Re: [PATCH v2 06/11] util/vfio-helpers: Pass Error handle to qemu_vfio_dma_map()

["Philippe Mathieu-Daudé" <philmd@redhat.com>]

+Markus

On 8/27/21 7:48 AM, Klaus Jensen wrote:
> On Aug 26 21:50, Philippe Mathieu-Daudé wrote:
>> Currently qemu_vfio_dma_map() displays errors on stderr.
>> When using management interface, this information is simply
>> lost. Pass qemu_vfio_dma_map() an Error** handle so it can
>> propagate the error to callers.
>>
>> Reviewed-by: Fam Zheng <fam@euphon.net>
>> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
>> Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
>> ---
>>  include/qemu/vfio-helpers.h |  2 +-
>>  block/nvme.c                | 22 +++++++++++-----------
>>  util/vfio-helpers.c         | 10 ++++++----
>>  3 files changed, 18 insertions(+), 16 deletions(-)
>>
>> diff --git a/include/qemu/vfio-helpers.h b/include/qemu/vfio-helpers.h
>> index 4491c8e1a6e..bde9495b254 100644
>> --- a/include/qemu/vfio-helpers.h
>> +++ b/include/qemu/vfio-helpers.h
>> @@ -18,7 +18,7 @@ typedef struct QEMUVFIOState QEMUVFIOState;
>>  QEMUVFIOState *qemu_vfio_open_pci(const char *device, Error **errp);
>>  void qemu_vfio_close(QEMUVFIOState *s);
>>  int qemu_vfio_dma_map(QEMUVFIOState *s, void *host, size_t size,
>> -                      bool temporary, uint64_t *iova_list);
>> +                      bool temporary, uint64_t *iova_list, Error **errp);
>>  int qemu_vfio_dma_reset_temporary(QEMUVFIOState *s);
>>  void qemu_vfio_dma_unmap(QEMUVFIOState *s, void *host);
>>  void *qemu_vfio_pci_map_bar(QEMUVFIOState *s, int index,
>> diff --git a/block/nvme.c b/block/nvme.c
>> index 0786c501e46..80546b0babd 100644
>> --- a/block/nvme.c
>> +++ b/block/nvme.c
>> @@ -176,12 +176,11 @@ static bool nvme_init_queue(BDRVNVMeState *s, NVMeQueue *q,
>>          return false;
>>      }
>>      memset(q->queue, 0, bytes);
>> -    r = qemu_vfio_dma_map(s->vfio, q->queue, bytes, false, &q->iova);
>> +    r = qemu_vfio_dma_map(s->vfio, q->queue, bytes, false, &q->iova, errp);
>>      if (r) {
>> -        error_setg(errp, "Cannot map queue");
>> -        return false;
>> +        error_prepend(errp, "Cannot map queue: ");
>>      }
>> -    return true;
>> +    return r == 0;
>>  }
>>  
>>  static void nvme_free_queue_pair(NVMeQueuePair *q)
>> @@ -239,9 +238,9 @@ static NVMeQueuePair *nvme_create_queue_pair(BDRVNVMeState *s,
>>      qemu_co_queue_init(&q->free_req_queue);
>>      q->completion_bh = aio_bh_new(aio_context, nvme_process_completion_bh, q);
>>      r = qemu_vfio_dma_map(s->vfio, q->prp_list_pages, bytes,
>> -                          false, &prp_list_iova);
>> +                          false, &prp_list_iova, errp);
>>      if (r) {
>> -        error_setg_errno(errp, -r, "Cannot map buffer for DMA");
>> +        error_prepend(errp, "Cannot map buffer for DMA: ");
>>          goto fail;
>>      }
>>      q->free_req_head = -1;
>> @@ -534,9 +533,9 @@ static bool nvme_identify(BlockDriverState *bs, int namespace, Error **errp)
>>          error_setg(errp, "Cannot allocate buffer for identify response");
>>          goto out;
>>      }
>> -    r = qemu_vfio_dma_map(s->vfio, id, id_size, true, &iova);
>> +    r = qemu_vfio_dma_map(s->vfio, id, id_size, true, &iova, errp);
>>      if (r) {
>> -        error_setg(errp, "Cannot map buffer for DMA");
>> +        error_prepend(errp, "Cannot map buffer for DMA: ");
>>          goto out;
>>      }
>>  
>> @@ -1032,7 +1031,7 @@ static coroutine_fn int nvme_cmd_map_qiov(BlockDriverState *bs, NvmeCmd *cmd,
>>  try_map:
>>          r = qemu_vfio_dma_map(s->vfio,
>>                                qiov->iov[i].iov_base,
>> -                              len, true, &iova);
>> +                              len, true, &iova, NULL);
>>          if (r == -ENOSPC) {
>>              /*
>>               * In addition to the -ENOMEM error, the VFIO_IOMMU_MAP_DMA
>> @@ -1524,14 +1523,15 @@ static void nvme_aio_unplug(BlockDriverState *bs)
>>  static void nvme_register_buf(BlockDriverState *bs, void *host, size_t size)
>>  {
>>      int ret;
>> +    Error *local_err = NULL;
>>      BDRVNVMeState *s = bs->opaque;
>>  
>> -    ret = qemu_vfio_dma_map(s->vfio, host, size, false, NULL);
>> +    ret = qemu_vfio_dma_map(s->vfio, host, size, false, NULL, &local_err);
>>      if (ret) {
>>          /* FIXME: we may run out of IOVA addresses after repeated
>>           * bdrv_register_buf/bdrv_unregister_buf, because nvme_vfio_dma_unmap
>>           * doesn't reclaim addresses for fixed mappings. */
>> -        error_report("nvme_register_buf failed: %s", strerror(-ret));
>> +        error_reportf_err(local_err, "nvme_register_buf failed: ");
>>      }
>>  }
>>  
>> diff --git a/util/vfio-helpers.c b/util/vfio-helpers.c
>> index e7909222cfd..77cdec845d9 100644
>> --- a/util/vfio-helpers.c
>> +++ b/util/vfio-helpers.c
>> @@ -463,13 +463,15 @@ static void qemu_vfio_ram_block_added(RAMBlockNotifier *n, void *host,
>>                                        size_t size, size_t max_size)
>>  {
>>      QEMUVFIOState *s = container_of(n, QEMUVFIOState, ram_notifier);
>> +    Error *local_err = NULL;
>>      int ret;
>>  
>>      trace_qemu_vfio_ram_block_added(s, host, max_size);
>> -    ret = qemu_vfio_dma_map(s, host, max_size, false, NULL);
>> +    ret = qemu_vfio_dma_map(s, host, max_size, false, NULL, &local_err);
>>      if (ret) {
>> -        error_report("qemu_vfio_dma_map(%p, %zu) failed: %s", host, max_size,
>> -                     strerror(-ret));
>> +        error_reportf_err(local_err,
>> +                          "qemu_vfio_dma_map(%p, %zu) failed: ",
>> +                          host, max_size);
>>      }
>>  }
>>  
>> @@ -725,7 +727,7 @@ qemu_vfio_find_temp_iova(QEMUVFIOState *s, size_t size, uint64_t *iova)
>>   * mapping status within this area is not allowed).
>>   */
>>  int qemu_vfio_dma_map(QEMUVFIOState *s, void *host, size_t size,
>> -                      bool temporary, uint64_t *iova)
>> +                      bool temporary, uint64_t *iova, Error **errp)
> 
> Won't this break at this point? You add an Error** to qemu_vfio_dma_map,
> but it is never used (so it probably stays NULL). So, above, if
> qemu_vfio_dma_map() returns something not zero, you use error_prepend(),
> which I think expects errp to be non-NULL?

I can't find such restriction documented in "qapi/error.h".

In error.c:

void error_prepend(Error *const *errp, const char *fmt, ...)
{
    va_list ap;

    va_start(ap, fmt);
    error_vprepend(errp, fmt, ap);
    ...

void error_vprepend(Error *const *errp, const char *fmt, va_list ap)
{
    GString *newmsg;

    if (!errp) {
        return;
    }
    ...

> 
> I realize that later patches fixes this, but this sticks out when
> reviewing this on its own.
> 
>>  {
>>      int ret = 0;
>>      int index;
>> -- 
>> 2.31.1
>>
>>
>