On Mon, Nov 9, 2020 at 4:48 PM Anatoly Burakov
<anatoly.bura...@intel.com> wrote:
>
> Any EAL memory allocation often goes through eal_get_virtual_area()
> function, which will print a warning whenever the resulting allocation
> didn't match the specified address requirements. This is useful for
> when we have requested a specific base virtual address, to let the user
> know that the mapping has deviated from that address.
>
> However, on Linux, we also have a default base address that's there to
> ensure better chances of successful secondary process initialization,
> as well as higher likelihood of the virtual areas to fit inside the
> IOMMU address width. Because of this default base address, there are
> warnings printed even when no base address was explicitly requested,
> which can be confusing to the user.
>
> Emit this warning with debug level unless base address was explicitly
> requested by the user.
>
> Cc: Damjan Marion <damar...@cisco.com>
>
> Signed-off-by: Anatoly Burakov <anatoly.bura...@intel.com>
> ---
>
> Notes:
> v2:
> - Fix the condition to not update the address incorrectly
> - Instead of removing the warning, let it have debug level unless base
> address
> was explicitly specified by the user
>
> I'm not entirely sure the trade off between user confusion and helpful
> debug
> information is worth it, but in my experience, i've stopped getting any
> emails
> about secondary processes a long time ago and this isn't a widely used
> feature,
> so i believe this is worth it.
>
> lib/librte_eal/common/eal_common_memory.c | 16 +++++++++++++---
> 1 file changed, 13 insertions(+), 3 deletions(-)
>
> diff --git a/lib/librte_eal/common/eal_common_memory.c
> b/lib/librte_eal/common/eal_common_memory.c
> index 33917fa835..1b50c2099d 100644
> --- a/lib/librte_eal/common/eal_common_memory.c
> +++ b/lib/librte_eal/common/eal_common_memory.c
> @@ -140,9 +140,19 @@ eal_get_virtual_area(void *requested_addr, size_t *size,
> return NULL;
> } else if (requested_addr != NULL && addr_is_hint &&
> aligned_addr != requested_addr) {
> - RTE_LOG(WARNING, EAL, "WARNING! Base virtual address hint (%p
> != %p) not respected!\n",
> - requested_addr, aligned_addr);
> - RTE_LOG(WARNING, EAL, " This may cause issues with mapping
> memory into secondary processes\n");
> + /*
> + * demote this warning to debug if we did not explicitly
> request
> + * a base virtual address.
> + */
> + if (internal_conf->base_virtaddr != 0) {
> + RTE_LOG(WARNING, EAL, "WARNING! Base virtual address
> hint (%p != %p) not respected!\n",
> + requested_addr, aligned_addr);
> + RTE_LOG(WARNING, EAL, " This may cause issues with
> mapping memory into secondary processes\n");
> + } else {
> + RTE_LOG(DEBUG, EAL, "WARNING! Base virtual address
> hint (%p != %p) not respected!\n",
> + requested_addr, aligned_addr);
> + RTE_LOG(DEBUG, EAL, " This may cause issues with
> mapping memory into secondary processes\n");
> + }
> } else if (next_baseaddr != NULL) {
> next_baseaddr = RTE_PTR_ADD(aligned_addr, *size);
> }
> --
> 2.17.1
>
EAL options like --in-memory or --no-shconf makes MP unusable.
If we add a rte_mp_disable() for them, we could check here for MP
status here and display nothing at all.
WDYT?
--
David Marchand
