_cse_iocs_zoned; > } else if (NVME_CC_CSS(n->bar.cc) == NVME_CC_CSS_NVM) { > ns->iocs = nvme_cse_iocs_nvm; > } > break; > > > Since I assume that the spec people intended reads/writes > to a ZNS namespace to still be possible when CC_CSS == NVM, > but who knows?
Yes, I think it should be this way, thanks. Now it is matched with what CSE log reports in this case. > > > Kind regards, > Niklas