cpu: reset byte_enable across writeMem calls

data_write_req byteEnable which is used in ARM SVE partial writes was not
being zeroed between writes.

As a result, non-SVE memory write instructions such as STP that followed
SVE memory write instructions could still have the write mask active.

This could lead to wrong simulation behaviour, and to an assertion failure:

src/mem/packet.hh:1211: void Packet::writeData(uint8_t*) const: Assertion
`req->getByteEnable().size() == getSize()' failed. '`

Change-Id: I74b5a82675e9923b0ffdf2c1dd9afb00c91cb204
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20448
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev: Templatize PioPort.

When creating a base class which needs to be a SimObject, it's
necessary to decide ahead of time whether to use PioDevice or
BasicPioDevice in the hierarchy because they inherit from SimObject. If
they were added into the hierarchy later, then the original class would
inherit from SimObject, as would PioDevice. That would create a diamond
inheritance structure which would require virtual inheritance, and
that's a can of worms we'd rather not get into.

A big part of the PioPort mechanism is the PioPort itself which holds
a pointer to its parent device and delegates reads/writes to it. It
does that with a PioDevice pointer, and PioDevice declares virtual
functions for all the callbacks the port can call into.

Instead of that, this change templatizes PioPort based on the class of
the device that holds it. That will let you use a PioPort on *any*
class, as long as it has the methods PioPort depends on. That removes
the need to create an inheritance diamond to add a PioPort down the
line since PioDevice is no longer strictly required.

The PioDevice and BasicPioDevice classes are still around since they
still provide some additional functionality and there are existing
classes which depend on them.

Change-Id: I753afc1e0fa54b91217d54c1f8743c150537e960
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20568
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

ruby: Fix the way stall map size is checked for availability

To ensure that enqueuer observes the practical availability. We
check the message buffer queue size at the start of the cycle.
We also add the size of the stall queue to consider the total
queue size. However, messages can be moved from regular queue
to stall map. This leads to messages being considered twice leading
to false flow control. This patch fixes it by storing the stall map
size at the beginning of the cycle and considering it for checking
availability.

Change-Id: I6ea94f34fe5279b91f74e106d43263e55ec4bf06
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20389
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>

configs: Fix replacement policy assignment

Commit d207e9ccee411877fdeac80bb68a27900560f50f reworked the tags
to split the replacement policies, however the name of the variable
that contains the replacement policy changed between patch revisions,
which was not updated accordingly in the configs files.

Change-Id: I2072529e2c7d54197c371bcaa323bfd9f34ec3ba
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20548
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

stats: Create HDF5 stat files relative to simout

When using the HDF5 stats backend, we currently pass the filename from
the command line straight to the HDF5 library. This behaviour is
different from the expected behaviour when using plain text stat files
where relative paths are resolved relative to simout. This change adds
support for resolving hdf5 paths relative to simout.

Change-Id: I09cb0b7468735e697125eda7f04c5a85725ca8d0
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20508
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>

stats: Catch exceptions by const reference

Catching exceptions by value causes -Werror=catch-value= to be
triggered during a build process. This change replaces the values
with const references in the catch blocks.

Change-Id: Iddabb595c5916d8120dcdb08e2c2f36ff21c3c9b
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20528
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

stats: Add support for listing available formats

Add a command line option to list available stat formats and their
documentation.

Change-Id: I7f5f2272d9b0176639f59f2efedb9cab2f7da5b9
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19670
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arm,kvm: Fix python imports from global namespace

Change-Id: I31bd3563c2427efd7e520f714b1ca6f480fa4e85
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Ciro Santilli <ciro.santilli@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20491
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Use SatCounter for RRPV

Use SatCounter in RRIP's RRPV. As such, move validation functionality
to a proper variable.

Change-Id: I142db2b7f6cd518ac3a2b68c9ed48005402b3464
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20452
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

base: Add function to saturate SatCounter

Create a saturation function for the SatCounter that makes its
internal counter reach its maximum value.

Change-Id: Ibd30fa27c4ae2714dd48e3eba85addd035fb737c
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20451
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

stats: Add beta support for HDF5 stat dumps

This changeset add support for stat dumps in the HDF5 file
format. HDF5 is a binary data format that represents data in a
file-system-like balanced tree. It has native support for
N-dimensional arrays and binary data (e.g., frame buffers).

It has the following benefits over traditional text stat files:

  * Efficient storage of time series (multiple stat dumps)

  * Fast lookup of stats

  * Plenty of existing tooling (e.g., Python libraries and graphical
    viewers)

  * File format can be used to store frame buffers together with
    normal stats.

Drawbacks:

  * Large startup cost (single stat dump larger than text equivalent)

  * Stat dumps are slower than text

Known limitations:

  * Distributions and histograms aren't supported.

HDF5 stat output can be enabled using the 'h5' URL scheme when
overriding the stat file name on gem5's command line. The following
parameters are supported:

  * chunking (unsigned): Number of time steps to pre-allocate
    (default: 10)

  * desc (bool): Output stat descriptions (default: True)

  * formulas (bool): Output derived stats (default: True)

Example gem5 command line:

./build/ARM/gem5.opt \
  --stats-file="h5://stats.h5?desc=False;formulas=False" \
  configs/example/fs.py

Example Python stat consumer that computes IPC:
  import h5py

  f = h5py.File('stats.h5', 'r')
  group = f['/system/cpu']
  for i, c in zip(group['committedInsts'], group['numCycles']):
      print i, c, i / c

Change-Id: I351c6cbff2fb7bef9012f47876ba227ed288975b
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/8121
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Ciro Santilli <ciro.santilli@arm.com>

mem: Convert CommMonitor to the new stat framework

Change-Id: I851c29909f3e6923c0233505a4d0f2d266bc254f
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19371
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>

cpu: Convert traffic gen to use new stats

Change-Id: Ife690a137c2dcfb6bcc8b22df996c84f0d231618
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19370
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>

stats: Add support for partial stat dumps

Add support for partial stat dumps by passing an optional 'root'
keyword argument to m5.stats.dump(). Specifying root slightly changes
the semantics of the dump command. For legacy reasons, gem5 only
allows one stat dump per tick. This is likely a limitation introduced
as a hack to prevent automatic dumping at the end of simulation from
interfering with explicit dumping from a simulation script. This
restriction does not apply when specifying a root. However, these stat
dumps will still prevent an additional stat dump in the same tick with
an unspecified root.

N.B.: This new API /only/ works for new-style stats that have an
explicit hierarchy. Legacy stats will not be dumped if a root is
specified.

Change-Id: Idc8ff448b9f70a796427b4a5231e7371485130b4
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19369
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

stats: Add support for hierarchical stats

This change makes the stat system aware of the hierarchical nature of
stats. The aim is to achieve the following goals:

  * Make the SimObject hierarchy explicit in the stat system (i.e.,
    get rid of name() + ".foo"). This makes stat naming less fragile
    and makes it possible to implement hierarchical formats like
    XML/HDF5/JSON in a clean way.

  * Make it more convenient to split stats into a separate
    struct/class that can be bound to a SimObject. This makes the
    namespace cleaner and makes stat accesses a bit more obvious.

  * Make it possible to build groups of stats in C++ that can be used
    in subcomponents in a SimObject (similar to what we do for
    checkpoint sections). This makes it easier to structure large
    components.

  * Enable partial stat dumps. Some of our internal users have been
    asking for this since a full stat dump can be large.

  * Enable better stat access from Python.

This changeset implements solves the first three points by introducing
a class (Stats::Group) that owns statistics belonging to the same
object. SimObjects inherit from Stats::Group since they typically have
statistics.

New-style statistics need to be associated with a parent group at
instantiation time. Instantiation typically sets the name and the
description, other parameters need to be set by overriding
Group::regStats() just like with legacy stats. Simple objects with
scalar stats can typically avoid implementing regStats() altogether
since the stat name and description are both specified in the
constructor.

For convenience reasons, statistics groups can be merged into other
groups. This means that a SimObject can create a stat struct that
inherits from Stats::Group and merge it into the parent group
(SimObject). This can make the code cleaner since statistics tracking
gets grouped into a single object.

Stat visitors have a new API to expose the group structure. The
Output::beginGroup(name) method is called at the beginning of a group
and the Output::endGroup() method is called when all stats, and
sub-groups, have been visited. Flat formats (e.g., the text format)
typically need to maintain a stack to track the full path to a stat.

Legacy, flat, statistics are still supported after applying this
change. These stats don't belong to any group and stat visitors will
not see a Output::beginGroup(name) call before their corresponding
Output::visit() methods are called.

Change-Id: I9025d61dfadeabcc8ecf30813ab2060def455648
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19368
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>

mem-ruby: Define BloomFilter namespace

Define a BloomFilter namespace and put all BloomFilter related
code in it.

As a side effect the BloomFilter classes have been renamed to
remove the "BloomFilter" suffix.

Change-Id: I3ee8cc225bf3b820e561c3e25a6bf38e0012e3a8
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18874
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-ruby: Make H3 inherit from MultiBitSelBloomFilter

Make MultiBitSelBloomFilter a generic BloomFilter that maps
multiple entries to an address, and therefore uses multiple
hash functions. This allows the common functionality of both
filters to be merged into one, since they only differ in the
hash functions being used.

Change-Id: I0984067b710a208715f5f2727b8c4312feb6529b
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18873
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-ruby: Finish implementing BloomFilter merge

Not all Bloom Filters had their union functionality implemented.
This change adds them.

Change-Id: I86af18d3c5eabd0da8280b57a88789b3af803c04
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18872
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>

mem-ruby: Remove NonCountingBloomFilter

Make BlockBloomFilter accept having a single bitfield, in which
case it behaves exactly as the NonCountingBloomFilter, and thus
the latter can be removed.

Change-Id: I56d96a89290c933293ce434bbe0e8bcd4bbcaa42
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18871
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-ruby: Make MultiGrainBloomFilter generic

Allow combining any number of Bloom Filters in the MultiGrain.

Change-Id: I73ae33063e1feed731af6f625d2f64245f21df18
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18869
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-ruby: Parameterize xor bits in BlockBloomFilter

Parameterize bitfield ranges in BlockBloomFilter such that the
hash is applied between masked bitfields of an address.

Change-Id: I008bd873458e9815e98530e308491adb65bb34cb
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18870
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>

cpu: Make get(Data|Inst)Port return a Port and not a MasterPort.

No caller uses any of the MasterPort specific properties of these
function's return values, so we can instead return a reference to the
base Port class. This makes it possible for the data and inst ports
to be of any port type, not just gem5 style MasterPorts. This makes
life simpler for, for example, systemc based CPUs which might have TLM
ports.

It also makes it possible for any two CPUs which have compatible ports
to be switched between, as long as the ports they use support being
unbound. Unfortunately that does not include TLM or systemc ports which
are bound permanently.

Change-Id: I98fce5a16d2ef1af051238e929dd96d57a4ac838
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20240
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Gabe Black <gabeblack@google.com>

cpu, mem: Add new getSendFunctional method to the base CPU.

This returns a sendFunctional delegate references which can be used
to send functional accesses directly, or more likely when constructing
a PortProxy subclass. In those cases only the functional capabilities
of those ports are needed so there's no reason to require a full port
which supports all three protocols. Also, this removes the last
remaining use of get(Data|Inst)Port which relies on those returning
a port which supports the gem5 protocols, except the default
implementations of this new function. If a CPU doesn't have
traditional gem5 style ports, it can override this function to
do whatever other behavior is necessary and return its real ports
through get(Data|Inst)Port.

Change-Id: Ide4da81e3bc679662cd85902ba6bd537cce54a53
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20237
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>

mem: Make PortProxy use a delegate for a sendFunctional function.

The only part of the MaserPort the PortProxy uses is the sendFunctional
function which is part of the functional protocol. Rather than require
a MasterPort which comes along with a lot of other mechanisms, this
change slightly adjusts the PortProxy to only require that function
through the use of a delegate. That allows lots of flexibility in how
the actual packet gets sent and what sends it.

In cases where code constructs a PortProxy and passes its constructor
an unbound MasterPort, the PortProxy will create a delegate to the
sendFunctional method on its own.

This should also make it easier for objects which don't have
traditional gem5 style ports, for instance systemc models, to implement
just the little bit of the protocol they need, rather than having to
stub out a whole port class, most of which will be ignored.

Change-Id: I234b42ce050f12313b551a61736186ddf2c9e2c7
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20229
Maintainer: Gabe Black <gabeblack@google.com>
Reviewed-by: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: Move the instruction port into o3's fetch stage.

That's where it's used, and that avoids having to pass it around using
the top level getInstPort accessor.

Change-Id: I489a3f3239b3116292f3dcd78a3945fb468c6311
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20239
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>

cpu: Move O3's data port into the LSQ.

That's where it's used, and putting it there avoids having to pass
around the port using the top level getDataPort function.

Change-Id: I0dea25d0c5f4bb3f58a6574a8f2b2d242784caf2
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20238
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>

mem: Eliminate the Base(Slave|Master)Port classes.

The Port class has assumed all the duties of the less generic
Base*Port classes, making them unnecessary. Since they don't add
anything but make the code more complex, this change eliminates them.

Change-Id: Ibb9c56def04465f353362595c1f1c5ac5083e5e9
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20236
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>

cpu, dev, mem: Use the new Port methods.

Use getPeer, takeOverFrom, and << to simplify the use of ports in some
areas.

Change-Id: Idfbda27411b5d6b742f5e4927894302ea6d6a53d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20235
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>

sim: Add a << overload for the Port class which prints its name.

This makes it easier/less verbose to print the name of a port, it's
most important and identifying feature, in a DPRINTF or other stream
based output.

Change-Id: I050d102844612577f9a83d550e619736507a6781
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20234
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>

sim: Add a takeOverFrom method to the base Port class.

This makes it easier, safer, and less verbose to swap out ports when
a CPU takes over for another CPU, for instance.

Change-Id: Ice08e4dcb4b04dc66b1841331092a78b4f6f5a96
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20233
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>

mem, sim, systemc: Reorganize Port and co.s bind, unbind slightly.

The base Port class can keep track of its peer, and also whether it's
connected. This is partially delegated away from the port subclasses
which still keep track of a cast version of their peer pointer for
their own conveneince, so that it can be used by generic code. Even
with the Port mechanism's new flexibility, each port still has
exactly one peer and is either connected or not based on whether there
is a peer currently.

Change-Id: Id3228617dd1604d196814254a1aadeac5ade7cde
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20232
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>

dev-arm: Fix GICv3 ITS indexing error

Table walks were not considering the entry size when evaluating
the address.

Change-Id: Ica6bf6d88632985ee8ed120448b32e0f7e918a8a
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20329
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev-arm: Fix GITS_BASER initialization/access

The patch is fixing/improving GITS_BASER registers initialization.

* Not using reserved table types anymore (GITS_BASER.TYPE)
* Using write mask for handling WI bits

Change-Id: Ibe24667fdf22b42b86496167c19fc00bbb0ba191
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20328
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-riscv: fix GDB register cache

Fixes the definition of the RISC-V GDB register cache.  The latest
version, of RISC-V gdb, commit c3eb4078520dad8234ffd7fbf893ac0da23ad3c8,
appears to only accept the 32 integer registers + the PC in the 'g'
packet.

This functions with the Linux toolchain (riscv64-unknown-linux-gnu-*),
but works best with the Newlib toolchain (riscv64-unknown-elf-*).

Change-Id: Ie35ea89a45870fb634e6c68236261bde27c86e41
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20028
Reviewed-by: Alec Roelke <alec.roelke@gmail.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Alec Roelke <alec.roelke@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem: Put gem5 protocols in their own directory.

This reduces clutter in the src/mem directory, and makes it clear that
those protocols are for the classic gem5 memory system, not ruby, TLM,
etc.

Change-Id: I6cf6b21134d82f4f01991e4fe92dbea8c7e82081
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20231
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>

mem: Move ruby protocols into a directory called ruby_protocol.

Now that the gem5 protocols are split out, it would be nice to put them
in their own protocol directory. It's also confusing to have files
called *_protocol which are not in the protocol directory.

Change-Id: I7475ee111630050a2421816dfd290921baab9f71
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20230
Reviewed-by: Gabe Black <gabeblack@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem: Split the various protocols out of the gem5 master/slave ports.

This makes the protocols easier to see in their entirity, and makes it
easier to add a new type of port which only supports the functional
protocol.

Change-Id: If5d639bef45062f0a23af2ac46f50933e6a8f144
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20228
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>

mem-ruby: fix build with PROTOCOL=MOESI_hammer

Was failing with:

Error: Unrecognized variable: l1i_victim_addr

since: I2c43f22aba5af3a57e54b1c435e5d3fbba86d1d5

Change-Id: I7df666acb724ee541804dd7557753a9ba4005516
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20261
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Reviewed-by: Pouya Fotouhi <pfotouhi@ucdavis.edu>
Maintainer: Anthony Gutierrez <anthony.gutierrez@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev-arm: Start using GITS_CTLR.quiescent bit

The GITS_CTLR.quiescent bit is used by priviledged sw to check when the
ITS has finished draining its state (all pending translations/table
walks have ended) once it has been disabled (by setting the
GITS_CTLR.enable bit to 0).
This patch is modelling this behaviour by

* Changing the reset state to enable=0, quiescent=1
* Making the GITS_CTLR.quiescent bit RO
* Updating the bit once a new translation/command is being processed
(quiescent=0) and when there are no pending translation/commands
(quiescent=1)

Change-Id: I7cfe94b25d603400364b1cdfc2d2397acf5dfad8
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20257
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev-arm: Allow 32 bit accesses to GITS_C(WRITER/READR/BASER)

For those registers (GITS_CWRITER, GITS_READR and GITS_CBASER)
Bits [63:32] and bits [31:0] are accessible separately.

Change-Id: Ibf60b5e4fd20efb21a63570e6012862e37946877
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20256
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev-arm,system-arm: missing GICv3 ranges property

This patch adds the device tree "ranges" property to GICv3 for
the VExpress_GEM5_V2 platform. It is also included in the GICv3 DTB
auto generation.
This allows the GICv3 ITS to be specified in the device tree.

Change-Id: I00e1bb0fd45521e34820c0a23ddf047afec7aa4c
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20255
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

configs: root, platform options in fs bigLITTLE

(1) Two new options are added to fs_bigLITTLE.py:
    - "root": disk/partition containing the rootfs (def. "/dev/vda1")
    - "machine-type": hardware platform class (def. "VExpress_GEM5_V1")
        + Accepts platform classes from PlatformConfig
(2) Default kernel is not available in public uploads, force the user
    to provide its own kernel instead of crashing.

Change-Id: I88283ae12cd7289e15b9277ea2cc382e9136f11c
Reviewed-by: Ciro Santilli <ciro.santilli@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20148
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-riscv: Update register file

This patch adds mcounteren, scounteren according to Risc-V
Privileged Architectures V1.10.

Change-Id: I6e138a50710bc0a1e9d9c38a11fc7fcc09ed500e
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20128
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Alec Roelke <alec.roelke@gmail.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Alec Roelke <alec.roelke@gmail.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm, cpu: fix ARM ubsan build on GCC 7.4.0

In src/cpu/reg_class.hh, numPinnedWrites was unset because the
constructors were not well factored out.

Change-Id: Ib2fc8d34a1adf5c48826d257a31dd24dfa64a08a
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20048
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

base: assert that stats bucket size is greater than 0

Change-Id: I9a2264e9ee74acea1f09f01ea8f5004567ba1ea8
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20049
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Maintainer: Anthony Gutierrez <anthony.gutierrez@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: Fix implicit fallthrough build errors

1942b21713 introduced implicit-fallthrough errors when compiled with
GCC 8.  This change adds M5_UNREACHABLE in the default case.

Change-Id: I220f2b3fe39b5c3a65c0dd390915bffeafb28962
Signed-off-by: Chun-Chen TK Hsu <chunchenhsu@google.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20268
Reviewed-by: Jordi Vaquero <jordi.vaquero@metempsy.com>
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev-arm: Add redistributor-stride property to GICv3

This is needed since by default the model is assuming a GICv4
memory layout.

Change-Id: Ic64e6a488cc1a43a56ce28f6d11b8868df102aa0
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Ciro Santilli <ciro.santilli@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20248
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: Replace occ of opModeToEL(currOpMode/cpsr) with currEL

Change-Id: I739a9be03ea5caa63540c62fd110eee86a058c4c
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20252
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: Replace direct use cpsr.el with currEL helper

The patch is replacing it in places where the current EL could be using
AArch32, hence leading to an incorrect ExceptionLevel.

Change-Id: I99b75af2668f2c38fd88bec62e985ab7dbea80dc
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20251
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: Overload currEL helper with CPSR argument

Change-Id: I1edabc61637ecb9d30bca34b5dbcf1de12b35fe0
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20250
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: Rewrite the currEL helper method to use opModeToEL

Direct use of cpsr.el should be discouraged: it should be used when
in AArch64 only; when in AArch32 it won't return the matching EL.

Eg: when in Supervisor Mode (EL1), CPSR.M<3,0> (mode) is 0b0011,
and cpsr.el will return 0 (EL0)

Change-Id: I5504bd1f59980f79b2607cce435ea09245de12e5
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20249
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev-arm: Add GITS_PIDR2 register to the ITS memory map

The GITS Peripheral Identification Register #2 bits assignments are the
same as those for GICD_PIDR2.

Change-Id: I235008a383e08dd557d899cb3aa18202ef943f8b
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20254
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev-arm: Add Gicv3Distributor members for GICv3 GICD_PIDRx

There is no need of calculating the value every time the registers
are read.

Change-Id: I58b87abb585fb9928959992927f00d9c000a4c35
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20253
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-ruby, arch-hsail: Removed hit latency from VIPERCoalescer

Removed the dcache hit latency from VIPERCoalescer so HSAIL_X86
compiles after commit 496d5ed3e1f7dad42b0c2ebe0050d84621be8f99

Change-Id: I050a58d90f0f6356824c3c3bcb3f0b3c76d145e0
Signed-off-by: Pablo Prieto <pablo.prieto@unican.es>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19148
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Maintainer: Anthony Gutierrez <anthony.gutierrez@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>

x86: Stop CPUID from claiming we support xsave.

xsave is a fairly complex feature which we don't support in gem5, but
we do report that we support it through CPUID. It looks like I confused
it with FXSAVE which is an instruction related to SSE. This change
turns that bit back off again.

Change-Id: I00fc79168c5f7095b5241e870a4c8782e4385425
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20169
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Pouya Fotouhi <pfotouhi@ucdavis.edu>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

x86: Make unsuccessful CPUID instructions zero the result.

The previous implementation left the registers unmodified which is
technically correct since there is no defined behavior in that case or
a fault to raise. That would make what happened when the following code
consumed the result unpredictable because it would depend on what junk
values were left in the registers. This was originally not a problem
since the space of supported functions were tightly packed, but someone
added a new function with a gap without adjusting this behavior.

This change makes CPUID zero out RAX, RBX, RCX, and RDX when it fails.
That should be more predictable and cause less flakey failures.

Change-Id: If6ffb17c2969d34aff1600c0ffc32333d0b9be44
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20168
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Pouya Fotouhi <pfotouhi@ucdavis.edu>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

sim: Add a hook Clocked objects can implement for frequency changes.

This hook will let them implement whatever additional behavior is
necessary for when the clock changes.

An alternative design for this might have made the "update" function
virtual, and required anyone overriding it to call into the base class.
I think that would be an inferior design for two reasons. First, the
subclass author might forget to call update. Second, while it might
*seem* like this would have some performance benefit since you wouldn't
call into the virtual function and then call update, incurring the
function call overhead twice, you're going to call into update once
regardless, and then you're either going to call the virtual funciton
which does nothing (the norm) or does something. In either case you
call the same functions the same number of times.

There may be a slight penalty in code size since the call to update
may be inlined in the call sights before the virtual function, and
there will almost certainly be more of those than there would be
implementations of the virtual function, but that should be negligable
when compared to gem5's size as a whole.

Change-Id: Id25a5359f2b1f7e42c6d1dcbc70a37d3ce092d38
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20089
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Chun-Chen TK Hsu <chunchenhsu@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Reviewed-by: Srikant Bharadwaj <srikant.bharadwaj@amd.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>

sim: Clean up some mild style bugs in clocked_object.hh.

Clean up some formatting, and also remove redundant inline keywords.
A function defined in place in a class definition is already
automatically inline.

Change-Id: Iaad3a8dda6498c6a6068c2aabc9d6eb11f3d2eb2
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20088
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-ruby: Use check_on_cache_probe on MI

This change uses check_on_cache_probe statement to check if the cacheline
subject to eviction is locked in MI.

Change-Id: I276822e987e52f7682ff30f55880f295b6af023d
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19888
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>

mem-ruby: Use check_on_cache_probe on MOESI hammer

This change uses check_on_cache_probe statement to check if the cacheline
subject to eviction is locked in MOESI hammer.

Change-Id: I2c43f22aba5af3a57e54b1c435e5d3fbba86d1d5
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19891
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>

mem-ruby: Use check_on_cache_probe on MOESI CMP

This change uses check_on_cache_probe statement to check if the cacheline
subject to eviction is locked in MOESI CMP.

Change-Id: I3a8879e10ebd94ef68194836475e656761fed62c
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19908
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>

mem-ruby: Use check_on_cache_probe on MOESI

This change uses check_on_cache_probe statement to check if the cacheline
subject to eviction is locked in MOESI.

Change-Id: Ie650ccdc15bb41b4088e534975b662408aaccf24
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19890
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>

arch-arm: Added LD/ST<op> atomic instruction family and SWP instrs

Adding LD/ST/SWP family of instructions, LD/ST include a set of
operations like ADD/CLR/EOR/SET/UMAX/UMIN/SMAX/SMIN
This commit includes:
+ Instruction decode
+ Instruction functional code
+ New set of skeletons for Ex/Com/Ini/Constructor and declaration.

Change-Id: Ieea8d4256807e004d2f8aca8f421b3df8d76b116
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19812
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>

mem-ruby: Use check_on_cache_probe to protect locked lines from eviction

This change uses check_on_cache_probe statement to check if the cacheline
subject to eviction is locked in MESI Three Level.

Change-Id: Ib0de54aa067c7603db1f7321cc4825b123b641ac
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19868
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-ruby: Use check_on_cache_probe to protect locked lines from eviction

This change uses check_on_cache_probe statement to check if the cacheline
subject to eviction is locked in MESI Two Level. Other protocols should
be updated accordingly.

Signed-off-by: Pouya Fotouhi <pfotouhi@ucdavis.edu>
Change-Id: Idcdbc8ee528eb5e4e2f8d56a268a3a92eadd95b1
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/16809
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev-arm: Enable DTB autogeneration in GICv3

Change-Id: I539ae5ae74bc6f42f291441594a0d14c98e687f4
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20053
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev-arm: Fix PCI node's interrupt-map property

The PCI host has an interrupt-map property which only works for a fixed
setup of parent/child interrupt/address cells, which currently overlaps
with GICv2.
We want to make this flexible, so that the interrupt-map doesn't break
if we change the interrupt/address-cells value, and the patch is aiming
in that direction.  This is also needed for GICv3 DTB autogeneration,
since it is using different values than GICv2.

Change-Id: If1c661ddcbc0c277c9d6b0e44a0fd3fe2427618c
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20052
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev-arm: Use FdtState to generate GIC properites

Rather than hardcoding property values, we use a FdtState variable, so
that it is possible to retrieve them from an external object.

Change-Id: Ifd90814b03c68a7f55ef3be6123dcfee5e1de568
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20051
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

python: FdtState using interrupt-cells

Change-Id: I37813748f518b442d2b53c2bc4f381edb2e26146
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20050
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: Adding CAS/CASP AMO instr including new TypedAtomic func

CAS/CASP atomic instruction implementation
This change includes:
+ Instructions decode
+ new amo64.isa file where CAS/CASP main functional code is implemented
+ mem64.isa include Execute/complete/initiatie skeletons,
contructor and declarator
+ Added TypedAtomic function for pair register CASP instruction

Change-Id: I4a4acdec4ab1c8b888f10ef5dc1e896be8c432bf
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19811
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>

sim-se: rename Process::setpgid member

The getter methods to access these types of members do not
have a 'get' string in the method names. To make the interface
a bit more consistent, remove the 'set' part of the member name.

Change-Id: I04c56bd9d9feb1cf68ff50a1152083ea57ea7c62
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/20008
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Maintainer: Brandon Potter <Brandon.Potter@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>

misc: Update MAINTAINERS with a system tag

The code in system/ currently doesn't have a well-defined tag. Arm
components have traditionally used 'system-arm' and in some cases just
'arm'. This change introduces a separate tag for system software and
related files components.

Change-Id: I5c0ecd057bd65ab9d88c25f7ec7572d6941e41b5
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19931
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu: Pull more arch specialization to the top of BaseCPU.py.

This simplifies the logic of the CPU python class, and brings us ever
so slightly closer to factoring hardcoded ISA behavior out of non-ISA
specific components.

Change-Id: I7e4511dd4e6076f5c214be5af2a0e33af0142563
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19889
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>

system-arm: Refactor makefile to create targets with functions

This change simplifies writing targets which has same prefix but
differrent number of CPUs.

Change-Id: I3b7d67a554f5d27714ace6b88c9784ddaa3b34d5
Signed-off-by: Chun-Chen TK Hsu <chunchenhsu@google.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19989
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>

x86: Move some fixed or dummy config information into X86LocalApic.py.

The X86 local APIC doesn't actually use the pio_addr set in the config
and instead computes what address it will respond to based on the
initial ID of the CPU it's attached to. gem5's BasicPioDevice, which
the X86LocalApic class inherits from, does not provide a default value
for that parameter and will complain if *something* isn't set. The
value used, 0x2000000000000000, is a dummy value which is the base of
the region of the physical address space set aside for messages to
local APICs from the CPU and from other local APICs.

Also, the clock for the local APIC's timer is defined to be the bus
clock. The assumption seems to be that this has a 16:1 ratio with the
CPU clock, and I vaguely remember finding that that was more or less
unofficially true, even if it isn't necessary stringently defined to
be that.

Since we were already just assuming that that ratio was correct and
always setting up the local APICs clock that way, we can do that in
the X86LocalApic class definition and remove some special x86 specific
setup that we'd otherwise need for the x86 version of the Interrupt
class. If that's not correct, it can still be overridden somewhere else
in the config.

Change-Id: I50e84f899f44b1191c2ad79d05803b44f07001f9
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19968
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch: Bump MaxVecRegLenInBytes to 4096

The GPU model uses the generic vector register
containers, however the maximum vector register
length is fixed at 256, which is an invalid assumption
for the GPU model as it can operate on vectors up
to 4096B.

Change-Id: Id85e0ed45c9a9c1a4bb6e712c44eaeec2d628fce
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17908
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Maintainer: Anthony Gutierrez <anthony.gutierrez@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>

tests: Add Arm full system regressions to the new framework

Change-Id: I7e0499c8c3d63798d44f936580eecd40dc650694
Signed-off-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18989
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>

tests: Add support for downloaded archive fixtures

This changes add support for specifying fixtures that download
archives and uncompress them to the desired directory.

Change-Id: Ib3f6ee111b8d6130200507cbd170ecaf9fb39445
Signed-off-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/18988
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>

tests: Refactor the Gem5Fixture to derive from UniqueFixture

Gem5Fixture is used to define a fixture for building the gem5
binary. Most tests are expected to define their own Gem5Fixture,
however, as some might depend on the same binary (e.g.,
./build/ARM/gem5.opt), they will try to re-define a fixture for the
same target. This patchset changes Gem5Fixture to derive from
UniqueFixture.

In addition, this patchset changes the way global fixtures are
discovered to work with the new Gem5Fixture class. Instead of
enumerating them when test definitions are loaded, we do so after the
tests have been filtered according to specified tags (e.g., include
opt variant, exclude fast, debug variants).

Change-Id: Ie868a7e18ef6c3271f3c8a658229657cd43997cb
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19251
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>

tests: Add base class for fixtures that generate a target file

The new TargetFixture can be used as a base class for fixtures that
generate/download a file. These fixtures are guarrantied to be unique
and their setup function is only executed once.

Change-Id: I6a8737b06c4e74f3e29736ec363f61251d85da8c
Signed-off-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19250
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

sim-se: minor refactor for ProcessParams::create

Remove the nullptr initialization and change the message
for object file failure.

Change-Id: I14b80b47b65893c55810e7f715c1b8fc643c5125
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19949
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

sim-se: remove unused parameter

The init function which processes invoke on their page tables
has a thread context pointer parameter. The parameter is not
used by the code so remove it.

Change-Id: Ic4766fbc105d81c1c9ee4b5c0f428497dff2ab30
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19948
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

sim-se, tests: add a new sim-se test

This changeset adds a test to check the redirection features
added in faux-filesystem changeset. The test contains a
"chdir" system call to "/proc" which should be redirected to
"$(gem5-dir)/m5out/fs/proc" (as specified by the config files).

After "chdir", the test subsequently outputs the "/proc/cpuinfo"
file which should output a configuration of a fake cpu with
values set by a Python configuration file.

Note, the test will call "clone" once. To avoid a runtime error,
make sure that you run this test with "-n2" supplied to the
"config/example/se.py" script.

Change-Id: I505b046b7a4feddfa93a6ef0f0773ac43078cc94
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17112
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>

cpu-o3: fix atomic instructions non-speculative

Fix problem with O3 and AMO instructions. At initial stages amo
instruction is considered a type of non-speculative store. After
the instruction has been commited and during the squash step,
acquire_release version of the AMO operation is considered speculative,
that differents results in an assert fault. This fix ensures that AMO
instructions are always considered non-speculative, during early stages
and during squas/removal of the instruction.

Change-Id: Ia0c5fbb9dc44a9991337b57eb759b1ed08e4149e
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19815
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

cpu-o3: added _amo_op parameter in o3 LSQ

Fix bug with AMO (or RMW) instructions where the amo_op variable
is not being propagated to the LSQ request.

Change-Id: I60c59641d9b497051376f638e27f3c4cc361f615
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19814
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Anthony Gutierrez <anthony.gutierrez@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>

arch-arm: Add TypeAtomicOp class to be used by new atomic instructions

Creating a new object TypeAtomicOp that will be used by the atomic
instructions following gem5 AMO feature.

Change-Id: If082b596fb37d7a1cb569a4320c23505591df6a5
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19810
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: adding register control flags enabling LSE implementation

Added changes on arch-arm architecture to accept Atomic instructions
following ARM v8.1 documentation. That includes enabling atomic bit
in ID registers and add have_lse variable into arm system.

Change-Id: Ic28d3215d74ff129142fb51cb2fa217d3b1482de
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19809
Reviewed-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Maintainer: Giacomo Travaglini <giacomo.travaglini@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

dev-arm: Perform SMMUv3 CFG Invalidation at device interface

In the current SMMUv3 model, multiple micro/mainTLB are present at the
device interface (SMMUv3SlaveInterface), caching translations specific
to a device.
Those distributed TLBs are checked for a translation before checking for
centralized TLBs (shared by devices), like the configuration cache, walk
cache etc.  This means that if a hit in these TLBs occurs, there won't
be a need to enter configuration stage (which is where the STE and CD
are retrieved).  So if we invalidate a cached configuration (in
ConfigCache), we need to invalidate those interface TLB entries as well,
otherwise in theory we will keep the same translation even after a
change in configuration tables.

Change-Id: I4aa36ba8392a530267517bef7562318b282bee25
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Michiel van Tol <michiel.vantol@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19813
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-cache: Fix non-virtual base destructor of Repl Entry

ReplaceableEntry contains a virtual method, yet its destructor
was not virtual, causing errors in some compilers.

Change-Id: I13deec843f4007d9deb924882a8d98ff6a89c84f
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19808
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

sim-se: add new getpgrp system call

This changeset adds new (relatively simple) system call
support. The getpgrp call returns a thread context's
pgid.

Change-Id: I361bdbfb9c01b761ddd5a4923d23f86971f8d614
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17111
Maintainer: Brandon Potter <Brandon.Potter@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ayaz Akram <yazakram@ucdavis.edu>

sim-se: adding pipe2 syscall

pipe2 builds on top of the pipe syscall implementation by
adding some extra flags for the files (to avoid have to
make separate calls to fcntl).

Change-Id: I88cf6f1387b9d14e60b33a32db412da9ed93a3e6
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/12310
Reviewed-by: Brandon Potter <Brandon.Potter@amd.com>
Maintainer: Brandon Potter <Brandon.Potter@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: Implement ARMv8.1-PAN, Privileged access never

ARMv8.1-PAN adds a new bit to PSTATE. When the value of this PAN state
bit is 1, any privileged data access from EL1 or EL2 to a virtual memory
address that is accessible at EL0 generates a Permission fault.
This feature is mandatory in ARMv8.1 implementations.
This feature is supported in AArch64 and AArch32 states.
The ID_AA64MMFR1_EL1.PAN, ID_MMFR3_EL1.PAN, and ID_MMFR3.PAN fields
identify the support for ARMv8.1-PAN.

Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Change-Id: I94a76311711739dd2394c72944d88ba9321fd159
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19729
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

arch-arm: Rewrite MSR immediate instruction class

MSR <pstatefield>, #imm is used for setting a PSTATE field using an
immediate. Current implementation has the following flaws:

* There is no base MSR immediate definition: all the existing
PSTATE fields have a different class definition
* Those implementation make use of a generic data64 base class
which results in a wrong disassembly (pstate register is printed as an
integer register).

This patch is fixing this by defining a new base class (MiscRegImmOp64)
and new related templates. In this way, we aim to ease addition of new
PSTATE fields (in ARMv8.x)

Change-Id: I71b630ff32abe1b105bbb3ab5781c6589b67d419
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19728
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

systemc: Provide Port wrapper classes for sc_port

Add Port wrappers for sc_port, sc_interface, and sc_export.

Change-Id: Ib70cb7466461c8d32e3a260bf2a2d60a6ded87c3
Signed-off-by: Chun-Chen TK Hsu <chunchenhsu@google.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19788
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem-ruby: Remove assertion with incorrect assumption

Current code assumes that only one cacheline would either be in
RW. This is not true for GPU protocols, and may not be true
for some CPU-only protocols with state violations.

Change-Id: I70db4fbb4e80663551e8635307bb937a4db8dc63
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19708
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>

sim-se: small refactor on pipe syscall

Change-Id: I02ffb1c4af980554ff12ac7d11d32ba80fe261c5
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/12308
Reviewed-by: Brandon Potter <Brandon.Potter@amd.com>
Maintainer: Brandon Potter <Brandon.Potter@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem: Move eraseIfNullEntry to when holder is updated

The entry should only be tested for deletion when holder is updated.

Change-Id: I5a10b6fa876912709b7467860d43c23c60f38568
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19750
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

mem: Encapsulate retry variables of SnoopFilter

Group all variables related to the restoration of a snoop filter
entry due to a crossbar retry.

Change-Id: I4e03edb3afd06563b7a5812959739876709eceeb
Signed-off-by: Daniel R. Carvalho <odanrc@yahoo.com.br>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19749
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Maintainer: Nikos Nikoleris <nikos.nikoleris@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>

sim-se: small performance optimization

A local variable was being set inside a loop when it should
have been set outside the loop. This changeset moves the
variable to the appropriate place.

Change-Id: If7655b501bd819c39d35dea4c316b4b9ed3173a2
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17108
Maintainer: Brandon Potter <Brandon.Potter@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>

sim-se: fstat64 bugfix

The fstat64 system call does an upcast on entries in the file
descriptor array to check if the file descriptor has a backing
host-filesystem file opened. It does so because it needs to pass
the host fd into the fstat call (since we rely on the host
filesystem to service filesystem system calls).

The upcast was overly specific. This changeset alters the system
call to use the most general base class of the file descriptor
entries that can satisfy the code.

Change-Id: I10daf820257cea4d678ee6917e01e9cc9cd1cf5e
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17110
Tested-by: kokoro <noreply+kokoro@google.com>
Maintainer: Brandon Potter <Brandon.Potter@amd.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>