hw/block/nvme: add block utilization tracking
PATCH 1/1 hw/block/nvme: add block utilization tracking
From: Klaus Jensen <[email protected]>
This adds support for reporting the Deallocated or Unwritten Logical Block error (DULBE). This requires tracking the allocated/deallocated status of all logical blocks.
Introduce a bitmap that does this. The bitmap is always intialized to all ones (aka, all blocks are allocated) on boot up. Blocks can then be specifically deallocated using Write Zeroes. This ensures that we can always guarantee zeroes to be returned from deallocated blocks.
When the device gains support for persistent state, the intention is to remove the "allocated by default" behavior.
--- hw/block/nvme-ns.h | 10 ++++++ include/block/nvme.h | 5 +++ hw/block/nvme-ns.c | 10 ++++++ hw/block/nvme.c | 81 +++++++++++++++++++++++++++++++++++++++++++ hw/block/trace-events | 2 ++ 5 files changed, 108 insertions(+)
hw/block/nvme-ns.h +10 -0
@@ -30,7 +30,17 @@ typedef struct NvmeNamespace {int64_t size;NvmeIdNs id_ns; +struct { +struct { +unsigned long *map; + } utilization; + } state; +NvmeNamespaceParams params; + +struct { +uint32_t err_rec; + } features; } NvmeNamespace;static inline uint32_t nvme_nsid (NvmeNamespace *ns)
include/block/nvme.h +5 -0
@@ -686,6 +686,7 @@ enum NvmeStatusCodes {NVME_E2E_REF_ERROR =0x0284 ,NVME_CMP_FAILURE =0x0285 ,NVME_ACCESS_DENIED =0x0286 , +NVME_DULB =0x0287 ,NVME_MORE =0x2000 ,NVME_DNR =0x4000 , NVME_NO_COMPLETE = 0xffff,
@@ -902,6 +903,9 @@ enum NvmeIdCtrlLpa {#define NVME_AEC_NS_ATTR (aec) ((aec >> 8) & 0x1)#define NVME_AEC_FW_ACTIVATION (aec) ((aec >> 9) & 0x1) +#define NVME_ERR_REC_TLER (err_rec) (err_rec & 0xffff) +#define NVME_ERR_REC_DULBE (err_rec) (err_rec & 0x10000) +enum NvmeFeatureIds {NVME_ARBITRATION =0x1 ,NVME_POWER_MANAGEMENT =0x2 ,
@@ -1022,6 +1026,7 @@ enum NvmeNsIdentifierType {#define NVME_ID_NS_NSFEAT_THIN (nsfeat) ((nsfeat & 0x1)) +#define NVME_ID_NS_NSFEAT_DULBE (nsfeat) ((nsfeat >> 2) & 0x1)#define NVME_ID_NS_FLBAS_EXTENDED (flbas) ((flbas >> 4) & 0x1)#define NVME_ID_NS_FLBAS_INDEX (flbas) ((flbas & 0xf))#define NVME_ID_NS_MC_SEPARATE (mc) ((mc >> 1) & 0x1)
hw/block/nvme-ns.c +10 -0
@@ -47,6 +47,9 @@ static void nvme_ns_init(NvmeNamespace *ns) lba_index =NVME_ID_NS_FLBAS_INDEX (ns->id_ns.flbas); ns->id_ns.lbaf[lba_index].ds =31 -clz32 (ns->blkconf.logical_block_size); + +/* support DULBE */ + id_ns->nsfeat |=0x4 ; }static int nvme_ns_init_blk (NvmeCtrl *n,NvmeNamespace *ns,Error **errp)
@@ -86,6 +89,8 @@ static int nvme_ns_check_constraints(NvmeNamespace *ns, Error **errp)int nvme_ns_setup (NvmeCtrl *n,NvmeNamespace *ns,Error **errp) { +uint64_t nlbas; +if (nvme_ns_check_constraints (ns, errp)) {return -1 ; }
@@ -95,6 +100,11 @@ int nvme_ns_setup(NvmeCtrl *n, NvmeNamespace *ns, Error **errp) }nvme_ns_init (ns); + + nlbas =nvme_ns_nlbas (ns); + ns->state.utilization.map =bitmap_new (nlbas); +bitmap_fill (ns->state.utilization.map, nlbas); +if (nvme_register_namespace (n, ns, errp)) {return -1 ; }
hw/block/nvme.c +81 -0
128 lines not on this page · raw message
hw/block/trace-events +2 -0
@@ -43,6 +43,8 @@ pci_nvme_admin_cmd(uint16_t cid, uint16_t sqid, uint8_t opcode, const char *opna pci_nvme_rw(uint16_t cid, const char *verb, uint32_t nsid, uint32_t nlb, uint64_t count, uint64_t lba) "cid %"PRIu16" opname '%s' nsid %"PRIu32" nlb %"PRIu32" count %"PRIu64" lba 0x%"PRIx64"" pci_nvme_rw_cb(uint16_t cid, const char *blkname) "cid %"PRIu16" blk '%s'" pci_nvme_write_zeroes(uint16_t cid, uint32_t nsid, uint64_t slba, uint32_t nlb) "cid %"PRIu16" nsid %"PRIu32" slba %"PRIu64" nlb %"PRIu32"" +pci_nvme_allocate(uint32_t ns, uint64_t slba, uint32_t nlb) "nsid %"PRIu32" slba 0x%"PRIx64" nlb %"PRIu32"" +pci_nvme_deallocate(uint32_t ns, uint64_t slba, uint32_t nlb) "nsid %"PRIu32" slba 0x%"PRIx64" nlb %"PRIu32"" pci_nvme_create_sq(uint64_t addr, uint16_t sqid, uint16_t cqid, uint16_t qsize, uint16_t qflags) "create submission queue, addr=0x%"PRIx64", sqid=%"PRIu16", cqid=%"PRIu16", qsize=%"PRIu16", qflags=%"PRIu16"" pci_nvme_create_cq(uint64_t addr, uint16_t cqid, uint16_t vector, uint16_t size, uint16_t qflags, int ien) "create completion queue, addr=0x%"PRIx64", cqid=%"PRIu16", vector=%"PRIu16", qsize=%"PRIu16", qflags=%"PRIu16", ien=%d" pci_nvme_del_sq(uint16_t qid) "deleting submission queue sqid=%"PRIu16""
Signature · 1 line
2.28.0
Quoted 6 lines · Klaus Jensen
From: Klaus Jensen <[email protected]>
This adds support for reporting the Deallocated or Unwritten Logical Block error (DULBE). This requires tracking the allocated/deallocated status of all logical blocks.
Quoted 7 lines
Introduce a bitmap that does this. The bitmap is always intialized to all ones (aka, all blocks are allocated) on boot up. Blocks can then be specifically deallocated using Write Zeroes. This ensures that we can always guarantee zeroes to be returned from deallocated blocks.
When the device gains support for persistent state, the intention is to remove the "allocated by default" behavior.
I think this is a rather odd feature for this device to support. The implementation is slow can end up allocating quite a bit of memory. If we were going to support it here, wouldn't it make more sense to tie it to the filesystem's ability to support fallocate hole punch for the backing namespace, and check if the range is allocated with SEEK_HOLE? Then you don't even need to track persistent state, and we're actually getting the genuine capability.
Quoted 12 lines · Keith Busch
Quoted 11 lines · Klaus Jensen
From: Klaus Jensen <[email protected]>
This adds support for reporting the Deallocated or Unwritten Logical Block error (DULBE). This requires tracking the allocated/deallocated status of all logical blocks.
Introduce a bitmap that does this. The bitmap is always intialized to all ones (aka, all blocks are allocated) on boot up. Blocks can then be specifically deallocated using Write Zeroes. This ensures that we can always guarantee zeroes to be returned from deallocated blocks.
Quoted 5 lines
Quoted 2 lines
When the device gains support for persistent state, the intention is to remove the "allocated by default" behavior.
I think this is a rather odd feature for this device to support. The implementation is slow can end up allocating quite a bit of memory.
Totally true - but the bitmap was the best way I could find to do this on a per LBA basis.
Quoted 6 lines
If we were going to support it here, wouldn't it make more sense to tie it to the filesystem's ability to support fallocate hole punch for the backing namespace, and check if the range is allocated with
SEEK_HOLE? Then you don't even need to track persistent state, and we're actually getting the genuine capability.
Yes, this would be optimal, definitely. I think we have to rely on what we can do with the QEMU block layer, so please see my v2 that uses bdrv_block_status. I did do something like this initially, but was unsure if we could live with the fact that block drivers such as qcow2 cannot honor a discard unless an entire cluster is discard/zeroed.
But see my commit message, I think we can work with it and still be in compliance with the spec.
Quoted 9 lines · Klaus Jensen
Quoted 6 lines · Keith Busch
If we were going to support it here, wouldn't it make more sense to tie it to the filesystem's ability to support fallocate hole punch for the backing namespace, and check if the range is allocated with
SEEK_HOLE? Then you don't even need to track persistent state, and we're actually getting the genuine capability.
Quoted 8 lines
Yes, this would be optimal, definitely. I think we have to rely on what we can do with the QEMU block layer, so please see my v2 that uses
bdrv_block_status. I did do something like this initially, but was unsure if we could live with the fact that block drivers such as qcow2 cannot honor a discard unless an entire cluster is discard/zeroed.But see my commit message, I think we can work with it and still be in compliance with the spec.
Yes, I think that's better, though I'm not familiar with that particular API. I definitely prefer implementing spec features in useful ways that work with other layers rather than faking it.