Re: [PATCH v5 3/3] nvme: allow cmb and pmr to be enabled on same device

[Andrzej Jakowski <andrzej.jakowski@linux.intel.com>]

On 7/27/20 2:06 AM, Klaus Jensen wrote:
> On Jul 23 09:03, Andrzej Jakowski wrote:
>> 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 <andrzej.jakowski@linux.intel.com>
>> ---
>>  hw/block/nvme.c      | 120 +++++++++++++++++++++++++++++--------------
>>  hw/block/nvme.h      |   1 +
>>  include/block/nvme.h |   4 +-
>>  3 files changed, 85 insertions(+), 40 deletions(-)
>>
>> diff --git a/hw/block/nvme.c b/hw/block/nvme.c
>> index 43866b744f..d55a71a346 100644
>> --- a/hw/block/nvme.c
>> +++ b/hw/block/nvme.c
>> @@ -22,12 +22,13 @@
>>   *              [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 offset that is 2MiB aligned. When CMB is emulated
>> + * on Linux guest it is recommended to make cmb_size_mb multiple of 2. Both
>> + * size and alignment restrictions are imposed by Linux guest.
>>   *
>> - * 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>, \
>> @@ -57,8 +58,8 @@
>>  #define NVME_MAX_IOQPAIRS 0xffff
>>  #define NVME_DB_SIZE  4
>>  #define NVME_SPEC_VER 0x00010300
>> -#define NVME_CMB_BIR 2
>>  #define NVME_PMR_BIR 2
>> +#define NVME_MSIX_BIR 4
> 
> I think that either we keep the CMB constant (but updated with '4' of
> course) or we just get rid of both NVME_{CMB,MSIX}_BIR and use a literal
> '4' in nvme_bar4_init. It is very clear that is only BAR 4 we use.
> 
>>  #define NVME_TEMPERATURE 0x143
>>  #define NVME_TEMPERATURE_WARNING 0x157
>>  #define NVME_TEMPERATURE_CRITICAL 0x175
>> @@ -111,16 +112,18 @@ static uint16_t nvme_sqid(NvmeRequest *req)
>>  
>>  static bool nvme_addr_is_cmb(NvmeCtrl *n, hwaddr addr)
>>  {
>> -    hwaddr low = n->ctrl_mem.addr;
>> -    hwaddr hi  = n->ctrl_mem.addr + int128_get64(n->ctrl_mem.size);
>> +    hwaddr low = memory_region_to_absolute_addr(&n->ctrl_mem, 0);
>> +    hwaddr hi  = low + int128_get64(n->ctrl_mem.size);
> 
> Are we really really sure we want to use a global helper like this? What
> are the chances/risk that we ever introduce another overlay? I'd say
> zero. We are not even using a *real* overlay, it's just an io memory
> region (ctrl_mem) on top of a pure container (bar4). Can't we live with
> an internal helper doing `n->bar4.addr + n->ctrl_mem.addr` and be done
> with it? It also removes a data structure walk on each invocation of
> nvme_addr_is_cmb (which is done for **all** addresses in PRP lists and
> SGLs).

Thx!
My understanding of memory_region_absolute_addr()([1]) function is that it walks
memory hierarchy up to root while incrementing absolute addr. It is very 
similar to n->bar4.addr + n->ctrl_mem.addr approach with following 
differences:
 * n->bar4.addr + n->ctrl_mem.addr assumes single level hierarchy. Updates would
   be needed when another memory level is added.
 * memory_region_to_absolute_addr() works for any-level hierarchy at tradeoff
   of dereferencing data structure. 

I don't have data for likelihood of adding new memory level, nor how much more
memory_region_to_absolute_addr() vs n->bar4.addr + n->ctrl_mem.addr costs.

Please let me know which approach is preferred.

[1]
hwaddr memory_region_to_absolute_addr(MemoryRegion *mr, hwaddr offset)
{
    MemoryRegion *root;
    hwaddr abs_addr = offset;

    abs_addr += mr->addr;
    for (root = mr; root->container; ) {
        root = root->container;
        abs_addr += root->addr;
    }    

    return abs_addr;
}

> 
>>  
>>      return addr >= low && addr < hi;
>>  }
>>  
>>  static void nvme_addr_read(NvmeCtrl *n, hwaddr addr, void *buf, int size)
>>  {
>> +    hwaddr cmb_addr = memory_region_to_absolute_addr(&n->ctrl_mem, 0);
>> +
>>      if (n->bar.cmbsz && nvme_addr_is_cmb(n, addr)) {
>> -        memcpy(buf, (void *)&n->cmbuf[addr - n->ctrl_mem.addr], size);
>> +        memcpy(buf, (void *)&n->cmbuf[addr - cmb_addr], size);
>>          return;
>>      }
>>  
>> @@ -207,17 +210,18 @@ 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 = memory_region_to_absolute_addr(&n->ctrl_mem, 0);
>>      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->bar.cmbsz && prp1 >= cmb_addr &&
>> +               prp1 < cmb_addr + int128_get64(n->ctrl_mem.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);
>> @@ -262,7 +266,7 @@ 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++;
>> @@ -275,7 +279,7 @@ 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);
>>              }
>>          }
>>      }
>> @@ -1980,7 +1984,7 @@ 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);
>> @@ -2042,33 +2046,73 @@ 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)
>> +static void nvme_bar4_init(PCIDevice *pci_dev, Error **errp)
> 
> Since everything else is nvme_init_yadayada, can we swap for
> nvme_init_bar4?
> 
>>  {
>> -    NVME_CMBLOC_SET_BIR(n->bar.cmbloc, NVME_CMB_BIR);
>> -    NVME_CMBLOC_SET_OFST(n->bar.cmbloc, 0);
>> +    NvmeCtrl *n = NVME(pci_dev);
>> +    int status;
>> +    uint64_t bar_size, cmb_offset = 0;
>> +    uint32_t msix_vectors;
>> +    uint32_t nvme_pba_offset;
>> +    uint32_t cmb_size_units;
>> +
>> +    msix_vectors = n->params.msix_qsize;
>> +    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);
>> +        /* Linux guest requires it to be 2MiB aligned */
>> +        cmb_offset = QEMU_ALIGN_UP(bar_size, 2 * cmb_size_units);
>> +
> 
> This only makes sense because cmb_size_units is fixed at 2, so instead
> of relying on the value, can we just use a constant of `2 * MiB` to make
> it completely clear?
> 
>> +        NVME_CMBLOC_SET_BIR(n->bar.cmbloc, NVME_MSIX_BIR);
> 
> Use the NVME_CMB_BIR constant here (or a literal '4').
> 
>> +        NVME_CMBLOC_SET_OFST(n->bar.cmbloc, cmb_offset / cmb_size_units);
>> +
>> +        n->cmbuf = g_malloc0(NVME_CMBSZ_GETSIZE(n->bar.cmbsz));
>> +
>> +        bar_size += cmb_offset;
>> +        bar_size += NVME_CMBSZ_GETSIZE(n->bar.cmbsz);
>> +    }
>> +
>> +    bar_size = pow2ceil(bar_size);
>> +
>> +    /*
>> +     * Create memory region for BAR4, then overlap cmb, msix and pba
>> +     * tables on top of it.
>> +     */
>> +    memory_region_init(&n->bar4, OBJECT(n), "nvme-bar4", bar_size);
>> +
>> +    if (n->params.cmb_size_mb) {
>> +        memory_region_init_io(&n->ctrl_mem, OBJECT(n), &nvme_cmb_ops, n,
>> +                              "nvme-cmb", NVME_CMBSZ_GETSIZE(n->bar.cmbsz));
>>  
>> -    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);
>> +        memory_region_add_subregion(&n->bar4, cmb_offset, &n->ctrl_mem);
>> +    }
>> +
>> +    status = msix_init(pci_dev, n->params.msix_qsize,
>> +                       &n->bar4, NVME_MSIX_BIR, 0,
>> +                       &n->bar4, NVME_MSIX_BIR, nvme_pba_offset,
>> +                       0, errp);
>> +
>> +    if (status) {
>> +        return;
>> +    }
>>  
>> -    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),
>> +    pci_register_bar(pci_dev, NVME_MSIX_BIR,
> 
> Since the bar has both MSIX and CMB, I'd rather use a literal '4' here
> (or a local constant).
> 
>>                       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);
>> @@ -2126,13 +2170,10 @@ 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);
>>      }
>>  }
>> @@ -2199,6 +2240,7 @@ 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 = NVME_SPEC_VER;
>>      n->bar.intmc = n->bar.intms = 0;
>> diff --git a/hw/block/nvme.h b/hw/block/nvme.h
>> index 0b6a8ae665..f291395cd0 100644
>> --- a/hw/block/nvme.h
>> +++ b/hw/block/nvme.h
>> @@ -91,6 +91,7 @@ typedef struct NvmeCtrl {
>>      PCIDevice    parent_obj;
>>      MemoryRegion iomem;
>>      MemoryRegion ctrl_mem;
>> +    MemoryRegion bar4;
>>      NvmeBar      bar;
>>      BlockConf    conf;
>>      NvmeParams   params;
>> diff --git a/include/block/nvme.h b/include/block/nvme.h
>> index d641ca6649..77b59d18dd 100644
>> --- a/include/block/nvme.h
>> +++ b/include/block/nvme.h
>> @@ -217,9 +217,11 @@ 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,
>> -- 
>> 2.21.1
>>
>>