Our usage of 'vlenb' is overwhelming superior than the use of 'vlen'.
We're using 'vlenb' most of the time, having to do 'vlen >> 3' or
'vlen / 8' in every instance.
In hindsight we would be better if the 'vlenb' property was introduced
instead of 'vlen'. That's not what happened, and now we can't easily get
rid of it due to user scripts all around. What we can do, however, is to
change our internal representation to use 'vlenb'.
Add a 'vlenb' field in cpu->cfg. It'll be set via the existing 'vlen'
property, i.e. setting 'vlen' will also set 'vlenb'.
We'll replace all 'vlen >> 3' code to use 'vlenb' directly. Start with
the single instance we have in target/riscv/cpu.c.
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20240122161107.26737-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
After adding a KVM finalize() implementation, turn cbom_blocksize into a
class property. Follow the same design we used with 'vlen' and 'elen'.
The duplicated 'cbom_blocksize' KVM property can be removed from
kvm_riscv_add_cpu_user_properties().
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
tested-by tags added, rebased with Alistair's riscv-to-apply.next.
Message-ID: <20240112140201.127083-3-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
To turn cbom_blocksize and cboz_blocksize into class properties we need
KVM specific changes.
KVM is creating its own version of these options with a customized
setter() that prevents users from picking an invalid value during init()
time. This comes at the cost of duplicating each option that KVM
supports. This will keep happening for each new shared option KVM
implements in the future.
We can avoid that by using the same property TCG uses and adding
specific KVM handling during finalize() time, like TCG already does with
riscv_tcg_cpu_finalize_features(). To do that, the common CPU property
offers a way of knowing if an option was user set or not, sparing us
from doing unneeded syscalls.
riscv_kvm_cpu_finalize_features() is then created using the same
KVMScratch CPU we already use during init() time, since finalize() time
is still too early to use the official KVM CPU for it. cbom_blocksize
and cboz_blocksize are then handled during finalize() in the same way
they're handled by their KVM specific setter.
With this change we can proceed with the blocksize changes in the common
code without breaking the KVM driver.
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
tested-by tags added, rebased with Alistair's riscv-to-apply.next.
Message-ID: <20240112140201.127083-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Turning 'vlen' into a class property will allow its default value to be
overwritten by cpu_init() later on, solving the issue we have now where
CPU specific settings are getting overwritten by the default.
Common validation bits are moved from riscv_cpu_validate_v() to
prop_vlen_set() to be shared with KVM.
And, as done with every option we migrated to riscv_cpu_properties[],
vendor CPUs can't have their 'vlen' value changed.
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-9-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
'priv_spec' and 'vext_spec' are two string options used as a fancy way
of setting integers in the CPU state (cpu->env.priv_ver and
cpu->env.vext_ver). It requires us to deal with string parsing and to
store them in cpu_cfg.
We must support these string options, but we don't need to store them.
We have a precedence for this kind of arrangement in target/ppc/compat.c,
ppc_compat_prop_get|set, getters and setters used for the
'max-cpu-compat' class property of the pseries ppc64 machine. We'll do
the same with both 'priv_spec' and 'vext_spec'.
For 'priv_spec', the validation from riscv_cpu_validate_priv_spec() will
be done by the prop_priv_spec_set() setter, while also preventing it to
be changed for vendor CPUs. Add two helpers that converts env->priv_ver
back and forth to its string representation. These helpers allow us to
get a string and set 'env->priv_ver' and return a string giving the
current env->priv_ver value. In other words, make the cpu->cfg.priv_spec
string obsolete.
Last but not the least, move the reworked 'priv_spec' option to
riscv_cpu_properties[].
After all said and done, we don't need to store the 'priv_spec' string in
the CPU state, and we're now protecting vendor CPUs from priv_ver
changes:
$ ./build/qemu-system-riscv64 -M virt -cpu sifive-e51,priv_spec="v1.12.0"
qemu-system-riscv64: can't apply global sifive-e51-riscv-cpu.priv_spec=v1.12.0:
CPU 'sifive-e51' does not allow changing the value of 'priv_spec'
Current 'priv_spec' val: v1.10.0
$
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-7-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Commit 7f0bdfb5bf ("target/riscv/cpu.c: remove cfg setup from
riscv_cpu_init()") already did some of the work by making some
cpu_init() functions to explictly enable their own 'mmu' default.
The generic CPUs didn't get update by that commit, so they are still
relying on the defaults set by the 'mmu' option. But having 'mmu' and
'pmp' being default=true will force CPUs that doesn't implement these
options to set them to 'false' in their cpu_init(), which isn't ideal.
We'll move 'mmu' to riscv_cpu_properties[] without any defaults, i.e.
the default will be 'false'. Compensate it by manually setting 'mmu =
true' to the generic CPUs that requires it.
Implement a setter for it to forbid the 'mmu' setting to be changed for
vendor CPUs. This will allow the option to exist for all CPUs and, at
the same time, protect vendor CPUs from undesired changes:
$ ./build/qemu-system-riscv64 -M virt -cpu sifive-e51,mmu=true
qemu-system-riscv64: can't apply global sifive-e51-riscv-cpu.mmu=true:
CPU 'sifive-e51' does not allow changing the value of 'mmu'
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-5-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Every property in riscv_cpu_options[] will be migrated to
riscv_cpu_properties[]. This will make their default values init
earlier, allowing cpu_init() functions to overwrite them. We'll also
implement common getters and setters that both accelerators will use,
allowing them to share validations that TCG is doing.
At the same time, some options (namely 'vlen', 'elen' and the cache
blocksizes) need a way of tracking if the user set a value for them.
This is benign for TCG since the cost of always validating these values
are small, but for KVM we need syscalls to read the host values to make
the validations, thus knowing whether the user didn't touch the values
makes a difference.
We'll track user setting for these properties using a hash, like we do
in the TCG driver. All riscv cpu options will update this hash in case
the user sets it. The KVM driver will use this hash to minimize the
amount of syscalls done.
For now, both 'pmu-mask' and 'pmu-num' shouldn't be changed for vendor
CPUs. The existing setter for 'pmu-num' is changed to add this
restriction. New getters and setters are required for 'pmu-mask'
While we're at it, add a 'static' modifier to 'prop_pmu_num' since we're
not exporting it.
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Tested-by: Vladimir Isaev <vladimir.isaev@syntacore.com>
Message-ID: <20240105230546.265053-4-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
tcg: Introduce TCG_COND_TST{EQ,NE}
target/alpha: Use TCG_COND_TST{EQ,NE}
target/m68k: Use TCG_COND_TST{EQ,NE} in gen_fcc_cond
target/sparc: Use TCG_COND_TSTEQ in gen_op_mulscc
target/s390x: Use TCG_COND_TSTNE for CC_OP_{TM,ICM}
target/s390x: Improve general case of disas_jcc
# -----BEGIN PGP SIGNATURE-----
#
# iQFRBAABCgA7FiEEekgeeIaLTbaoWgXAZN846K9+IV8FAmXBpTAdHHJpY2hhcmQu
# aGVuZGVyc29uQGxpbmFyby5vcmcACgkQZN846K9+IV/p6gf9HAasTSRECk2cvjW9
# /mcJy0AIaespnI50fG8fm48OoFl0847CdrsJycpZ1spw3W3Wb0cVbMbq/teNMjXZ
# 0SGQJFk9Baq7wMhW7VzhSzJ96pcorpQprp7XBMdheLXqpT4zsM/EuwEAepBk8RUG
# 3kCeo38dswXE681ZafZkd/8pPzII19sQK8eiMpceeYkBsbbep+DDcnE18Ee4kISS
# u0SbuslKVahxd86LKuzrcz0pNFcmFuR5jRP9hmbQ0MfeAn0Pxlndi+ayZNghfgPf
# 3hDjskiionFwxb/OoRj45BssTWfDiluWl7IUsHfegPXCQ2Y+woT5Vq6TVGZn0GqS
# c6RLQQ==
# =TMiE
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 06 Feb 2024 03:19:12 GMT
# gpg: using RSA key 7A481E78868B4DB6A85A05C064DF38E8AF7E215F
# gpg: issuer "richard.henderson@linaro.org"
# gpg: Good signature from "Richard Henderson <richard.henderson@linaro.org>" [full]
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A 05C0 64DF 38E8 AF7E 215F
* tag 'pull-tcg-20240205-2' of https://gitlab.com/rth7680/qemu: (39 commits)
tcg/tci: Support TCG_COND_TST{EQ,NE}
tcg/s390x: Support TCG_COND_TST{EQ,NE}
tcg/s390x: Add TCG_CT_CONST_CMP
tcg/s390x: Split constraint A into J+U
tcg/ppc: Support TCG_COND_TST{EQ,NE}
tcg/ppc: Add TCG_CT_CONST_CMP
tcg/ppc: Tidy up tcg_target_const_match
tcg/ppc: Use cr0 in tcg_to_bc and tcg_to_isel
tcg/ppc: Sink tcg_to_bc usage into tcg_out_bc
tcg/sparc64: Support TCG_COND_TST{EQ,NE}
tcg/sparc64: Pass TCGCond to tcg_out_cmp
tcg/sparc64: Hoist read of tcg_cond_to_rcond
tcg/i386: Use TEST r,r to test 8/16/32 bits
tcg/i386: Improve TSTNE/TESTEQ vs powers of two
tcg/i386: Support TCG_COND_TST{EQ,NE}
tcg/i386: Move tcg_cond_to_jcc[] into tcg_out_cmp
tcg/i386: Pass x86 condition codes to tcg_out_cmov
tcg/arm: Support TCG_COND_TST{EQ,NE}
tcg/arm: Split out tcg_out_cmp()
tcg/aarch64: Generate CBNZ for TSTNE of UINT32_MAX
...
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
make vm-build-freebsd fails with:
ld: error: undefined symbol: inotify_init1
>>> referenced by filemonitor-inotify.c:183 (../src/util/filemonitor-inotify.c:183)
>>> util_filemonitor-inotify.c.o:(qemu_file_monitor_new) in archive libqemuutil.a
On FreeBSD the inotify functions are defined in libinotify.so. Add it
to the dependencies.
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20240206002344.12372-5-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Unlike on Linux, on FreeBSD renaming a file when the destination
already exists results in an IN_DELETE event for that existing file:
$ FILEMONITOR_DEBUG=1 build/tests/unit/test-util-filemonitor
Rename /tmp/test-util-filemonitor-K13LI2/fish/one.txt -> /tmp/test-util-filemonitor-K13LI2/two.txt
Event id=200000000 event=2 file=one.txt
Queue event id 200000000 event 2 file one.txt
Queue event id 100000000 event 2 file two.txt
Queue event id 100000002 event 2 file two.txt
Queue event id 100000000 event 0 file two.txt
Queue event id 100000002 event 0 file two.txt
Event id=100000000 event=0 file=two.txt
Expected event 0 but got 2
This difference in behavior is not expected to break the real users, so
teach the test to accept it.
Suggested-by: "Daniel P. Berrange" <berrange@redhat.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-ID: <20240206002344.12372-4-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
After console_sshd_config(), the SSH server needs to be nudged to pick
up the new configs. The scripts for the other BSD flavors already do
this with a reboot, but a simple reload is sufficient.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-ID: <20240206002344.12372-3-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
make vm-build-freebsd sometimes fails with "Connection timed out during
banner exchange". The client strace shows:
13:59:30 write(3, "SSH-2.0-OpenSSH_9.3\r\n", 21) = 21
13:59:30 getpid() = 252655
13:59:30 poll([{fd=3, events=POLLIN}], 1, 5000) = 1 ([{fd=3, revents=POLLIN}])
13:59:32 read(3, "S", 1) = 1
13:59:32 poll([{fd=3, events=POLLIN}], 1, 3625) = 1 ([{fd=3, revents=POLLIN}])
13:59:32 read(3, "S", 1) = 1
13:59:32 poll([{fd=3, events=POLLIN}], 1, 3625) = 1 ([{fd=3, revents=POLLIN}])
13:59:32 read(3, "H", 1) = 1
There is a 2s delay during connection, and ConnectTimeout is set to 1.
Raising it makes the issue go away, but we can do better. The server
truss shows:
888: 27.811414714 socket(PF_INET,SOCK_DGRAM|SOCK_CLOEXEC,0) = 5 (0x5)
888: 27.811765030 connect(5,{ AF_INET 10.0.2.3:53 },16) = 0 (0x0)
888: 27.812166941 sendto(5,"\^Z/\^A\0\0\^A\0\0\0\0\0\0\^A2"...,39,0,NULL,0) = 39 (0x27)
888: 29.363970743 poll({ 5/POLLRDNORM },1,5000) = 1 (0x1)
So the delay is due to a DNS query. Disable DNS queries in the server
config.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-ID: <20240206002344.12372-2-iii@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
QEMU initializes preallocated backend memory as the objects are parsed from
the command line. This is not optimal in some cases (e.g. memory spanning
multiple NUMA nodes) because the memory objects are initialized in series.
Allow the initialization to occur in parallel (asynchronously). In order to
ensure optimal thread placement, asynchronous initialization requires prealloc
context threads to be in use.
Signed-off-by: Mark Kanda <mark.kanda@oracle.com>
Message-ID: <20240131165327.3154970-2-mark.kanda@oracle.com>
Tested-by: Mario Casquero <mcasquer@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
We used to check that the memory region size is multiples of the overall
requested address alignment for the device memory address.
We removed that check, because there are cases (i.e., hv-balloon) where
devices unconditionally request an address alignment that has a very large
alignment (i.e., 32 GiB), but the actual memory device size might not be
multiples of that alignment.
However, this change:
(a) allows for some practically impossible DIMM sizes, like "1GB+1 byte".
(b) allows for DIMMs that partially cover hugetlb pages, previously
reported in [1].
Both scenarios don't make any sense: we might even waste memory.
So let's reintroduce that check, but only check that the
memory region size is multiples of the memory region alignment (i.e.,
page size, huge page size), but not any additional memory device
requirements communicated using md->get_min_alignment().
The following examples now fail again as expected:
(a) 1M with 2M THP
qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \
-object memory-backend-ram,id=mem1,size=1M \
-device pc-dimm,id=dimm1,memdev=mem1
-> backend memory size must be multiple of 0x200000
(b) 1G+1byte
qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \
-object memory-backend-ram,id=mem1,size=1073741825B \
-device pc-dimm,id=dimm1,memdev=mem1
-> backend memory size must be multiple of 0x200000
(c) Unliagned hugetlb size (2M)
qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \
-object memory-backend-file,id=mem1,mem-path=/dev/hugepages/tmp,size=511M \
-device pc-dimm,id=dimm1,memdev=mem1
backend memory size must be multiple of 0x200000
(d) Unliagned hugetlb size (1G)
qemu-system-x86_64 -m 4g,maxmem=16g,slots=1 -S -nodefaults -nographic \
-object memory-backend-file,id=mem1,mem-path=/dev/hugepages1G/tmp,size=2047M \
-device pc-dimm,id=dimm1,memdev=mem1
-> backend memory size must be multiple of 0x40000000
Note that this fix depends on a hv-balloon change to communicate its
additional alignment requirements using get_min_alignment() instead of
through the memory region.
[1] https://lkml.kernel.org/r/f77d641d500324525ac036fe1827b3070de75fc1.1701088320.git.mprivozn@redhat.com
Message-ID: <20240117135554.787344-3-david@redhat.com>
Reported-by: Zhenyu Zhang <zhenyzha@redhat.com>
Reported-by: Michal Privoznik <mprivozn@redhat.com>
Fixes: eb1b7c4bd4 ("memory-device: Drop size alignment check")
Tested-by: Zhenyu Zhang <zhenyzha@redhat.com>
Tested-by: Mario Casquero <mcasquer@redhat.com>
Reviewed-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
The character "+" is now forbidden in QOM device names (see commit
b447378e12 - "Limit type names to alphanumerical and some few special
characters"). For the "power5+" and "power7+" CPU names, there is
currently a hack in type_name_is_valid() to still allow them for
compatibility reasons. However, there is a much nicer solution for this:
Simply use aliases! This way we can still support the old names without
the need for the ugly hack in type_name_is_valid().
Message-ID: <20240117141054.73841-2-thuth@redhat.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Harsh Prateek Bora <harshpb@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
When the maximum count of SCRIPTS instructions is reached, the code
stops execution and returns, but fails to decrement the reentrancy
counter. This effectively renders the SCSI controller unusable
because on next entry the reentrancy counter is still above the limit.
This bug was seen on HP-UX 10.20 which seems to trigger SCRIPTS
loops.
Fixes: b987718bbb ("hw/scsi/lsi53c895a: Fix reentrancy issues in the LSI controller (CVE-2023-0330)")
Signed-off-by: Sven Schnelle <svens@stackframe.org>
Message-ID: <20240128202214.2644768-1-svens@stackframe.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Tested-by: Helge Deller <deller@gmx.de>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Let's implement the get_min_alignment() callback for memory devices, and
copy for the device memory region the alignment of the host memory
region. This mimics what virtio-mem does, and allows for re-introducing
proper alignment checks for the memory region size (where we don't care
about additional device requirements) in memory device core.
Message-ID: <20240117135554.787344-2-david@redhat.com>
Reviewed-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Better constraint for tcg_out_cmp, based on the comparison.
We can't yet remove the fallback to load constants into a
scratch because of tcg_out_cmp2, but that path should not
be as frequent.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
