nvme: allow cmb and pmr emulation on same device
PATCH RESEND v2 1/1 nvme: allow cmb and pmr emulation on same device
Hi All,
Resending series recently posted on mailing list related to nvme device extension.
This patch series does following:
- Fixes problem where CMBS bit was not set in controller capabilities register, so support for CMB was not correctly advertised to guest. This is resend of patch that has been submitted and reviewed by Klaus [1]
- Introduces BAR4 sharing between MSI-X vectors and CMB. This allows to have CMB and PMR emulated on the same device. This extension was indicated by Keith [2]
v2:
- rebase on Kevin's latest block branch (Klaus [3])
- improved comments section (Klaus [3])
- simplified calculation of BAR4 size (Klaus [3])
v1:
- initial push of the patch
[1]: https://lore.kernel.org/qemu-devel/[email protected]/
[2]: https://lore.kernel.org/qemu-devel/[email protected]/
[3]: https://lore.kernel.org/qemu-devel/[email protected]/
PATCH RESEND v2 1/2 nvme: indicate CMB support through controller capabilities register
This patch sets CMBS bit in controller capabilities register when user configures NVMe driver with CMB support, so capabilites are correctly reported to guest OS.
--- hw/block/nvme.c | 2 +- include/block/nvme.h | 6 +++++- 2 files changed, 6 insertions(+), 2 deletions(-)
hw/block/nvme.c +1 -1
@@ -1582,6 +1582,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev)NVME_CAP_SET_TO (n->bar.cap,0xf );NVME_CAP_SET_CSS (n->bar.cap,1 );NVME_CAP_SET_MPSMAX (n->bar.cap,4 ); +NVME_CAP_SET_CMBS (n->bar.cap, n->params.cmb_size_mb ?1 :0 ); n->bar.vs =0x00010200 ; n->bar.intmc = n->bar.intms =0 ;
@@ -1591,7 +1592,6 @@ static void nvme_realize(PCIDevice *pci_dev, Error **errp) {NvmeCtrl *n =NVME (pci_dev);Error *local_err =NULL ; -int i;nvme_check_constraints (n, &local_err);
include/block/nvme.h +5 -1
@@ -35,6 +35,7 @@ enum NvmeCapShift {CAP_MPSMIN_SHIFT =48 ,CAP_MPSMAX_SHIFT =52 ,CAP_PMR_SHIFT =56 , + CAP_CMB_SHIFT = 57, }; enum NvmeCapMask {
@@ -48,6 +49,7 @@ enum NvmeCapMask {CAP_MPSMIN_MASK =0xf ,CAP_MPSMAX_MASK =0xf ,CAP_PMR_MASK =0x1 , + CAP_CMB_MASK = 0x1, };#define NVME_CAP_MQES (cap) (((cap) >> CAP_MQES_SHIFT) & CAP_MQES_MASK)
@@ -78,8 +80,10 @@ enum NvmeCapMask { << CAP_MPSMIN_SHIFT)#define NVME_CAP_SET_MPSMAX (cap, val) (cap |= (uint64_t)(val & CAP_MPSMAX_MASK)\ << CAP_MPSMAX_SHIFT) -#define NVME_CAP_SET_PMRS (cap, val) (cap |= (uint64_t)(val & CAP_PMR_MASK)\ +#define NVME_CAP_SET_PMRS (cap, val) (cap |= (uint64_t)(val & CAP_PMR_MASK) \ << CAP_PMR_SHIFT) +#define NVME_CAP_SET_CMBS (cap, val) (cap |= (uint64_t)(val & CAP_CMB_MASK) \ + << CAP_CMB_SHIFT)enum NvmeCcShift {CC_EN_SHIFT =0 ,
Signature · 1 line
2.21.1
PATCH RESEND v2 2/2 nvme: allow cmb and pmr to be enabled on same device
In reply to Andrzej JakowskiSo far it was not possible to have CMB and PMR emulated on the same device, because BAR2 was used exclusively either of PMR or CMB. This patch places CMB at BAR4 offset so it not conflicts with MSI-X vectors.
--- hw/block/nvme.c | 122 ++++++++++++++++++++++++++++--------------- hw/block/nvme.h | 5 +- include/block/nvme.h | 4 +- 3 files changed, 86 insertions(+), 45 deletions(-)
hw/block/nvme.c +79 -43
@@ -22,12 +22,12 @@* [pmrdev=<mem_backend_file_id>,] \ * max_ioqpairs=<N[optional]> * -* Note cmb_size_mb denotes size of CMB in MB. CMB is assumed to be at -* offset 0 in BAR2 and supports only WDS, RDS and SQS for now. +* Note cmb_size_mb denotes size of CMB in MB. CMB when configured is assumed +* to be resident in BAR4 at certain offset - this is because BAR4 is also +* used for storing MSI-X table that is available at offset 0 in BAR4. * -* cmb_size_mb= and pmrdev= options are mutually exclusive due to limitation -* in available BAR's. cmb_size_mb= will take precedence over pmrdev= when -* both provided. +* pmrdev is assumed to be resident in BAR2/BAR3. When configured it consumes +* whole BAR2/BAR3 exclusively. * Enabling pmr emulation can be achieved by pointing to memory-backend-file. * For example: * -object memory-backend-file,id=<mem_id>,share=on,mem-path=<file_path>, \
@@ -69,18 +69,19 @@static void nvme_process_sq (void *opaque); -static bool nvme_addr_is_cmb (NvmeCtrl *n,hwaddr addr) +static bool nvme_addr_is_cmb (NvmeCtrl *n,hwaddr addr,int size) { -hwaddr low = n->ctrl_mem.addr; -hwaddr hi = n->ctrl_mem.addr +int128_get64 (n->ctrl_mem.size); +hwaddr low = n->bar4.addr + n->cmb_offset; +hwaddr hi = low +NVME_CMBSZ_GETSIZE (n->bar.cmbsz); -return addr >= low && addr < hi; +return addr >= low && (addr + size) < hi; }static void nvme_addr_read (NvmeCtrl *n,hwaddr addr,void *buf,int size) { -if (n->bar.cmbsz &&nvme_addr_is_cmb (n, addr)) { -memcpy (buf, (void *)&n->cmbuf[addr - n->ctrl_mem.addr], size); +hwaddr cmb_addr = n->bar4.addr + n->cmb_offset; +if (n->cmbsz &&nvme_addr_is_cmb (n, addr, size)) { +memcpy (buf, (void *)&n->cmbuf[addr - cmb_addr], size);return ; }
@@ -167,17 +168,18 @@ static uint16_t nvme_map_prp(QEMUSGList *qsg, QEMUIOVector *iov, uint64_t prp1, uint64_t prp2, uint32_t len, NvmeCtrl *n) { hwaddr trans_len = n->page_size - (prp1 % n->page_size); +hwaddr cmb_addr = n->bar4.addr + n->cmb_offset; trans_len =MIN (len, trans_len);int num_prps = (len >> n->page_bits) +1 ;if (unlikely (!prp1)) {trace_pci_nvme_err_invalid_prp ();return NVME_INVALID_FIELD |NVME_DNR ; - }else if (n->bar.cmbsz && prp1 >= n->ctrl_mem.addr && - prp1 < n->ctrl_mem.addr +int128_get64 (n->ctrl_mem.size)) { + }else if (n->cmbsz && prp1 >= cmb_addr && + prp1 < cmb_addr +int128_get64 (n->bar4.size)) { qsg->nsg =0 ;qemu_iovec_init (iov, num_prps); -qemu_iovec_add (iov, (void *)&n->cmbuf[prp1 - n->ctrl_mem.addr], trans_len); +qemu_iovec_add (iov, (void *)&n->cmbuf[prp1 - cmb_addr], trans_len); }else {pci_dma_sglist_init (qsg, &n->parent_obj, num_prps);qemu_sglist_add (qsg, prp1, trans_len);
@@ -222,7 +224,8 @@ static uint16_t nvme_map_prp(QEMUSGList *qsg, QEMUIOVector *iov, uint64_t prp1,if (qsg->nsg){qemu_sglist_add (qsg, prp_ent, trans_len); }else { -qemu_iovec_add (iov, (void *)&n->cmbuf[prp_ent - n->ctrl_mem.addr], trans_len); +qemu_iovec_add (iov, (void *)&n->cmbuf[prp_ent - cmb_addr], + trans_len); } len -= trans_len; i++;
@@ -235,7 +238,8 @@ static uint16_t nvme_map_prp(QEMUSGList *qsg, QEMUIOVector *iov, uint64_t prp1,if (qsg->nsg) {qemu_sglist_add (qsg, prp2, len); }else { -qemu_iovec_add (iov, (void *)&n->cmbuf[prp2 - n->ctrl_mem.addr], trans_len); +qemu_iovec_add (iov, (void *)&n->cmbuf[prp2 - cmb_addr], + trans_len); } } }
@@ -1360,6 +1364,7 @@ static const MemoryRegionOps nvme_cmb_ops = { }, }; +static void nvme_check_constraints (NvmeCtrl *n,Error **errp) {NvmeParams *params = &n->params;
@@ -1395,7 +1400,7 @@ static void nvme_check_constraints(NvmeCtrl *n, Error **errp)return ; } -if (!n->params.cmb_size_mb && n->pmrdev) { +if (n->pmrdev) {if (host_memory_backend_is_mapped (n->pmrdev)) {char *path =object_get_canonical_path_component (OBJECT (n->pmrdev));error_setg (errp,"can't use already busy memdev: %s" , path);
@@ -1453,33 +1458,66 @@ static void nvme_init_namespace(NvmeCtrl *n, NvmeNamespace *ns, Error **errp) id_ns->nuse = id_ns->ncap; } -static void nvme_init_cmb (NvmeCtrl *n,PCIDevice *pci_dev) +#define NVME_MSIX_BIR (4) +static void nvme_bar4_init (PCIDevice *pci_dev,Error **errp) { -NVME_CMBLOC_SET_BIR (n->bar.cmbloc,NVME_CMB_BIR ); -NVME_CMBLOC_SET_OFST (n->bar.cmbloc,0 ); - -NVME_CMBSZ_SET_SQS (n->bar.cmbsz,1 ); -NVME_CMBSZ_SET_CQS (n->bar.cmbsz,0 ); -NVME_CMBSZ_SET_LISTS (n->bar.cmbsz,0 ); -NVME_CMBSZ_SET_RDS (n->bar.cmbsz,1 ); -NVME_CMBSZ_SET_WDS (n->bar.cmbsz,1 ); -NVME_CMBSZ_SET_SZU (n->bar.cmbsz,2 );/* MBs */ -NVME_CMBSZ_SET_SZ (n->bar.cmbsz, n->params.cmb_size_mb); - - n->cmbuf =g_malloc0 (NVME_CMBSZ_GETSIZE (n->bar.cmbsz)); -memory_region_init_io (&n->ctrl_mem,OBJECT (n), &nvme_cmb_ops, n, -"nvme-cmb" ,NVME_CMBSZ_GETSIZE (n->bar.cmbsz)); -pci_register_bar (pci_dev,NVME_CMBLOC_BIR (n->bar.cmbloc), +NvmeCtrl *n =NVME (pci_dev); +int status; +uint64_t bar_size; +uint32_t msix_vectors; +uint32_t nvme_pba_offset; +uint32_t cmb_size_units; + + msix_vectors = n->params.max_ioqpairs +1 ; + nvme_pba_offset =PCI_MSIX_ENTRY_SIZE * msix_vectors; + bar_size = nvme_pba_offset +QEMU_ALIGN_UP (msix_vectors,64 ) /8 ; + +if (n->params.cmb_size_mb) { +NVME_CMBSZ_SET_SQS (n->bar.cmbsz,1 ); +NVME_CMBSZ_SET_CQS (n->bar.cmbsz,0 ); +NVME_CMBSZ_SET_LISTS (n->bar.cmbsz,0 ); +NVME_CMBSZ_SET_RDS (n->bar.cmbsz,1 ); +NVME_CMBSZ_SET_WDS (n->bar.cmbsz,1 ); +NVME_CMBSZ_SET_SZU (n->bar.cmbsz,2 );/* MBs */ +NVME_CMBSZ_SET_SZ (n->bar.cmbsz, n->params.cmb_size_mb); + + cmb_size_units =NVME_CMBSZ_GETSIZEUNITS (n->bar.cmbsz); + n->cmb_offset =QEMU_ALIGN_UP (bar_size, cmb_size_units); + +NVME_CMBLOC_SET_BIR (n->bar.cmbloc,NVME_MSIX_BIR ); +NVME_CMBLOC_SET_OFST (n->bar.cmbloc, n->cmb_offset / cmb_size_units); + + n->cmbloc = n->bar.cmbloc; + n->cmbsz = n->bar.cmbsz; + + n->cmbuf =g_malloc0 (NVME_CMBSZ_GETSIZE (n->bar.cmbsz)); + + bar_size += n->cmb_offset; + bar_size +=NVME_CMBSZ_GETSIZE (n->bar.cmbsz); + } + + bar_size =pow2ceil (bar_size); + +memory_region_init_io (&n->bar4,OBJECT (n), &nvme_cmb_ops, n, +"nvme-bar4" , bar_size); + + status =msix_init (pci_dev, n->params.max_ioqpairs +1 , + &n->bar4,NVME_MSIX_BIR ,0 , + &n->bar4,NVME_MSIX_BIR , nvme_pba_offset, +0 , errp); + +if (status) { +return ; + } + +pci_register_bar (pci_dev,NVME_MSIX_BIR ,PCI_BASE_ADDRESS_SPACE_MEMORY |PCI_BASE_ADDRESS_MEM_TYPE_64 | -PCI_BASE_ADDRESS_MEM_PREFETCH , &n->ctrl_mem); +PCI_BASE_ADDRESS_MEM_PREFETCH , &n->bar4); }static void nvme_init_pmr (NvmeCtrl *n,PCIDevice *pci_dev) { -/* Controller Capabilities register */ -NVME_CAP_SET_PMRS (n->bar.cap,1 ); -/* PMR Capabities register */ n->bar.pmrcap =0 ;NVME_PMRCAP_SET_RDS (n->bar.pmrcap,0 );
@@ -1537,13 +1575,10 @@ static void nvme_init_pci(NvmeCtrl *n, PCIDevice *pci_dev, Error **errp) n->reg_size);pci_register_bar (pci_dev,0 ,PCI_BASE_ADDRESS_SPACE_MEMORY |PCI_BASE_ADDRESS_MEM_TYPE_64 , &n->iomem); -if (msix_init_exclusive_bar (pci_dev, n->params.msix_qsize,4 , errp)) { -return ; - } -if (n->params.cmb_size_mb) { -nvme_init_cmb (n, pci_dev); - }else if (n->pmrdev) { +nvme_bar4_init (pci_dev, errp); + +if (n->pmrdev) {nvme_init_pmr (n, pci_dev); } }
@@ -1583,6 +1618,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev)NVME_CAP_SET_CSS (n->bar.cap,1 );NVME_CAP_SET_MPSMAX (n->bar.cap,4 );NVME_CAP_SET_CMBS (n->bar.cap, n->params.cmb_size_mb ?1 :0 ); +NVME_CAP_SET_PMRS (n->bar.cap, n->pmrdev ?1 :0 ); n->bar.vs =0x00010200 ; n->bar.intmc = n->bar.intms =0 ;
hw/block/nvme.h +4 -1
@@ -80,7 +80,7 @@ static inline uint8_t nvme_ns_lbads(NvmeNamespace *ns) typedef struct NvmeCtrl { PCIDevice parent_obj; MemoryRegion iomem; - MemoryRegion ctrl_mem; + MemoryRegion bar4; NvmeBar bar; BlockConf conf; NvmeParams params;
@@ -94,6 +94,9 @@ typedef struct NvmeCtrl {uint32_t num_namespaces;uint32_t max_q_ents;uint64_t ns_size; +uint32_t cmbsz; +uint32_t cmbloc; +uint32_t cmb_offset;uint8_t *cmbuf;uint32_t irq_status;uint64_t host_timestamp;/* Timestamp sent by the host */
include/block/nvme.h +3 -1
@@ -216,9 +216,11 @@ enum NvmeCmbszMask { (cmbsz |= (uint64_t)(val & CMBSZ_SZU_MASK) << CMBSZ_SZU_SHIFT)#define NVME_CMBSZ_SET_SZ (cmbsz, val) \ (cmbsz |= (uint64_t)(val & CMBSZ_SZ_MASK) << CMBSZ_SZ_SHIFT) +#define NVME_CMBSZ_GETSIZEUNITS (cmbsz) \ + (1 << (12 + 4 * NVME_CMBSZ_SZU(cmbsz)))#define NVME_CMBSZ_GETSIZE (cmbsz) \ - (NVME_CMBSZ_SZ(cmbsz) * (1 << (12 + 4 * NVME_CMBSZ_SZU(cmbsz)))) + (NVME_CMBSZ_SZ(cmbsz) * NVME_CMBSZ_GETSIZEUNITS(cmbsz))enum NvmePmrcapShift {PMRCAP_RDS_SHIFT =3 ,
Signature · 1 line
2.21.1
Re: [PATCH RESEND v2 2/2] nvme: allow cmb and pmr to be enabled on same device
Quoted 16 lines · Andrzej Jakowski
So far it was not possible to have CMB and PMR emulated on the same device, because BAR2 was used exclusively either of PMR or CMB. This patch places CMB at BAR4 offset so it not conflicts with MSI-X vectors.
Signed-off-by Andrzej Jakowski <[email protected]>--- hw/block/nvme.c | 122 ++++++++++++++++++++++++++++--------------- hw/block/nvme.h | 5 +- include/block/nvme.h | 4 +- 3 files changed, 86 insertions(+), 45 deletions(-)hw/block/nvme.c +0
-0
hw/block/nvme.c line 27
Quoted 8 lines
hw/block/nvme.c +3
-2@@ -22,12 +22,12 @@* [pmrdev=<mem_backend_file_id>,] \ * max_ioqpairs=<N[optional]> * -* Note cmb_size_mb denotes size of CMB in MB. CMB is assumed to be at -* offset 0 in BAR2 and supports only WDS, RDS and SQS for now. +* Note cmb_size_mb denotes size of CMB in MB. CMB when configured is assumed +* to be resident in BAR4 at certain offset - this is because BAR4 is also +* used for storing MSI-X table that is available at offset 0 in BAR4.
I would still really like a R-b by a more PCI-competent reviewer to ensure that it is sane to have the MSI-X table here in prefetchable 64-bit address space.
Quoted 10 lines
hw/block/nvme.c +2
-3* -* cmb_size_mb= and pmrdev= options are mutually exclusive due to limitation -* in available BAR's. cmb_size_mb= will take precedence over pmrdev= when -* both provided. +* pmrdev is assumed to be resident in BAR2/BAR3. When configured it consumes +* whole BAR2/BAR3 exclusively. * Enabling pmr emulation can be achieved by pointing to memory-backend-file. * For example: * -object memory-backend-file,id=<mem_id>,share=on,mem-path=<file_path>, \
hw/block/nvme.c line 75
Quoted 10 lines
hw/block/nvme.c +3
-3@@ -69,18 +69,19 @@static void nvme_process_sq (void *opaque); -static bool nvme_addr_is_cmb (NvmeCtrl *n,hwaddr addr) +static bool nvme_addr_is_cmb (NvmeCtrl *n,hwaddr addr,int size) { -hwaddr low = n->ctrl_mem.addr; -hwaddr hi = n->ctrl_mem.addr +int128_get64 (n->ctrl_mem.size); +hwaddr low = n->bar4.addr + n->cmb_offset; +hwaddr hi = low +NVME_CMBSZ_GETSIZE (n->bar.cmbsz);
Isn't it better to use n->bar4.size?
hw/block/nvme.c line 78
Quoted 4 lines
hw/block/nvme.c +1
-1-return addr >= low && addr < hi; +return addr >= low && (addr + size) < hi; }
I think nvme_addr_is_cmb should do what it says on the tin - that is, check that addr is within the CMB. The size check belongs in nvme_addr_read.
Quoted 54 lines
hw/block/nvme.c +11
-7static void nvme_addr_read (NvmeCtrl *n,hwaddr addr,void *buf,int size) { -if (n->bar.cmbsz &&nvme_addr_is_cmb (n, addr)) { -memcpy (buf, (void *)&n->cmbuf[addr - n->ctrl_mem.addr], size); +hwaddr cmb_addr = n->bar4.addr + n->cmb_offset; +if (n->cmbsz &&nvme_addr_is_cmb (n, addr, size)) { +memcpy (buf, (void *)&n->cmbuf[addr - cmb_addr], size);return ; }@@ -167,17 +168,18 @@ static uint16_t nvme_map_prp(QEMUSGList *qsg, QEMUIOVector *iov, uint64_t prp1, uint64_t prp2, uint32_t len, NvmeCtrl *n) { hwaddr trans_len = n->page_size - (prp1 % n->page_size); +hwaddr cmb_addr = n->bar4.addr + n->cmb_offset; trans_len =MIN (len, trans_len);int num_prps = (len >> n->page_bits) +1 ;if (unlikely (!prp1)) {trace_pci_nvme_err_invalid_prp ();return NVME_INVALID_FIELD |NVME_DNR ; - }else if (n->bar.cmbsz && prp1 >= n->ctrl_mem.addr && - prp1 < n->ctrl_mem.addr +int128_get64 (n->ctrl_mem.size)) { + }else if (n->cmbsz && prp1 >= cmb_addr && + prp1 < cmb_addr +int128_get64 (n->bar4.size)) { qsg->nsg =0 ;qemu_iovec_init (iov, num_prps); -qemu_iovec_add (iov, (void *)&n->cmbuf[prp1 - n->ctrl_mem.addr], trans_len); +qemu_iovec_add (iov, (void *)&n->cmbuf[prp1 - cmb_addr], trans_len); }else {pci_dma_sglist_init (qsg, &n->parent_obj, num_prps);qemu_sglist_add (qsg, prp1, trans_len);@@ -222,7 +224,8 @@ static uint16_t nvme_map_prp(QEMUSGList *qsg, QEMUIOVector *iov, uint64_t prp1,if (qsg->nsg){qemu_sglist_add (qsg, prp_ent, trans_len); }else { -qemu_iovec_add (iov, (void *)&n->cmbuf[prp_ent - n->ctrl_mem.addr], trans_len); +qemu_iovec_add (iov, (void *)&n->cmbuf[prp_ent - cmb_addr], + trans_len); } len -= trans_len; i++;@@ -235,7 +238,8 @@ static uint16_t nvme_map_prp(QEMUSGList *qsg, QEMUIOVector *iov, uint64_t prp1,if (qsg->nsg) {qemu_sglist_add (qsg, prp2, len); }else { -qemu_iovec_add (iov, (void *)&n->cmbuf[prp2 - n->ctrl_mem.addr], trans_len); +qemu_iovec_add (iov, (void *)&n->cmbuf[prp2 - cmb_addr], + trans_len); } } }
hw/block/nvme.c static const MemoryRegionOps nvme_cmb_ops = { · line 1367
Quoted 4 lines
hw/block/nvme.c +1
-0@@ -1360,6 +1364,7 @@ static const MemoryRegionOps nvme_cmb_ops = { }, }; +
Spurious line break.
Quoted 13 lines
hw/block/nvme.c +1
-1static void nvme_check_constraints (NvmeCtrl *n,Error **errp) {NvmeParams *params = &n->params;@@ -1395,7 +1400,7 @@ static void nvme_check_constraints(NvmeCtrl *n, Error **errp)return ; } -if (!n->params.cmb_size_mb && n->pmrdev) { +if (n->pmrdev) {if (host_memory_backend_is_mapped (n->pmrdev)) {char *path =object_get_canonical_path_component (OBJECT (n->pmrdev));error_setg (errp,"can't use already busy memdev: %s" , path);
hw/block/nvme.c static void nvme_init_namespace(NvmeCtrl *n, NvmeNamespace *ns, Error **errp) · line 1461
Quoted 5 lines
hw/block/nvme.c +1
-1@@ -1453,33 +1458,66 @@ static void nvme_init_namespace(NvmeCtrl *n, NvmeNamespace *ns, Error **errp) id_ns->nuse = id_ns->ncap; } -static void nvme_init_cmb (NvmeCtrl *n,PCIDevice *pci_dev) +#define NVME_MSIX_BIR (4)
I think this should be at the top with the other constants.
Quoted 51 lines
hw/block/nvme.c +35
-15+static void nvme_bar4_init (PCIDevice *pci_dev,Error **errp) { -NVME_CMBLOC_SET_BIR (n->bar.cmbloc,NVME_CMB_BIR ); -NVME_CMBLOC_SET_OFST (n->bar.cmbloc,0 ); - -NVME_CMBSZ_SET_SQS (n->bar.cmbsz,1 ); -NVME_CMBSZ_SET_CQS (n->bar.cmbsz,0 ); -NVME_CMBSZ_SET_LISTS (n->bar.cmbsz,0 ); -NVME_CMBSZ_SET_RDS (n->bar.cmbsz,1 ); -NVME_CMBSZ_SET_WDS (n->bar.cmbsz,1 ); -NVME_CMBSZ_SET_SZU (n->bar.cmbsz,2 );/* MBs */ -NVME_CMBSZ_SET_SZ (n->bar.cmbsz, n->params.cmb_size_mb); - - n->cmbuf =g_malloc0 (NVME_CMBSZ_GETSIZE (n->bar.cmbsz)); -memory_region_init_io (&n->ctrl_mem,OBJECT (n), &nvme_cmb_ops, n, -"nvme-cmb" ,NVME_CMBSZ_GETSIZE (n->bar.cmbsz)); - pci_register_bar(pci_dev, NVME_CMBLOC_BIR(n->bar.cmbloc), +NvmeCtrl *n =NVME (pci_dev); +int status; +uint64_t bar_size; +uint32_t msix_vectors; +uint32_t nvme_pba_offset; +uint32_t cmb_size_units; + + msix_vectors = n->params.max_ioqpairs +1 ; + nvme_pba_offset =PCI_MSIX_ENTRY_SIZE * msix_vectors; + bar_size = nvme_pba_offset +QEMU_ALIGN_UP (msix_vectors,64 ) /8 ; + +if (n->params.cmb_size_mb) { +NVME_CMBSZ_SET_SQS (n->bar.cmbsz,1 ); +NVME_CMBSZ_SET_CQS (n->bar.cmbsz,0 ); +NVME_CMBSZ_SET_LISTS (n->bar.cmbsz,0 ); +NVME_CMBSZ_SET_RDS (n->bar.cmbsz,1 ); +NVME_CMBSZ_SET_WDS (n->bar.cmbsz,1 ); +NVME_CMBSZ_SET_SZU (n->bar.cmbsz,2 );/* MBs */ +NVME_CMBSZ_SET_SZ (n->bar.cmbsz, n->params.cmb_size_mb); + + cmb_size_units =NVME_CMBSZ_GETSIZEUNITS (n->bar.cmbsz); + n->cmb_offset =QEMU_ALIGN_UP (bar_size, cmb_size_units); + +NVME_CMBLOC_SET_BIR (n->bar.cmbloc,NVME_MSIX_BIR ); +NVME_CMBLOC_SET_OFST (n->bar.cmbloc, n->cmb_offset / cmb_size_units); + + n->cmbloc = n->bar.cmbloc; + n->cmbsz = n->bar.cmbsz; + + n->cmbuf =g_malloc0 (NVME_CMBSZ_GETSIZE (n->bar.cmbsz)); + + bar_size += n->cmb_offset; + bar_size +=NVME_CMBSZ_GETSIZE (n->bar.cmbsz); + }
hw/block/nvme.c static void nvme_init_namespace(NvmeCtrl *n, NvmeNamespace *ns, Error **errp) · line 1507
Quoted 10 lines
hw/block/nvme.c +10
-0+ + bar_size =pow2ceil (bar_size); + +memory_region_init_io (&n->bar4,OBJECT (n), &nvme_cmb_ops, n, +"nvme-bar4" , bar_size); + + status =msix_init (pci_dev, n->params.max_ioqpairs +1 , + &n->bar4,NVME_MSIX_BIR ,0 , + &n->bar4,NVME_MSIX_BIR , nvme_pba_offset, +0 , errp);
Align with opening brace.
Quoted 59 lines
hw/block/nvme.c +11
-10+ +if (status) { +return ; + } + +pci_register_bar (pci_dev,NVME_MSIX_BIR ,PCI_BASE_ADDRESS_SPACE_MEMORY |PCI_BASE_ADDRESS_MEM_TYPE_64 | - PCI_BASE_ADDRESS_MEM_PREFETCH, &n->ctrl_mem); +PCI_BASE_ADDRESS_MEM_PREFETCH , &n->bar4); }static void nvme_init_pmr (NvmeCtrl *n,PCIDevice *pci_dev) { -/* Controller Capabilities register */ -NVME_CAP_SET_PMRS (n->bar.cap,1 ); -/* PMR Capabities register */ n->bar.pmrcap =0 ;NVME_PMRCAP_SET_RDS (n->bar.pmrcap,0 );@@ -1537,13 +1575,10 @@ static void nvme_init_pci(NvmeCtrl *n, PCIDevice *pci_dev, Error **errp) n->reg_size);pci_register_bar (pci_dev,0 ,PCI_BASE_ADDRESS_SPACE_MEMORY |PCI_BASE_ADDRESS_MEM_TYPE_64 , &n->iomem); -if (msix_init_exclusive_bar (pci_dev, n->params.msix_qsize,4 , errp)) { -return ; - } -if (n->params.cmb_size_mb) { -nvme_init_cmb (n, pci_dev); - }else if (n->pmrdev) { +nvme_bar4_init (pci_dev, errp); + +if (n->pmrdev) {nvme_init_pmr (n, pci_dev); } }@@ -1583,6 +1618,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev)NVME_CAP_SET_CSS (n->bar.cap,1 );NVME_CAP_SET_MPSMAX (n->bar.cap,4 );NVME_CAP_SET_CMBS (n->bar.cap, n->params.cmb_size_mb ?1 :0 ); +NVME_CAP_SET_PMRS (n->bar.cap, n->pmrdev ?1 :0 ); n->bar.vs =0x00010200 ; n->bar.intmc = n->bar.intms =0 ;hw/block/nvme.h +1
-1@@ -80,7 +80,7 @@ static inline uint8_t nvme_ns_lbads(NvmeNamespace *ns) typedef struct NvmeCtrl { PCIDevice parent_obj; MemoryRegion iomem; - MemoryRegion ctrl_mem; + MemoryRegion bar4; NvmeBar bar; BlockConf conf; NvmeParams params;
hw/block/nvme.h typedef struct NvmeCtrl { · line 99
Quoted 6 lines
hw/block/nvme.h +3
-0@@ -94,6 +94,9 @@ typedef struct NvmeCtrl {uint32_t num_namespaces;uint32_t max_q_ents;uint64_t ns_size; +uint32_t cmbsz; +uint32_t cmbloc; +uint32_t cmb_offset;
So, in commit c923aac9e00d ("hw/block/nvme: remove redundant cmbloc/cmbsz members") I actually specifically got rid of cmbloc and cmbsz members and they are now being reintroduced :P
Since we are not really using the value, just testing if they are non-zero, can we use n->bar.cmbsz instead? n->cmbloc is not used at all actually.
Quoted 24 lines
hw/block/nvme.h +0
-0uint8_t *cmbuf;uint32_t irq_status;uint64_t host_timestamp;/* Timestamp sent by the host */ include/block/nvme.h +3
-1@@ -216,9 +216,11 @@ enum NvmeCmbszMask { (cmbsz |= (uint64_t)(val & CMBSZ_SZU_MASK) << CMBSZ_SZU_SHIFT)#define NVME_CMBSZ_SET_SZ (cmbsz, val) \ (cmbsz |= (uint64_t)(val & CMBSZ_SZ_MASK) << CMBSZ_SZ_SHIFT) +#define NVME_CMBSZ_GETSIZEUNITS (cmbsz) \ + (1 << (12 + 4 * NVME_CMBSZ_SZU(cmbsz)))#define NVME_CMBSZ_GETSIZE (cmbsz) \ - (NVME_CMBSZ_SZ(cmbsz) * (1 << (12 + 4 * NVME_CMBSZ_SZU(cmbsz)))) + (NVME_CMBSZ_SZ(cmbsz) * NVME_CMBSZ_GETSIZEUNITS(cmbsz))enum NvmePmrcapShift {PMRCAP_RDS_SHIFT =3 ,Signature · 1 line
2.21.1
Re: [PATCH RESEND v2 2/2] nvme: allow cmb and pmr to be enabled on same device
Quoted 21 lines · Klaus Jensen
Quoted 20 lines · Andrzej Jakowski
So far it was not possible to have CMB and PMR emulated on the same device, because BAR2 was used exclusively either of PMR or CMB. This patch places CMB at BAR4 offset so it not conflicts with MSI-X vectors.
Signed-off-by Andrzej Jakowski <[email protected]>--- hw/block/nvme.c | 122 ++++++++++++++++++++++++++++--------------- hw/block/nvme.h | 5 +- include/block/nvme.h | 4 +- 3 files changed, 86 insertions(+), 45 deletions(-)hw/block/nvme.c +0
-1@@ -22,12 +22,12 @@* [pmrdev=<mem_backend_file_id>,] \ * max_ioqpairs=<N[optional]> * -* Note cmb_size_mb denotes size of CMB in MB. CMB is assumed to be at
Quoted 8 lines
hw/block/nvme.c line 27
Quoted 4 lines
hw/block/nvme.c +3
-1-* offset 0 in BAR2 and supports only WDS, RDS and SQS for now. +* Note cmb_size_mb denotes size of CMB in MB. CMB when configured is assumed +* to be resident in BAR4 at certain offset - this is because BAR4 is also +* used for storing MSI-X table that is available at offset 0 in BAR4. I would still really like a R-b by a more PCI-competent reviewer to ensure that it is sane to have the MSI-X table here in prefetchable 64-bit address space.
Having Reviewed-by for that make sense to me. And let me offer some evidences of real devices having MSI-X in prefetchable region and non-prefetchable region. Based on those examples I don't think it matters where you place your MSI-X vector.
Why do you think it may not be sane to place MSI-x table in prefetchable region?
Device with MSI-X in non-prefetchable region:
Region 0: Memory at fb42c000 (64-bit, non-prefetchable) [size=16K]
Capabilities: [80] MSI-X: Enable+ Count=1 Masked-
Vector table: BAR=0 offset=00002000
PBA: BAR=0 offset=00003000Device with MSI-X in prefetchable region
Region 0: Memory at fbc00000 (64-bit, prefetchable) [size=2M]
Region 2: I/O ports at e020 [size=32]
Region 4: Memory at fbe04000 (64-bit, prefetchable) [size=16K]
Capabilities: [40] Power Management version 3
Flags: PMEClk- DSI+ D1- D2- AuxCurrent=0mA PME(D0+,D1-,D2-,D3hot+,D3cold+)
Status: D0 NoSoftRst- PME-Enable- DSel=0 DScale=1 PME-
Capabilities: [50] MSI: Enable- Count=1/1 Maskable+ 64bit+
Address: 0000000000000000 Data: 0000
Masking: 00000000 Pending: 00000000
Capabilities: [70] MSI-X: Enable+ Count=64 Masked-
Vector table: BAR=4 offset=00000000
PBA: BAR=4 offset=00002000Quoted 14 lines
Quoted 14 lines
hw/block/nvme.c +2
-3* -* cmb_size_mb= and pmrdev= options are mutually exclusive due to limitation -* in available BAR's. cmb_size_mb= will take precedence over pmrdev= when -* both provided. +* pmrdev is assumed to be resident in BAR2/BAR3. When configured it consumes +* whole BAR2/BAR3 exclusively. * Enabling pmr emulation can be achieved by pointing to memory-backend-file. * For example: * -object memory-backend-file,id=<mem_id>,share=on,mem-path=<file_path>, \ @@ -69,18 +69,19 @@static void nvme_process_sq (void *opaque);
Quoted 9 lines
hw/block/nvme.c line 75
Quoted 7 lines
hw/block/nvme.c +3
-3-static bool nvme_addr_is_cmb (NvmeCtrl *n,hwaddr addr) +static bool nvme_addr_is_cmb (NvmeCtrl *n,hwaddr addr,int size) { -hwaddr low = n->ctrl_mem.addr; -hwaddr hi = n->ctrl_mem.addr +int128_get64 (n->ctrl_mem.size); +hwaddr low = n->bar4.addr + n->cmb_offset; +hwaddr hi = low +NVME_CMBSZ_GETSIZE (n->bar.cmbsz);Isn't it better to use
n->bar4.size?
My understanding is that cmb doesn't necessarily need to occupy whole BAR, which is required to be power-of-two in size.
Quoted 9 lines
hw/block/nvme.c line 78
Quoted 5 lines
hw/block/nvme.c +1
-1-return addr >= low && addr < hi; +return addr >= low && (addr + size) < hi; }I think
nvme_addr_is_cmbshould do what it says on the tin - that is, check that addr is within the CMB. The size check belongs innvme_addr_read.
Yep, that was my intention. New check confirms that start and end of requested addresses are within CMB range. While previous code only checked start address. You may end up with situation where end adress could be outside of CMB range while start in CMB range.
Quoted 58 lines
Quoted 58 lines
hw/block/nvme.c +11
-7static void nvme_addr_read (NvmeCtrl *n,hwaddr addr,void *buf,int size) { -if (n->bar.cmbsz &&nvme_addr_is_cmb (n, addr)) { -memcpy (buf, (void *)&n->cmbuf[addr - n->ctrl_mem.addr], size); +hwaddr cmb_addr = n->bar4.addr + n->cmb_offset; +if (n->cmbsz &&nvme_addr_is_cmb (n, addr, size)) { +memcpy (buf, (void *)&n->cmbuf[addr - cmb_addr], size);return ; }@@ -167,17 +168,18 @@ static uint16_t nvme_map_prp(QEMUSGList *qsg, QEMUIOVector *iov, uint64_t prp1, uint64_t prp2, uint32_t len, NvmeCtrl *n) { hwaddr trans_len = n->page_size - (prp1 % n->page_size); +hwaddr cmb_addr = n->bar4.addr + n->cmb_offset; trans_len =MIN (len, trans_len);int num_prps = (len >> n->page_bits) +1 ;if (unlikely (!prp1)) {trace_pci_nvme_err_invalid_prp ();return NVME_INVALID_FIELD |NVME_DNR ; - }else if (n->bar.cmbsz && prp1 >= n->ctrl_mem.addr && - prp1 < n->ctrl_mem.addr +int128_get64 (n->ctrl_mem.size)) { + }else if (n->cmbsz && prp1 >= cmb_addr && + prp1 < cmb_addr +int128_get64 (n->bar4.size)) { qsg->nsg =0 ;qemu_iovec_init (iov, num_prps); -qemu_iovec_add (iov, (void *)&n->cmbuf[prp1 - n->ctrl_mem.addr], trans_len); +qemu_iovec_add (iov, (void *)&n->cmbuf[prp1 - cmb_addr], trans_len); }else {pci_dma_sglist_init (qsg, &n->parent_obj, num_prps);qemu_sglist_add (qsg, prp1, trans_len);@@ -222,7 +224,8 @@ static uint16_t nvme_map_prp(QEMUSGList *qsg, QEMUIOVector *iov, uint64_t prp1,if (qsg->nsg){qemu_sglist_add (qsg, prp_ent, trans_len); }else { -qemu_iovec_add (iov, (void *)&n->cmbuf[prp_ent - n->ctrl_mem.addr], trans_len); +qemu_iovec_add (iov, (void *)&n->cmbuf[prp_ent - cmb_addr], + trans_len); } len -= trans_len; i++;@@ -235,7 +238,8 @@ static uint16_t nvme_map_prp(QEMUSGList *qsg, QEMUIOVector *iov, uint64_t prp1,if (qsg->nsg) {qemu_sglist_add (qsg, prp2, len); }else { -qemu_iovec_add (iov, (void *)&n->cmbuf[prp2 - n->ctrl_mem.addr], trans_len); +qemu_iovec_add (iov, (void *)&n->cmbuf[prp2 - cmb_addr], + trans_len); } } }@@ -1360,6 +1364,7 @@ static const MemoryRegionOps nvme_cmb_ops = { }, };
Quoted 3 lines
hw/block/nvme.c static const MemoryRegionOps nvme_cmb_ops = { · line 1367
Quoted 1 line
hw/block/nvme.c +1
-0+Spurious line break.
Yep. Will remove that change
Quoted 17 lines
Quoted 17 lines
hw/block/nvme.c +1
-1static void nvme_check_constraints (NvmeCtrl *n,Error **errp) {NvmeParams *params = &n->params;@@ -1395,7 +1400,7 @@ static void nvme_check_constraints(NvmeCtrl *n, Error **errp)return ; } -if (!n->params.cmb_size_mb && n->pmrdev) { +if (n->pmrdev) {if (host_memory_backend_is_mapped (n->pmrdev)) {char *path =object_get_canonical_path_component (OBJECT (n->pmrdev));error_setg (errp,"can't use already busy memdev: %s" , path);@@ -1453,33 +1458,66 @@ static void nvme_init_namespace(NvmeCtrl *n, NvmeNamespace *ns, Error **errp) id_ns->nuse = id_ns->ncap; }
Quoted 4 lines
hw/block/nvme.c static void nvme_init_namespace(NvmeCtrl *n, NvmeNamespace *ns, Error **errp) · line 1461
Quoted 2 lines
hw/block/nvme.c +1
-1-static void nvme_init_cmb (NvmeCtrl *n,PCIDevice *pci_dev) +#define NVME_MSIX_BIR (4)I think this should be at the top with the other constants.
Makes sense.
Quoted 56 lines
Quoted 56 lines
hw/block/nvme.c +39
-15+static void nvme_bar4_init (PCIDevice *pci_dev,Error **errp) { -NVME_CMBLOC_SET_BIR (n->bar.cmbloc,NVME_CMB_BIR ); -NVME_CMBLOC_SET_OFST (n->bar.cmbloc,0 ); - -NVME_CMBSZ_SET_SQS (n->bar.cmbsz,1 ); -NVME_CMBSZ_SET_CQS (n->bar.cmbsz,0 ); -NVME_CMBSZ_SET_LISTS (n->bar.cmbsz,0 ); -NVME_CMBSZ_SET_RDS (n->bar.cmbsz,1 ); -NVME_CMBSZ_SET_WDS (n->bar.cmbsz,1 ); -NVME_CMBSZ_SET_SZU (n->bar.cmbsz,2 );/* MBs */ -NVME_CMBSZ_SET_SZ (n->bar.cmbsz, n->params.cmb_size_mb); - - n->cmbuf =g_malloc0 (NVME_CMBSZ_GETSIZE (n->bar.cmbsz)); -memory_region_init_io (&n->ctrl_mem,OBJECT (n), &nvme_cmb_ops, n, -"nvme-cmb" ,NVME_CMBSZ_GETSIZE (n->bar.cmbsz)); - pci_register_bar(pci_dev, NVME_CMBLOC_BIR(n->bar.cmbloc), +NvmeCtrl *n =NVME (pci_dev); +int status; +uint64_t bar_size; +uint32_t msix_vectors; +uint32_t nvme_pba_offset; +uint32_t cmb_size_units; + + msix_vectors = n->params.max_ioqpairs +1 ; + nvme_pba_offset =PCI_MSIX_ENTRY_SIZE * msix_vectors; + bar_size = nvme_pba_offset +QEMU_ALIGN_UP (msix_vectors,64 ) /8 ; + +if (n->params.cmb_size_mb) { +NVME_CMBSZ_SET_SQS (n->bar.cmbsz,1 ); +NVME_CMBSZ_SET_CQS (n->bar.cmbsz,0 ); +NVME_CMBSZ_SET_LISTS (n->bar.cmbsz,0 ); +NVME_CMBSZ_SET_RDS (n->bar.cmbsz,1 ); +NVME_CMBSZ_SET_WDS (n->bar.cmbsz,1 ); +NVME_CMBSZ_SET_SZU (n->bar.cmbsz,2 );/* MBs */ +NVME_CMBSZ_SET_SZ (n->bar.cmbsz, n->params.cmb_size_mb); + + cmb_size_units =NVME_CMBSZ_GETSIZEUNITS (n->bar.cmbsz); + n->cmb_offset =QEMU_ALIGN_UP (bar_size, cmb_size_units); + +NVME_CMBLOC_SET_BIR (n->bar.cmbloc,NVME_MSIX_BIR ); +NVME_CMBLOC_SET_OFST (n->bar.cmbloc, n->cmb_offset / cmb_size_units); + + n->cmbloc = n->bar.cmbloc; + n->cmbsz = n->bar.cmbsz; + + n->cmbuf =g_malloc0 (NVME_CMBSZ_GETSIZE (n->bar.cmbsz)); + + bar_size += n->cmb_offset; + bar_size +=NVME_CMBSZ_GETSIZE (n->bar.cmbsz); + } + + bar_size =pow2ceil (bar_size); + + memory_region_init_io(&n->bar4, OBJECT(n), &nvme_cmb_ops, n,
Quoted 8 lines
hw/block/nvme.c static void nvme_init_namespace(NvmeCtrl *n, NvmeNamespace *ns, Error **errp) · line 1507
Quoted 6 lines
hw/block/nvme.c +6
-0+ "nvme-bar4", bar_size); + + status =msix_init (pci_dev, n->params.max_ioqpairs +1 , + &n->bar4,NVME_MSIX_BIR ,0 , + &n->bar4,NVME_MSIX_BIR , nvme_pba_offset, +0 , errp);Align with opening brace.
Yep, makes sense
Quoted 71 lines
Quoted 60 lines
hw/block/nvme.c +11
-10+ +if (status) { +return ; + } + +pci_register_bar (pci_dev,NVME_MSIX_BIR ,PCI_BASE_ADDRESS_SPACE_MEMORY |PCI_BASE_ADDRESS_MEM_TYPE_64 | - PCI_BASE_ADDRESS_MEM_PREFETCH, &n->ctrl_mem); +PCI_BASE_ADDRESS_MEM_PREFETCH , &n->bar4); }static void nvme_init_pmr (NvmeCtrl *n,PCIDevice *pci_dev) { -/* Controller Capabilities register */ -NVME_CAP_SET_PMRS (n->bar.cap,1 ); -/* PMR Capabities register */ n->bar.pmrcap =0 ;NVME_PMRCAP_SET_RDS (n->bar.pmrcap,0 );@@ -1537,13 +1575,10 @@ static void nvme_init_pci(NvmeCtrl *n, PCIDevice *pci_dev, Error **errp) n->reg_size);pci_register_bar (pci_dev,0 ,PCI_BASE_ADDRESS_SPACE_MEMORY |PCI_BASE_ADDRESS_MEM_TYPE_64 , &n->iomem); -if (msix_init_exclusive_bar (pci_dev, n->params.msix_qsize,4 , errp)) { -return ; - } -if (n->params.cmb_size_mb) { -nvme_init_cmb (n, pci_dev); - }else if (n->pmrdev) { +nvme_bar4_init (pci_dev, errp); + +if (n->pmrdev) {nvme_init_pmr (n, pci_dev); } }@@ -1583,6 +1618,7 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev)NVME_CAP_SET_CSS (n->bar.cap,1 );NVME_CAP_SET_MPSMAX (n->bar.cap,4 );NVME_CAP_SET_CMBS (n->bar.cap, n->params.cmb_size_mb ?1 :0 ); +NVME_CAP_SET_PMRS (n->bar.cap, n->pmrdev ?1 :0 ); n->bar.vs =0x00010200 ; n->bar.intmc = n->bar.intms =0 ;hw/block/nvme.h +1
-1@@ -80,7 +80,7 @@ static inline uint8_t nvme_ns_lbads(NvmeNamespace *ns) typedef struct NvmeCtrl { PCIDevice parent_obj; MemoryRegion iomem; - MemoryRegion ctrl_mem; + MemoryRegion bar4; NvmeBar bar; BlockConf conf; NvmeParams params;hw/block/nvme.h typedef struct NvmeCtrl { · line 99
Quoted 6 lines
hw/block/nvme.h +3
-0@@ -94,6 +94,9 @@ typedef struct NvmeCtrl {uint32_t num_namespaces;uint32_t max_q_ents;uint64_t ns_size; +uint32_t cmbsz; +uint32_t cmbloc; +uint32_t cmb_offset;So, in commit
c923aac9e00d("hw/block/nvme: remove redundant cmbloc/cmbsz members") I actually specifically got rid of cmbloc and cmbsz members and they are now being reintroduced :P
Quoted 3 lines
Since we are not really using the value, just testing if they are non-zero, can we use
n->bar.cmbszinstead?n->cmblocis not used at all actually.
Let me take a look on that.
Quoted 25 lines
Quoted 25 lines
hw/block/nvme.h +0
-0uint8_t *cmbuf;uint32_t irq_status;uint64_t host_timestamp;/* Timestamp sent by the host */ include/block/nvme.h +3
-1@@ -216,9 +216,11 @@ enum NvmeCmbszMask { (cmbsz |= (uint64_t)(val & CMBSZ_SZU_MASK) << CMBSZ_SZU_SHIFT)#define NVME_CMBSZ_SET_SZ (cmbsz, val) \ (cmbsz |= (uint64_t)(val & CMBSZ_SZ_MASK) << CMBSZ_SZ_SHIFT) +#define NVME_CMBSZ_GETSIZEUNITS (cmbsz) \ + (1 << (12 + 4 * NVME_CMBSZ_SZU(cmbsz)))#define NVME_CMBSZ_GETSIZE (cmbsz) \ - (NVME_CMBSZ_SZ(cmbsz) * (1 << (12 + 4 * NVME_CMBSZ_SZU(cmbsz)))) + (NVME_CMBSZ_SZ(cmbsz) * NVME_CMBSZ_GETSIZEUNITS(cmbsz))enum NvmePmrcapShift {PMRCAP_RDS_SHIFT =3 ,Signature · 1 line
2.21.1
Re: [PATCH RESEND v2 2/2] nvme: allow cmb and pmr to be enabled on same device
Quoted 47 lines · Andrzej Jakowski
Quoted 21 lines · Klaus Jensen
Quoted 20 lines · Andrzej Jakowski
So far it was not possible to have CMB and PMR emulated on the same device, because BAR2 was used exclusively either of PMR or CMB. This patch places CMB at BAR4 offset so it not conflicts with MSI-X vectors.
Signed-off-by Andrzej Jakowski <[email protected]>--- hw/block/nvme.c | 122 ++++++++++++++++++++++++++++--------------- hw/block/nvme.h | 5 +- include/block/nvme.h | 4 +- 3 files changed, 86 insertions(+), 45 deletions(-)hw/block/nvme.c +0
-1@@ -22,12 +22,12 @@* [pmrdev=<mem_backend_file_id>,] \ * max_ioqpairs=<N[optional]> * -* Note cmb_size_mb denotes size of CMB in MB. CMB is assumed to be at Quoted 8 lines
hw/block/nvme.c line 27
Quoted 4 lines
hw/block/nvme.c +3
-1-* offset 0 in BAR2 and supports only WDS, RDS and SQS for now. +* Note cmb_size_mb denotes size of CMB in MB. CMB when configured is assumed +* to be resident in BAR4 at certain offset - this is because BAR4 is also +* used for storing MSI-X table that is available at offset 0 in BAR4. I would still really like a R-b by a more PCI-competent reviewer to ensure that it is sane to have the MSI-X table here in prefetchable 64-bit address space.
Having Reviewed-by for that make sense to me. And let me offer some evidences of real devices having MSI-X in prefetchable region and non-prefetchable region. Based on those examples I don't think it matters where you place your MSI-X vector.
Why do you think it may not be sane to place MSI-x table in prefetchable region?
Device with MSI-X in non-prefetchable region:
Region 0: Memory atfb42c000(64-bit, non-prefetchable) [size=16K]Capabilities: [80] MSI-X: Enable+ Count=1 Masked- Vector table: BAR=0 offset=00002000 PBA: BAR=0 offset=00003000Device with MSI-X in prefetchable region
Region 0: Memory atfbc00000(64-bit, prefetchable) [size=2M]
Quoted 13 lines
Region 2: I/O ports at e020 [size=32]
Region 4: Memory at fbe04000 (64-bit, prefetchable) [size=16K]
Capabilities: [40] Power Management version 3
Flags: PMEClk- DSI+ D1- D2- AuxCurrent=0mA PME(D0+,D1-,D2-,D3hot+,D3cold+)
Status: D0 NoSoftRst- PME-Enable- DSel=0 DScale=1 PME-
Capabilities: [50] MSI: Enable- Count=1/1 Maskable+ 64bit+
Address: 0000000000000000 Data: 0000
Masking: 00000000 Pending: 00000000
Capabilities: [70] MSI-X: Enable+ Count=64 Masked-
Vector table: BAR=4 offset=00000000
PBA: BAR=4 offset=00002000That's good enough for me. I rest my case!
Quoted 25 lines
Quoted 15 lines
Quoted 15 lines
hw/block/nvme.c +2
-3* -* cmb_size_mb= and pmrdev= options are mutually exclusive due to limitation -* in available BAR's. cmb_size_mb= will take precedence over pmrdev= when -* both provided. +* pmrdev is assumed to be resident in BAR2/BAR3. When configured it consumes +* whole BAR2/BAR3 exclusively. * Enabling pmr emulation can be achieved by pointing to memory-backend-file. * For example: * -object memory-backend-file,id=<mem_id>,share=on,mem-path=<file_path>, \ @@ -69,18 +69,19 @@static void nvme_process_sq (void *opaque);Quoted 9 lines
hw/block/nvme.c line 75
Quoted 7 lines
hw/block/nvme.c +3
-3-static bool nvme_addr_is_cmb (NvmeCtrl *n,hwaddr addr) +static bool nvme_addr_is_cmb (NvmeCtrl *n,hwaddr addr,int size) { -hwaddr low = n->ctrl_mem.addr; -hwaddr hi = n->ctrl_mem.addr +int128_get64 (n->ctrl_mem.size); +hwaddr low = n->bar4.addr + n->cmb_offset; +hwaddr hi = low +NVME_CMBSZ_GETSIZE (n->bar.cmbsz);Isn't it better to use
n->bar4.size?
Quoted 3 lines
My understanding is that cmb doesn't necessarily need to occupy whole BAR, which is required to be power-of-two in size.
You are completely right of course.