The current logic for calculating 'maxdomain' making it a sum of
numa_state->num_nodes with spapr->gpu_numa_id. spapr->gpu_numa_id is
used as a index to determine the next available NUMA id that a
given NVGPU can use.
The problem is that the initial value of gpu_numa_id, for any topology
that has more than one NUMA node, is equal to numa_state->num_nodes.
This means that our maxdomain will always be, at least, twice the
amount of existing NUMA nodes. This means that a guest with 4 NUMA
nodes will end up with the following max-associativity-domains:
rtas/ibm,max-associativity-domains
00000004 00000008 00000008 00000008 00000008
This overtuning of maxdomains doesn't go unnoticed in the guest, being
detected in SLUB during boot:
dmesg | grep SLUB
[ 0.000000] SLUB: HWalign=128, Order=0-3, MinObjects=0, CPUs=4, Nodes=8
SLUB is detecting 8 total nodes, with 4 nodes being online.
This patch fixes ibm,max-associativity-domains by considering the amount
of NVGPUs NUMA nodes presented in the guest, instead of just
spapr->gpu_numa_id.
Reported-by: Cédric Le Goater <clg@kaod.org>
Tested-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20210128174213.1349181-4-danielhb413@gmail.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
We'll need to check the initial value given to spapr->gpu_numa_id when
building the rtas DT, so put it in a helper for easier access and to
avoid repetition.
Tested-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20210128174213.1349181-3-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This function is used only in spapr_numa.c.
Tested-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20210128174213.1349181-2-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
A new function called spapr_numa_define_associativity_domains()
is created to calculate the associativity domains and change
the associativity arrays considering user input. This is how
the associativity domain between two NUMA nodes A and B is
calculated:
- get the distance D between them
- get the correspondent NUMA level 'n_level' for D. This is done
via a helper called spapr_numa_get_numa_level()
- all associativity arrays were initialized with their own
numa_ids, and we're calculating the distance in node_id ascending
order, starting from node id 0 (the first node retrieved by
numa_state). This will have a cascade effect in the algorithm because
the associativity domains that node 0 defines will be carried over to
other nodes, and node 1 associativities will be carried over after
taking node 0 associativities into account, and so on. This
happens because we'll assign assoc_src as the associativity domain
of dst as well, for all NUMA levels beyond and including n_level.
The PPC kernel expects the associativity domains of the first node
(node id 0) to be always 0 [1], and this algorithm will grant that
by default.
Ultimately, all of this results in a best effort approximation for
the actual NUMA distances the user input in the command line. Given
the nature of how PAPR itself interprets NUMA distances versus the
expectations risen by how ACPI SLIT works, there might be better
algorithms but, in the end, it'll also result in another way to
approximate what the user really wanted.
To keep this commit message no longer than it already is, the next
patch will update the existing documentation in ppc-spapr-numa.rst
with more in depth details and design considerations/drawbacks.
[1] https://lore.kernel.org/linuxppc-dev/5e8fbea3-8faf-0951-172a-b41a2138fbcf@gmail.com/
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20201007172849.302240-5-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This is the first guest visible change introduced in
spapr_numa.c. The previous settings of both reference-points
and maxdomains were too restrictive, but enough for the
existing associativity we're setting in the resources.
We'll change that in the following patches, populating the
associativity arrays based on user input. For those changes
to be effective, reference-points and maxdomains must be
more flexible. After this patch, we'll have 4 distinct
levels of NUMA (0x4, 0x3, 0x2, 0x1) and maxdomains will
allow for any type of configuration the user intends to
do - under the scope and limitations of PAPR itself, of
course.
Reviewed-by: Greg Kurz <groug@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20201007172849.302240-4-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The pSeries machine does not support asymmetrical NUMA
configurations. This doesn't make much of a different
since we're not using user input for pSeries NUMA setup,
but this will change in the next patches.
To avoid breaking existing setups, gate this change by
checking for legacy NUMA support.
Reviewed-by: Greg Kurz <groug@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20201007172849.302240-3-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The current implementation of h_home_node_associativity hard codes
the values of associativity domains of the vcpus. Let's make
it consider the values already initialized in spapr->numa_assoc_array,
via the spapr_numa_get_vcpu_assoc() helper.
We want to set it and forget it, and for that we also need to
assert that we don't overflow the registers of the hypercall.
>From R4 to R9 we can squeeze in 12 associativity domains for
vcpus, so let's assert that VCPU_ASSOC_SIZE -1 isn't greater
than that.
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20200904172422.617460-4-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The work to be done in h_home_node_associativity() intersects
with what is already done in spapr_numa_fixup_cpu_dt(). This
patch creates a new helper, spapr_numa_get_vcpu_assoc(), to
be used for both spapr_numa_fixup_cpu_dt() and
h_home_node_associativity().
While we're at it, use memcpy() instead of loop assignment
to created the returned array.
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20200904172422.617460-3-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The implementation of this hypercall will be modified to use
spapr->numa_assoc_arrays input. Moving it to spapr_numa.c makes
make more sense.
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20200904172422.617460-2-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The NVLink2 GPUs works like a regular NUMA node with its
own associativity values, regardless of user input.
This can be handled inside spapr_numa_associativity_init(),
initializing NVGPU_MAX_NUM associativity arrays that can
be used by the GPUs.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20200903220639.563090-5-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
In a similar fashion as the previous patch, let's move the
handling of ibm,associativity-lookup-arrays from spapr.c to
spapr_numa.c. A spapr_numa_write_assoc_lookup_arrays() helper was
created, and spapr_dt_dynamic_reconfiguration_memory() can now
use it to advertise the lookup-arrays.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20200903220639.563090-4-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Vcpus have an additional paramenter to be appended, vcpu_id. This
also changes the size of the of property itself, which is being
represented in index 0 of numa_assoc_array[cpu->node_id],
and defaults to MAX_DISTANCE_REF_POINTS for all cases but
vcpus.
All this logic makes more sense in spapr_numa.c, where we handle
everything NUMA and associativity. A new helper spapr_numa_fixup_cpu_dt()
was added, and spapr.c uses it the same way as it was using the former
spapr_fixup_cpu_numa_dt().
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20200903220639.563090-3-danielhb413@gmail.com>
[dwg: Correct uint to int type, which can break windows builds]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The next step to centralize all NUMA/associativity handling in
the spapr machine is to create a 'one stop place' for all
things ibm,associativity.
This patch introduces numa_assoc_array, a 2 dimensional array
that will store all ibm,associativity arrays of all NUMA nodes.
This array is initialized in a new spapr_numa_associativity_init()
function, called in spapr_machine_init(). It is being initialized
with the same values used in other ibm,associativity properties
around spapr files (i.e. all zeros, last value is node_id).
The idea is to remove all hardcoded definitions and FDT writes
of ibm,associativity arrays, doing instead a call to the new
helper spapr_numa_write_associativity_dt() helper, that will
be able to write the DT with the correct values.
We'll start small, handling the trivial cases first. The
remaining instances of ibm,associativity will be handled
next.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20200903220639.563090-2-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
We're going to make changes in how spapr handles all
ibm,associativity* related properties to enhance our current NUMA
support.
At this moment we have associativity code scattered all around
spapr_* files, with hardcoded values and array sizes. This
makes it harder to change any NUMA specific parameters in
the future. Having everything in the same place allows not
only for easier tuning, but also easier understanding since all
NUMA related code is on the same file.
This patch introduces a new file to gather all NUMA/associativity
handling code in spapr, spapr_numa.c. To get things started, let's
remove associativity-reference-points and max-associativity-domains
code from spapr_dt_rtas() to a new helper called spapr_numa_write_rtas_dt().
This will decouple spapr_dt_rtas() from the NUMA changes that
are going to happen in those two properties.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20200901125645.118026-2-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
