Information about what kernel to load and how to load it was built
into the System object and its subclasses. That overloaded the System
object and made it responsible for too many things, and also was
somewhat awkward when working with SE mode which doesn't have a kernel.
This change extracts the kernel and information related to it from the
System object and puts into into a OsKernel or Workload object.
Currently the idea of a "Workload" to run and a kernel are a bit
muddled, an unfortunate carry-over from the original code. It's also an
implication of trying not to make too sweeping of a change, and to
minimize the number of times configs need to change, ie avoiding
creating a "kernel" parameter which would shortly thereafter be
renamed to "workload".
In future changes, the ideas of a kernel and a workload will be
disentangled, and workloads will be expanded to include emulated
operating systems which shephard and contain Process-es for syscall
emulation.
This change was originally split into pieces to make reviewing it
easier. Those reviews are here:
https: //gem5-review.googlesource.com/c/public/gem5/+/22243
https: //gem5-review.googlesource.com/c/public/gem5/+/24144
https: //gem5-review.googlesource.com/c/public/gem5/+/24145
https: //gem5-review.googlesource.com/c/public/gem5/+/24146
https: //gem5-review.googlesource.com/c/public/gem5/+/24147
https: //gem5-review.googlesource.com/c/public/gem5/+/24286
Change-Id: Ia3d863db276a023b6a2c7ee7a656d8142ff75589
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/26466
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Currently the System class has a mechanism to wait for a GDB connection
for each CPU which has requested it through one of its parameters.
Unfortunately, not every thread context/CPU will be ready for GDB at
that point, particularly considering that in an FS simulation the
kernel won't have been read so there will be no symbols, none of the
registers or the entry point will have been set.
Also in the fast models, the CPUs haven't had a chance to initialize
themselves enough by that point to respond to the API calls which are
used to implement GDB support.
Change-Id: If27cb3e0259a1f67599ab0493695b2f8af640d8e
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/24963
Reviewed-by: Gabe Black <gabeblack@google.com>
Reviewed-by: Chun-Chen TK Hsu <chunchenhsu@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
By using braced initializer lists and dropping the default
unimplementedFunc implementation function, the SyscallDesc tables
become a lot less crowded, and it's now very obvious which syscalls
are implemented just by quickly visually scanning the table.
This will also make it a lot easier to change the underlying type
stored in the table without having to adjust all of the instances
within them.
Jira Issue: https://gem5.atlassian.net/browse/GEM5-187
Change-Id: I7821de74812e1c02ca4550fc9c46cc2188cf1bd0
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/23189
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The logic that determines which syscall to call was built into the
implementation of faults/exceptions or even into the instruction
decoder, but that logic can depend on what OS is being used, and
sometimes even what version, for example 32bit vs. 64bit.
This change pushes that logic up into the Process objects since those
already handle a lot of the aspects of emulating the guest OS. Instead,
the ISA or fault implementations just notify the rest of the system
that a nebulous syscall has happened, and that gets propogated upward
until the process does something with it. That's very analogous to how
a system call would work on a real machine.
When a system call happens, the low level component which detects that
should call tc->syscall(&fault), where tc is the relevant thread (or
execution) context, and fault is a Fault which can ultimately be set
by the system call implementation.
The TC implementor (probably a CPU) will then have a chance to do
whatever it needs to to handle a system call. Currently only O3 does
anything special here. That implementor will end up calling the
Process's syscall() method.
Once in Process::syscall, the process object will use it's contextual
knowledge to determine what system call is being requested. It then
calls Process::doSyscall with the right syscall number, where doSyscall
centralizes the common mechanism for actually retrieving and calling
into the system call implementation.
Jira Issue: https://gem5.atlassian.net/browse/GEM5-187
Change-Id: I937ec1ef0576142c2a182ff33ca508d77ad0e7a1
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/23176
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Brandon Potter <Brandon.Potter@amd.com>
This abstraction will allow scheduling PCEvents for a particular
ThreadContext, all contexts on a CPU, all contexts in a system, etc.,
and delegates scheduling and removing events to each particular scope.
Right now the PCEventQueue is the only implementor of the PCEventSCope
interface.
Change-Id: I8fb62931511136229915c2e19d36aae7ffdec9df
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/22099
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Tested-by: kokoro <noreply+kokoro@google.com>
This change creates a distinction between object files which hold
executable code, and flat files which don't. The first type of files
have entry points, symbols, etc., while the others are just blobs which
can be shoved into memory. Rather than have those aspects but stub
them out, this change creates a new base class which simply doesn't
have them.
This change also restructures the ELF loader since it's main function
was quite long and doing multiple jobs.
It stops passing the architecture and operating system to the
ObjectFile constructor, since those might not be known at the very top
of the constructor. Instead, those default to Uknown*, and then are
filled in in the constructor body if appropriate. This removes a lot
of plumbing that was hard to actually use in practice.
It also introduces a mechanism to collect generic object file formats
so that they can be tried one by one by the general createObjectFile
function, rather than listing them all there one by one. It's unlikely
that new types of object files will need to be added in a modular way
without being able to modify the core loader code, but it's cleaner to
have that abstraction and modularization like is already there for
process loaders.
Finally, to make it possible to share the code which handles zipped
files for both true object files and also files which will be loaded
into memory but are just blobs, that mechanism is pulled out into a
new class called ImageFileData. It holds a collection of segments
which are set up by the object file and may refer to regions of the
original file, buffers maintained elsewhere, or even nothing to support
bss-es. shared_ptr is used to make it easier to keep track of that
information without having to do so explicitly or worry about deleting
a buffer before everyone was done using it.
Change-Id: I92890266f2ba0a703803cccad675a3ab41f2c4af
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/21467
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Brandon Potter <Brandon.Potter@amd.com>
Maintainer: Gabe Black <gabeblack@google.com>
A memory image can be described by an object file, but an object file
is more than a memory image. Also, it makes sense to manipulate a
memory image to, for instance, change how it's loaded into memory. That
takes on larger implications (relocations, the entry point, symbols,
etc.) when talking about the whole object file, and also modifies
aspects which may not need to change. For instance if an image needs
to be loaded into memory at addresses different from what's in the
object file, but other things like symbols need to stay unmodified.
Change-Id: Ia360405ffb2c1c48e0cc201ac0a0764357996a54
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/21466
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Brandon Potter <Brandon.Potter@amd.com>
Maintainer: Gabe Black <gabeblack@google.com>
The ObjectFile class has hardcoded assumptions that there are three
segments, text, bss and data. There are some files which have one
"segment" like raw files, where the entire file's contents are
considered a single segment. There are also ELF files which can have
an arbitrary number of segments, and those segments can hold any
number of sections, including the text, data and/or bss sections.
Removing this assumption frees up some object file formats from having
to twist themselves to fit in that structure, possibly introducing
ambiguities when some segments may fulfill multiple roles.
Change-Id: I976e06a3a90ef852b17a6485e2595b006b2090d5
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/21463
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
ELF is, in my opinion, the most important object file format gem5
currently understands, and in ELF terminolgy the blob of data that
needs to be loaded into memory to a particular location is called a
segment. A section is a software level view of what's in a region
of memory, and a single segment may contain multiple sections which
happen to follow each other in memory.
Change-Id: Ib810c5050723d5a96bd7550515b08ac695fb1b02
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/21462
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Gabe Black <gabeblack@google.com>
Set the default release to that single value for all ISAs.
glibc has checks for the kernel version based on uname, and refuses
to start any syscall emulation programs if those checks don't pass with
error:
FATAL: kernel too old
The ideal solution to this problem is to actually implement all missing
system calls for the required kernel version and bumping the release
accordingly.
However, it is very hard to implement all missing syscalls and verify
compliance.
Previously, we have simply bumped the version manually from time to
time when major glibc versions started breaking.
This commit alleviates the problem in two ways.
Firstly, having a single kernel version for all versions means that it is
easier to bump all versions at once.
Secondly, it makes it is possible to set the release with a parameter,
which in turn can be set from the command line with:
se.py --param 'system.cpu[:].workload[:].release = "4.18.0"'
Change-Id: I9e3c31073bfe68735f7b0775c8e299aa62b98222
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/17849
Maintainer: Brandon Potter <Brandon.Potter@amd.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Tested-by: kokoro <noreply+kokoro@google.com>
The system calls had four parameters. One of the parameters
is ThreadContext and another is Process. The ThreadContext
holds the value of the current process so the Process parameter
is redundant since the system call functions already have
indirect access.
With the old API, it is possible to call into the functions with
the wrong supplied Process which could end up being a confusing
error.
This patch removes the redundancy by forcing access through the
ThreadContext field within each system call.
Change-Id: Ib43d3f65824f6d425260dfd9f67de1892b6e8b7c
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/12299
Reviewed-by: Brandon Potter <Brandon.Potter@amd.com>
Maintainer: Brandon Potter <Brandon.Potter@amd.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Fix poll so that it will use the syscall retry capability
instead of causing a blocking call.
Add the accept and wait4 system calls.
Add polling to read to remove deadlocks that occur in the
event queue that are caused by blocking system calls.
Modify the write system call to return an error number in
case of error.
Change-Id: I0b4091a2e41e4187ebf69d63e0088f988f37d5da
Reviewed-on: https://gem5-review.googlesource.com/c/12115
Reviewed-by: Anthony Gutierrez <anthony.gutierrez@amd.com>
Maintainer: Anthony Gutierrez <anthony.gutierrez@amd.com>
Some parts of clone are architecture dependent. In some cases, we are
able to use architecture-specific helper functions or register
aliases. However, there is still some architecture-specific that is
protected by ifdefs in the common clone implementation.
Move these architecture-specific bits to the architecture-specific OS
class instead to avoid these ifdefs and make the code a bit more
readable.
Change-Id: Ia0903d738d0ba890863bddfa77e3b717db7f45de
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Cc: Giacomo Travaglini <giacomo.travaglini@arm.com>
Cc: Javier Setoain <javier.setoain@arm.com>
Cc: Brandon Potter <Brandon.Potter@amd.com>
Reviewed-on: https://gem5-review.googlesource.com/c/15435
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Brandon Potter <Brandon.Potter@amd.com>
This is especially important because the Ubuntu 18.04 packaged
arm-linux-gnueabihf-gcc uses the system call on the program initialization,
which leads all programs to fail with:
fatal: syscall openat (#322) unimplemented.
Change-Id: I5596162ad19644df7b6d21f2a46acc07030001ae
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/13004
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Trying to compile an ARM C hello world with arm-linux-gnueabihf-gcc from
Ubuntu 16.04 leads to a runtime failure with se.py:
FATAL: kernel too old
because the glibc tests if the kernel is at least 3.2.0, and gem5 was
reporting 3.0.0.
Furthermore, it is hard to obtain such toolchain at all: for example
crosstool-NG currently only allows for minimum kernels above 3.2.0.
3.7.0+ was chosen to match the aarch64 value, as it is likely that the
level of support will be very similar.
This commit does not guarantee that full 3.7.0 is supported, but it is
not likely that we had full 3.0.0 support previously either.
However, it is more likely that such support will be eventually achieved
if users can at least try out their programs and implement the missing
system calls as they are found.
Change-Id: I8df3763ae49788a6cb11cb0920e8202cd56b0f09
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/12986
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Jason Lowe-Power <jason@lowepower.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
A System object has a _numContexts member variable which represent the
number of ThreadContext registered in the System. Since this has to
match the size of the ThreadContext vector, this patch removes the
manually cached size. This was usually used as a for-loop index, whereas
we want to enforce the use of range-based loops whenever possible.
Change-Id: I1ba317c0393bcc9c1aeebbb1fc22d7b2bc2cf90c
Signed-off-by: Giacomo Travaglini <giacomo.travaglini@arm.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/8062
Reviewed-by: Gabe Black <gabeblack@google.com>
Maintainer: Brandon Potter <Brandon.Potter@amd.com>
32bit and 64bit Linux have different arguments passed to the
__switch_to() function that gem5 hooks into in order to collect context
switch statistics. 64bit Linux provides the task_struct pointer to the
next task that will be switched to, which means we don't have to look
up the task_struct from thread_info as we do in 32bit ARM Linux.
This patch adds a second set of accessors to ThreadInfo to extract
details such as the pid, tgid, task name, etc., directly from a
task_struct. The existing accessors maintain their existing behavior by
first looking up the task_struct and then calling these new accessors.
A 64-bit variant of the DumpStatsPCEvent class is added that uses these
new accessors to get the task details for the context switch dumps
directly from the task_struct passed to __switch_to().
Change-Id: I63c4b3e1ad64446751a91f6340901d5180d7382d
Reviewed-on: https://gem5-review.googlesource.com/2640
Reviewed-by: Curtis Dunham <curtis.dunham@arm.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Pau Cabre <pau.cabre@metempsy.com>
Reviewed-by: Andreas Sandberg <andreas.sandberg@arm.com>
Maintainer: Andreas Sandberg <andreas.sandberg@arm.com>
Modifies the clone system call and adds execve system call. Requires allowing
processes to steal thread contexts from other processes in the same system
object and the ability to detach pieces of process state (such as MemState)
to allow dynamic sharing.
The EIOProcess class was removed recently and it was the only other class
which derived from Process. Since every Process invocation is also a
LiveProcess invocation, it makes sense to simplify the organization by
combining the fields from LiveProcess into Process.
The class was crammed into syscall_emul.hh which has tons of forward
declarations and template definitions. To clean it up a bit, moved the
class into separate files and commented the class with doxygen style
comments. Also, provided some encapsulation by adding some accessors and
a mutator.
The syscallreturn.hh file was renamed syscall_return.hh to make it consistent
with other similarly named files in the src/sim directory.
The DPRINTF_SYSCALL macro was moved into its own header file with the
include the Base and Verbose flags as well.
--HG--
rename : src/sim/syscallreturn.hh => src/sim/syscall_return.hh
Add helper functions to dump the guest kernel's dmesg buffer to a text
file in m5out. This functionality is split into two parts. First, a
dmesg dump function that can be used in other places:
void Linux::dumpDmesg(ThreadContext *, std::ostream &)
This function is used to implement two PCEvents: DmesgDumpEvent and
KernelPanic event. The only difference between the two is that the
latter produces a gem5 panic instead of a warning in addition to
dumping the kernel log.
Change-Id: I6d2af1d666ace57124089648ea906f6c787ac63c
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
Reviewed-by: Gabor Dozsa <gabor.dozsa@arm.com>
The LinuxArmSystem class normally provides support for panicing gem5
if the simulated kernel panics. When this is turned off (default),
gem5 uses a BreakPCEvent to provide a debugger hook into the simulator
when the kernel crashes. This hook unconditionally kills gem5 with a
SIGTRAP unless gem5 is compiled in fast mode. This is undesirable
since the panic_on_panic param already provides similar functionality.
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-by: Nikos Nikoleris <nikos.nikoleris@arm.com>
The openFlagTable and mmapFlagTables for emulated Linux
platforms are basically identical, but are specified
repetitively for every platform. Use a common file
that gets included for each platform so that we only
have one copy, making them more consistent and simplifying
changes (like adding #ifdefs).
In the process, made some minor fixes that slipped through
due to previous inconsistencies, and added more #ifdefs
to try to fix building on alternative hosts.
Libraries are loaded into the process address space using the
mmap system call. Conveniently, this happens to be a good
time to update the process symbol table with the library's
incoming symbols so we handle the table update from within the
system call.
This works just like an application's normal symbols. The only
difference between a dynamic library and a main executable is
when the symbol table update occurs. The symbol table update for
an executable happens at program load time and is finished before
the process ever begins executing. Since dynamic linking happens
at runtime, the symbol loading happens after the library is
first loaded into the process address space. The library binary
is examined at this time for a symbol section and that section
is parsed for symbol types with specific bindings (global,
local, weak). Subsequently, these symbols are added to the table
and are available for use by gem5 for things like trace
generation.
Checkpointing should work just as it did previously. The address
space (and therefore the library) will be recorded and the symbol
table will be entirely recorded. (It's not possible to do anything
clever like checkpoint a program and then load the program back
with different libraries with LD_LIBRARY_PATH, because the
library becomes part of the address space after being loaded.)
For O3, which has a stat that counts reg reads, there is an additional
reg read per mmap() call since there's an arg we no longer ignore.
Otherwise, stats should not be affected.
The structure definition only had the open system call flag set in mind when
it was named, so we rename it here with the intention of using it to define
additional tables to translate flags for other system calls in the future.
This changeset adds support for changing the simulator output
directory. This can be useful when the simulation goes through several
stages (e.g., a warming phase, a simulation phase, and a verification
phase) since it allows the output from each stage to be located in a
different directory. Relocation is done by calling core.setOutputDir()
from Python or simout.setOutputDirectory() from C++.
This change affects several parts of the design of the gem5's output
subsystem. First, files returned by an OutputDirectory instance (e.g.,
simout) are of the type OutputStream instead of a std::ostream. This
allows us to do some more book keeping and control re-opening of files
when the output directory is changed. Second, new subdirectories are
OutputDirectory instances, which should be used to create files in
that sub-directory.
Signed-off-by: Andreas Sandberg <andreas@sandberg.pp.se>
[sascha.bischoff@arm.com: Rebased patches onto a newer gem5 version]
Signed-off-by: Sascha Bischoff <sascha.bischoff@arm.com>
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
