date:20240223

[PATCH] spapr: avoid overhead of finding vhyp class in critical operations

2024-02-23 Thread Nicholas Piggin

PPC_VIRTUAL_HYPERVISOR_GET_CLASS is used in critical operations like
interrupts and TLB misses and is quite costly. Running the
kvm-unit-tests sieve program with radix MMU enabled thrashes the TCG
TLB and spends a lot of time in TLB and page table walking code. The
test takes 67 seconds to complete with a lot of time being spent in
code related to finding the vhyp class:

   12.01%  [.] g_str_hash
8.94%  [.] g_hash_table_lookup
8.06%  [.] object_class_dynamic_cast
6.21%  [.] address_space_ldq
4.94%  [.] __strcmp_avx2
4.28%  [.] tlb_set_page_full
4.08%  [.] address_space_translate_internal
3.17%  [.] object_class_dynamic_cast_assert
2.84%  [.] ppc_radix64_xlate

Keep a pointer to the class and avoid this lookup. This reduces the
execution time to 40 seconds.

Signed-off-by: Nicholas Piggin 
---
This feels a bit ugly, but the performance problem of looking up the
class in fast paths can't be ignored. Is there a "nicer" way to get the
same result?

Thanks,
Nick

 target/ppc/cpu.h   |  3 ++-
 target/ppc/mmu-book3s-v3.h |  4 +---
 hw/ppc/pegasos2.c  |  1 +
 target/ppc/cpu_init.c  |  9 +++--
 target/ppc/excp_helper.c   | 16 
 target/ppc/kvm.c   |  4 +---
 target/ppc/mmu-hash64.c| 16 
 target/ppc/mmu-radix64.c   |  4 +---
 8 files changed, 17 insertions(+), 40 deletions(-)

diff --git a/target/ppc/cpu.h b/target/ppc/cpu.h
index ec14574d14..eb85d9aa71 100644
--- a/target/ppc/cpu.h
+++ b/target/ppc/cpu.h
@@ -1437,6 +1437,7 @@ struct ArchCPU {
 int vcpu_id;
 uint32_t compat_pvr;
 PPCVirtualHypervisor *vhyp;
+PPCVirtualHypervisorClass *vhyp_class;
 void *machine_data;
 int32_t node_id; /* NUMA node this CPU belongs to */
 PPCHash64Options *hash64_opts;
@@ -1535,7 +1536,7 @@ DECLARE_OBJ_CHECKERS(PPCVirtualHypervisor, 
PPCVirtualHypervisorClass,
 
 static inline bool vhyp_cpu_in_nested(PowerPCCPU *cpu)
 {
-return PPC_VIRTUAL_HYPERVISOR_GET_CLASS(cpu->vhyp)->cpu_in_nested(cpu);
+return cpu->vhyp_class->cpu_in_nested(cpu);
 }
 #endif /* CONFIG_USER_ONLY */
 
diff --git a/target/ppc/mmu-book3s-v3.h b/target/ppc/mmu-book3s-v3.h
index 674377a19e..f3f7993958 100644
--- a/target/ppc/mmu-book3s-v3.h
+++ b/target/ppc/mmu-book3s-v3.h
@@ -108,9 +108,7 @@ static inline hwaddr ppc_hash64_hpt_mask(PowerPCCPU *cpu)
 uint64_t base;
 
 if (cpu->vhyp) {
-PPCVirtualHypervisorClass *vhc =
-PPC_VIRTUAL_HYPERVISOR_GET_CLASS(cpu->vhyp);
-return vhc->hpt_mask(cpu->vhyp);
+return cpu->vhyp_class->hpt_mask(cpu->vhyp);
 }
 if (cpu->env.mmu_model == POWERPC_MMU_3_00) {
 ppc_v3_pate_t pate;
diff --git a/hw/ppc/pegasos2.c b/hw/ppc/pegasos2.c
index 04d6decb2b..c22e8b336d 100644
--- a/hw/ppc/pegasos2.c
+++ b/hw/ppc/pegasos2.c
@@ -400,6 +400,7 @@ static void pegasos2_machine_reset(MachineState *machine, 
ShutdownCause reason)
 machine->fdt = fdt;
 
 pm->cpu->vhyp = PPC_VIRTUAL_HYPERVISOR(machine);
+pm->cpu->vhyp_class = PPC_VIRTUAL_HYPERVISOR_GET_CLASS(pm->cpu->vhyp);
 }
 
 enum pegasos2_rtas_tokens {
diff --git a/target/ppc/cpu_init.c b/target/ppc/cpu_init.c
index 9bccddb350..63d0094024 100644
--- a/target/ppc/cpu_init.c
+++ b/target/ppc/cpu_init.c
@@ -6631,6 +6631,7 @@ void cpu_ppc_set_vhyp(PowerPCCPU *cpu, 
PPCVirtualHypervisor *vhyp)
 CPUPPCState *env = >env;
 
 cpu->vhyp = vhyp;
+cpu->vhyp_class = PPC_VIRTUAL_HYPERVISOR_GET_CLASS(vhyp);
 
 /*
  * With a virtual hypervisor mode we never allow the CPU to go
@@ -7224,9 +7225,7 @@ static void ppc_cpu_exec_enter(CPUState *cs)
 PowerPCCPU *cpu = POWERPC_CPU(cs);
 
 if (cpu->vhyp) {
-PPCVirtualHypervisorClass *vhc =
-PPC_VIRTUAL_HYPERVISOR_GET_CLASS(cpu->vhyp);
-vhc->cpu_exec_enter(cpu->vhyp, cpu);
+cpu->vhyp_class->cpu_exec_enter(cpu->vhyp, cpu);
 }
 }
 
@@ -7235,9 +7234,7 @@ static void ppc_cpu_exec_exit(CPUState *cs)
 PowerPCCPU *cpu = POWERPC_CPU(cs);
 
 if (cpu->vhyp) {
-PPCVirtualHypervisorClass *vhc =
-PPC_VIRTUAL_HYPERVISOR_GET_CLASS(cpu->vhyp);
-vhc->cpu_exec_exit(cpu->vhyp, cpu);
+cpu->vhyp_class->cpu_exec_exit(cpu->vhyp, cpu);
 }
 }
 #endif /* CONFIG_TCG */
diff --git a/target/ppc/excp_helper.c b/target/ppc/excp_helper.c
index 98952de267..445350488c 100644
--- a/target/ppc/excp_helper.c
+++ b/target/ppc/excp_helper.c
@@ -840,9 +840,7 @@ static void powerpc_excp_7xx(PowerPCCPU *cpu, int excp)
  * HV mode, we need to keep hypercall support.
  */
 if (lev == 1 && cpu->vhyp) {
-PPCVirtualHypervisorClass *vhc =
-PPC_VIRTUAL_HYPERVISOR_GET_CLASS(cpu->vhyp);
-vhc->hypercall(cpu->vhyp, cpu);
+cpu->vhyp_class->hypercall(cpu->vhyp, cpu);
 powerpc_reset_excp_state(cpu);
 return;
 }
@@ -1012,9 +1010,7 @@ static void powerpc_excp_74xx(PowerPCCPU *cpu, int excp)
  * HV

Re: Support Android hypervisors

2024-02-23 Thread RR NN

{
  "emoji": "",
  "version": 1
}

[PULL v2 31/39] *-user: Deprecate and disable -p pagesize

2024-02-23 Thread Richard Henderson

This option controls the host page size.  From the mis-usage in
our own testsuite, this is easily confused with guest page size.

The only thing that occurs when changing the host page size is
that stuff breaks, because one cannot actually change the host
page size.  Therefore reject all but the no-op setting as part
of the deprecation process.

Reviewed-by: Warner Losh 
Signed-off-by: Richard Henderson 
Reviewed-by: Philippe Mathieu-Daudé 
Acked-by: Helge Deller 
Message-Id: <20240102015808.132373-27-richard.hender...@linaro.org>
---
 docs/about/deprecated.rst | 10 ++
 docs/user/main.rst|  3 ---
 bsd-user/main.c   | 11 ++-
 linux-user/main.c | 12 ++--
 4 files changed, 22 insertions(+), 14 deletions(-)

diff --git a/docs/about/deprecated.rst b/docs/about/deprecated.rst
index 5a2305ccd6..3074303b9c 100644
--- a/docs/about/deprecated.rst
+++ b/docs/about/deprecated.rst
@@ -63,6 +63,16 @@ as short-form boolean values, and passed to plugins as 
``arg_name=on``.
 However, short-form booleans are deprecated and full explicit ``arg_name=on``
 form is preferred.
 
+User-mode emulator command line arguments
+-
+
+``-p`` (since 9.0)
+''
+
+The ``-p`` option pretends to control the host page size.  However,
+it is not possible to change the host page size, and using the
+option only causes failures.
+
 QEMU Machine Protocol (QMP) commands
 
 
diff --git a/docs/user/main.rst b/docs/user/main.rst
index 7e7ad07409..d5fbb78d3c 100644
--- a/docs/user/main.rst
+++ b/docs/user/main.rst
@@ -87,9 +87,6 @@ Debug options:
Activate logging of the specified items (use '-d help' for a list of
log items)
 
-``-p pagesize``
-   Act as if the host page size was 'pagesize' bytes
-
 ``-g port``
Wait gdb connection to port
 
diff --git a/bsd-user/main.c b/bsd-user/main.c
index e5efb7b845..6ab3efd6c0 100644
--- a/bsd-user/main.c
+++ b/bsd-user/main.c
@@ -364,11 +364,12 @@ int main(int argc, char **argv)
 } else if (!strcmp(r, "L")) {
 interp_prefix = argv[optind++];
 } else if (!strcmp(r, "p")) {
-qemu_host_page_size = atoi(argv[optind++]);
-if (qemu_host_page_size == 0 ||
-(qemu_host_page_size & (qemu_host_page_size - 1)) != 0) {
-fprintf(stderr, "page size must be a power of two\n");
-exit(1);
+unsigned size, want = qemu_real_host_page_size();
+
+r = argv[optind++];
+if (qemu_strtoui(r, NULL, 10, ) || size != want) {
+warn_report("Deprecated page size option cannot "
+"change host page size (%u)", want);
 }
 } else if (!strcmp(r, "g")) {
 gdbstub = g_strdup(argv[optind++]);
diff --git a/linux-user/main.c b/linux-user/main.c
index e540acb84a..bad03f06d3 100644
--- a/linux-user/main.c
+++ b/linux-user/main.c
@@ -332,11 +332,11 @@ static void handle_arg_ld_prefix(const char *arg)
 
 static void handle_arg_pagesize(const char *arg)
 {
-qemu_host_page_size = atoi(arg);
-if (qemu_host_page_size == 0 ||
-(qemu_host_page_size & (qemu_host_page_size - 1)) != 0) {
-fprintf(stderr, "page size must be a power of two\n");
-exit(EXIT_FAILURE);
+unsigned size, want = qemu_real_host_page_size();
+
+if (qemu_strtoui(arg, NULL, 10, ) || size != want) {
+warn_report("Deprecated page size option cannot "
+"change host page size (%u)", want);
 }
 }
 
@@ -496,7 +496,7 @@ static const struct qemu_argument arg_table[] = {
 {"D",  "QEMU_LOG_FILENAME", true, handle_arg_log_filename,
  "logfile", "write logs to 'logfile' (default stderr)"},
 {"p",  "QEMU_PAGESIZE",true,  handle_arg_pagesize,
- "pagesize",   "set the host page size to 'pagesize'"},
+ "pagesize",   "deprecated change to host page size"},
 {"one-insn-per-tb",
"QEMU_ONE_INSN_PER_TB",  false, handle_arg_one_insn_per_tb,
  "",   "run with one guest instruction per emulated TB"},
-- 
2.34.1

[PULL v2 00/39] tcg and linux-user patch queue

2024-02-23 Thread Richard Henderson

v2: Fix bsd-user build errors.


r~


The following changes since commit 3d54cbf269d63ff1d500b35b2bcf4565ff8ad485:

  Merge tag 'hw-misc-20240222' of https://github.com/philmd/qemu into staging 
(2024-02-22 15:44:29 +)

are available in the Git repository at:

  https://gitlab.com/rth7680/qemu.git tags/pull-tcg-20240222-2

for you to fetch changes up to fcc6ad372f56d3f47b6d5457a904916b48b9e114:

  linux-user: Remove pgb_dynamic alignment assertion (2024-02-23 15:07:03 -0800)


tcg/aarch64: Apple does not align __int128_t in even registers
accel/tcg: Fixes for page tables in mmio memory
linux-user: Remove qemu_host_page_{size,mask}, HOST_PAGE_ALIGN
migration: Remove qemu_host_page_size
hw/tpm: Remove qemu_host_page_size
softmmu: Remove qemu_host_page_{size,mask}, HOST_PAGE_ALIGN
linux-user: Split and reorganize target_mmap.
*-user: Deprecate and disable -p pagesize
linux-user: Allow TARGET_PAGE_BITS_VARY
target/alpha: Enable TARGET_PAGE_BITS_VARY for user-only
target/arm: Enable TARGET_PAGE_BITS_VARY for AArch64 user-only
target/ppc: Enable TARGET_PAGE_BITS_VARY for user-only
linux-user: Remove pgb_dynamic alignment assertion


Jonathan Cameron (1):
  tcg: Avoid double lock if page tables happen to be in mmio memory.

Peter Maydell (1):
  accel/tcg: Set can_do_io at at start of lookup_tb_ptr helper

Richard Henderson (37):
  tcg/aarch64: Apple does not align __int128_t in even registers
  accel/tcg: Remove qemu_host_page_size from page_protect/page_unprotect
  linux-user: Adjust SVr4 NULL page mapping
  linux-user: Remove qemu_host_page_{size, mask} in probe_guest_base
  linux-user: Remove qemu_host_page_size from create_elf_tables
  linux-user/hppa: Simplify init_guest_commpage
  linux-user/nios2: Remove qemu_host_page_size from init_guest_commpage
  linux-user/arm: Remove qemu_host_page_size from init_guest_commpage
  linux-user: Remove qemu_host_page_{size, mask} from mmap.c
  linux-user: Remove REAL_HOST_PAGE_ALIGN from mmap.c
  linux-user: Remove HOST_PAGE_ALIGN from mmap.c
  migration: Remove qemu_host_page_size
  hw/tpm: Remove HOST_PAGE_ALIGN from tpm_ppi_init
  softmmu/physmem: Remove qemu_host_page_size
  softmmu/physmem: Remove HOST_PAGE_ALIGN
  linux-user: Remove qemu_host_page_size from main
  linux-user: Split out target_mmap__locked
  linux-user: Move some mmap checks outside the lock
  linux-user: Fix sub-host-page mmap
  linux-user: Split out mmap_end
  linux-user: Do early mmap placement only for reserved_va
  linux-user: Split out do_munmap
  linux-user: Use do_munmap for target_mmap failure
  linux-user: Split out mmap_h_eq_g
  linux-user: Split out mmap_h_lt_g
  linux-user: Split out mmap_h_gt_g
  tests/tcg: Remove run-test-mmap-*
  tests/tcg: Extend file in linux-madvise.c
  *-user: Deprecate and disable -p pagesize
  cpu: Remove page_size_init
  accel/tcg: Disconnect TargetPageDataNode from page size
  linux-user: Allow TARGET_PAGE_BITS_VARY
  target/arm: Enable TARGET_PAGE_BITS_VARY for AArch64 user-only
  linux-user: Bound mmap_min_addr by host page size
  target/ppc: Enable TARGET_PAGE_BITS_VARY for user-only
  target/alpha: Enable TARGET_PAGE_BITS_VARY for user-only
  linux-user: Remove pgb_dynamic alignment assertion

 docs/about/deprecated.rst |  10 +
 docs/user/main.rst|   3 -
 bsd-user/qemu.h   |   7 +
 include/exec/cpu-common.h |   7 -
 include/hw/core/cpu.h |   2 -
 target/alpha/cpu-param.h  |  16 +-
 target/arm/cpu-param.h|   6 +-
 target/ppc/cpu-param.h|   9 +-
 tcg/aarch64/tcg-target.h  |   6 +-
 accel/tcg/cpu-exec.c  |   8 +
 accel/tcg/cputlb.c|  34 +-
 accel/tcg/translate-all.c |   1 -
 accel/tcg/user-exec.c |  31 +-
 bsd-user/main.c   |  23 +-
 cpu-target.c  |  16 -
 hw/tpm/tpm_ppi.c  |   6 +-
 linux-user/elfload.c  |  68 +--
 linux-user/main.c |  34 +-
 linux-user/mmap.c | 767 ++
 migration/ram.c   |  22 +-
 system/physmem.c  |  17 +-
 system/vl.c   |   1 -
 target/arm/cpu.c  |  51 +-
 tests/tcg/multiarch/linux/linux-madvise.c |   2 +
 tests/tcg/alpha/Makefile.target   |   3 -
 tests/tcg/arm/Makefile.target |   3 -
 tests/tcg/hppa/Makefile.target|   3 -
 tests/tcg/i386/Makefile.target|   3 -
 tests/tcg/m68k/Makefile.target

[PULL v2 32/39] cpu: Remove page_size_init

2024-02-23 Thread Richard Henderson

Move qemu_host_page_{size,mask} and HOST_PAGE_ALIGN into bsd-user.
It should be removed from bsd-user as well, but defer that cleanup.

Reviewed-by: Warner Losh 
Reviewed-by: Philippe Mathieu-Daudé 
Signed-off-by: Richard Henderson 
Tested-by: Ilya Leoshkevich 
Acked-by: Helge Deller 
Message-Id: <20240102015808.132373-28-richard.hender...@linaro.org>
---
 bsd-user/qemu.h   |  7 +++
 include/exec/cpu-common.h |  7 ---
 include/hw/core/cpu.h |  2 --
 accel/tcg/translate-all.c |  1 -
 bsd-user/main.c   | 12 
 cpu-target.c  | 16 
 system/vl.c   |  1 -
 7 files changed, 19 insertions(+), 27 deletions(-)

diff --git a/bsd-user/qemu.h b/bsd-user/qemu.h
index dc842fffa7..c05c512767 100644
--- a/bsd-user/qemu.h
+++ b/bsd-user/qemu.h
@@ -39,6 +39,13 @@ extern char **environ;
 #include "qemu/clang-tsa.h"
 
 #include "qemu-os.h"
+/*
+ * TODO: Remove these and rely only on qemu_real_host_page_size().
+ */
+extern uintptr_t qemu_host_page_size;
+extern intptr_t qemu_host_page_mask;
+#define HOST_PAGE_ALIGN(addr) ROUND_UP((addr), qemu_host_page_size)
+
 /*
  * This struct is used to hold certain information about the image.  Basically,
  * it replicates in user space what would be certain task_struct fields in the
diff --git a/include/exec/cpu-common.h b/include/exec/cpu-common.h
index 9ead1be100..6346df17ce 100644
--- a/include/exec/cpu-common.h
+++ b/include/exec/cpu-common.h
@@ -20,13 +20,6 @@
 void cpu_exec_init_all(void);
 void cpu_exec_step_atomic(CPUState *cpu);
 
-/* Using intptr_t ensures that qemu_*_page_mask is sign-extended even
- * when intptr_t is 32-bit and we are aligning a long long.
- */
-extern uintptr_t qemu_host_page_size;
-extern intptr_t qemu_host_page_mask;
-
-#define HOST_PAGE_ALIGN(addr) ROUND_UP((addr), qemu_host_page_size)
 #define REAL_HOST_PAGE_ALIGN(addr) ROUND_UP((addr), qemu_real_host_page_size())
 
 /* The CPU list lock nests outside page_(un)lock or mmap_(un)lock */
diff --git a/include/hw/core/cpu.h b/include/hw/core/cpu.h
index 4385ce54c9..5c2d55f6d2 100644
--- a/include/hw/core/cpu.h
+++ b/include/hw/core/cpu.h
@@ -1179,8 +1179,6 @@ bool target_words_bigendian(void);
 
 const char *target_name(void);
 
-void page_size_init(void);
-
 #ifdef NEED_CPU_H
 
 #ifndef CONFIG_USER_ONLY
diff --git a/accel/tcg/translate-all.c b/accel/tcg/translate-all.c
index 1c695efe02..c1f57e894a 100644
--- a/accel/tcg/translate-all.c
+++ b/accel/tcg/translate-all.c
@@ -256,7 +256,6 @@ bool cpu_unwind_state_data(CPUState *cpu, uintptr_t 
host_pc, uint64_t *data)
 
 void page_init(void)
 {
-page_size_init();
 page_table_config_init();
 }
 
diff --git a/bsd-user/main.c b/bsd-user/main.c
index 6ab3efd6c0..512d4ab69f 100644
--- a/bsd-user/main.c
+++ b/bsd-user/main.c
@@ -49,6 +49,13 @@
 #include "host-os.h"
 #include "target_arch_cpu.h"
 
+
+/*
+ * TODO: Remove these and rely only on qemu_real_host_page_size().
+ */
+uintptr_t qemu_host_page_size;
+intptr_t qemu_host_page_mask;
+
 static bool opt_one_insn_per_tb;
 uintptr_t guest_base;
 bool have_guest_base;
@@ -307,6 +314,9 @@ int main(int argc, char **argv)
 (void) envlist_setenv(envlist, *wrk);
 }
 
+qemu_host_page_size = getpagesize();
+qemu_host_page_size = MAX(qemu_host_page_size, TARGET_PAGE_SIZE);
+
 cpu_model = NULL;
 
 qemu_add_opts(_trace_opts);
@@ -404,6 +414,8 @@ int main(int argc, char **argv)
 }
 }
 
+qemu_host_page_mask = -qemu_host_page_size;
+
 /* init debug */
 {
 int mask = 0;
diff --git a/cpu-target.c b/cpu-target.c
index 86444cc2c6..4c0621bf33 100644
--- a/cpu-target.c
+++ b/cpu-target.c
@@ -45,9 +45,6 @@
 #include "trace/trace-root.h"
 #include "qemu/accel.h"
 
-uintptr_t qemu_host_page_size;
-intptr_t qemu_host_page_mask;
-
 #ifndef CONFIG_USER_ONLY
 static int cpu_common_post_load(void *opaque, int version_id)
 {
@@ -474,16 +471,3 @@ const char *target_name(void)
 {
 return TARGET_NAME;
 }
-
-void page_size_init(void)
-{
-/* NOTE: we can always suppose that qemu_host_page_size >=
-   TARGET_PAGE_SIZE */
-if (qemu_host_page_size == 0) {
-qemu_host_page_size = qemu_real_host_page_size();
-}
-if (qemu_host_page_size < TARGET_PAGE_SIZE) {
-qemu_host_page_size = TARGET_PAGE_SIZE;
-}
-qemu_host_page_mask = -(intptr_t)qemu_host_page_size;
-}
diff --git a/system/vl.c b/system/vl.c
index b8469d9965..7913cc28aa 100644
--- a/system/vl.c
+++ b/system/vl.c
@@ -2118,7 +2118,6 @@ static void qemu_create_machine(QDict *qdict)
 }
 
 cpu_exec_init_all();
-page_size_init();
 
 if (machine_class->hw_version) {
 qemu_set_hw_version(machine_class->hw_version);
-- 
2.34.1

[PATCH] virtio-gpu: Fix HW cursor visibility

2024-02-23 Thread Satyeshwar Singh

When show-cursor is on, most of the time Windows VM draws the cursor by
itself and hands it over to Qemu as a separate resource. However,
sometimes, Windows OS gives control of the cursor to applications like
Notepad. In such cases a software cursor which is part of the overall
framebuffer is drawn by the application. Windows intimates the indirect
display driver (IDD) about this change through a flag and IDD is in turn
responsible for communicating with the hypervisor (Qemu) to hide the HW
cursor. This show/hide cursor can happen any time dynamically.

Unfortunately, it seems like Qemu doesn't expect this change to happen
dynamically. The code in virtio-gpu.c was written such that
update_cursor would first call dpy_cursor_define if the cursor shape has
changed and this is not a simple move operation (which indeed is the
case when moving to a SW cursor) and then call dpy_mouse_set.
dpy_cursor_define calls toolkits like GTK but in addition to creating
the cursor content, it also calls gdk_window_set_cursor thereby setting
the cursor. It is therefore, the right function to either show or hide
the cursor but there was no code present to hide the cursor. Changing
the cursor visibility in dpy_mouse_set has two issues. First,
dpy_cursor_define already decided to display the cursor so showing the
cursor in the previous call only to hide it in dpy_mouse_set doesn't
sound right. Second, dpy_mouse_set skips the rest of the code if we
are in absolute mode so adding this code there wouldn't work in that
mode.

Qemu makes the decision of whether to show or hide the cursor by
observing the cursor->resource_id flag in virtio-gpu.c as is evident
from the lines
dpy_mouse_set(s->con, cursor->pos.x, cursor->pos.y,
  cursor->resource_id ? 1 : 0);
The last parameter is considered the "visible" parameter in gdk code.
Therefore, in this patch we continue with the same model. Instead of
changing the function prototype and changing a dozen plus files, we are
adding the visible field in QEMUCursor data structure and passing
it from virtio-gpu to the GTK code where we check if the cursor is
visible or not. If not, we simply call gdk_window_set_cursor with NULL
parameter, thereby hiding the cursor. Once Windows VM switches back to
the HW cursor, then IDD would again provide a new resource_id to Qemu
and we can start displaying it once more.

Signed-off-by: Satyeshwar Singh 
Cc: Marc-André Lureau 
Cc: Vivek Kasireddy 
Cc: Dongwon Kim 
---
 hw/display/virtio-gpu.c | 1 +
 include/ui/console.h| 1 +
 ui/gtk.c| 5 +
 3 files changed, 7 insertions(+)

diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c
index 1c1ee230b3..1ae9be605b 100644
--- a/hw/display/virtio-gpu.c
+++ b/hw/display/virtio-gpu.c
@@ -98,6 +98,7 @@ static void update_cursor(VirtIOGPU *g, struct 
virtio_gpu_update_cursor *cursor)
 
 s->current_cursor->hot_x = cursor->hot_x;
 s->current_cursor->hot_y = cursor->hot_y;
+s->current_cursor->visible = cursor->resource_id ? 1 : 0;
 
 if (cursor->resource_id > 0) {
 vgc->update_cursor_data(g, s, cursor->resource_id);
diff --git a/include/ui/console.h b/include/ui/console.h
index a4a49ffc64..47c39ed405 100644
--- a/include/ui/console.h
+++ b/include/ui/console.h
@@ -161,6 +161,7 @@ typedef struct QEMUCursor {
 uint16_twidth, height;
 int hot_x, hot_y;
 int refcount;
+int visible;
 uint32_tdata[];
 } QEMUCursor;
 
diff --git a/ui/gtk.c b/ui/gtk.c
index 810d7fc796..12a76de570 100644
--- a/ui/gtk.c
+++ b/ui/gtk.c
@@ -478,6 +478,11 @@ static void gd_cursor_define(DisplayChangeListener *dcl,
 return;
 }
 
+if(!c->visible) {
+gdk_window_set_cursor(gtk_widget_get_window(vc->gfx.drawing_area), 
NULL);
+return;
+}
+
 pixbuf = gdk_pixbuf_new_from_data((guchar *)(c->data),
   GDK_COLORSPACE_RGB, true, 8,
   c->width, c->height, c->width * 4,
-- 
2.33.1

[PATCH 1/1] virtio-gpu: Fix HW cursor visibility

2024-02-23 Thread Singh, Satyeshwar

When show-cursor is on, most of the time Windows VM draws the cursor by
itself and hands it over to Qemu as a separate resource. However,
sometimes, Windows OS gives control of the cursor to applications like
Notepad. In such cases a software cursor which is part of the overall
framebuffer is drawn by the application. Windows intimates the indirect
display driver (IDD) about this change through a flag and IDD is in turn
responsible for communicating with the hypervisor (Qemu) to hide the HW
cursor. This show/hide cursor can happen any time dynamically.

Unfortunately, it seems like Qemu doesn't expect this change to happen
dynamically. The code in virtio-gpu.c was written such that
update_cursor would first call dpy_cursor_define if the cursor shape has
changed and this is not a simple move operation (which indeed is the
case when moving to a SW cursor) and then call dpy_mouse_set.
dpy_cursor_define calls toolkits like GTK but in addition to creating
the cursor content, it also calls gdk_window_set_cursor thereby setting
the cursor. It is therefore, the right function to either show or hide
the cursor but there was no code present to hide the cursor. Changing
the cursor visibility in dpy_mouse_set has two issues. First,
dpy_cursor_define already decided to display the cursor so showing the
cursor in the previous call only to hide it in dpy_mouse_set doesn't
sound right. Second, dpy_mouse_set skips the rest of the code if we
are in absolute mode so adding this code there wouldn't work in that
mode.

Qemu makes the decision of whether to show or hide the cursor by
observing the cursor->resource_id flag in virtio-gpu.c as is evident
from the lines
dpy_mouse_set(s->con, cursor->pos.x, cursor->pos.y,
  cursor->resource_id ? 1 : 0);
The last parameter is considered the "visible" parameter in gdk code.
Therefore, in this patch we continue with the same model. Instead of
changing the function prototype and changing a dozen plus files, we are
adding the visible field in QEMUCursor data structure and passing
it from virtio-gpu to the GTK code where we check if the cursor is
visible or not. If not, we simply call gdk_window_set_cursor with NULL
parameter, thereby hiding the cursor. Once Windows VM switches back to
the HW cursor, then IDD would again provide a new resource_id to Qemu
and we can start displaying it once more.

Signed-off-by: Satyeshwar Singh 
mailto:satyeshwar.si...@intel.com>>
Cc: Marc-André Lureau 
mailto:marcandre.lur...@redhat.com>>
Cc: Vivek Kasireddy 
mailto:vivek.kasire...@intel.com>>
Cc: Dongwon Kim mailto:dongwon@intel.com>>

---
hw/display/virtio-gpu.c | 1 +
include/ui/console.h| 1 +
ui/gtk.c| 5 +
3 files changed, 7 insertions(+)

diff --git a/hw/display/virtio-gpu.c b/hw/display/virtio-gpu.c
index 1c1ee230b3..1ae9be605b 100644
--- a/hw/display/virtio-gpu.c
+++ b/hw/display/virtio-gpu.c
@@ -98,6 +98,7 @@ static void update_cursor(VirtIOGPU *g, struct 
virtio_gpu_update_cursor *cursor)
 s->current_cursor->hot_x = cursor->hot_x;
 s->current_cursor->hot_y = cursor->hot_y;
+s->current_cursor->visible = cursor->resource_id ? 1 : 0;
 if (cursor->resource_id > 0) {
 vgc->update_cursor_data(g, s, cursor->resource_id);
diff --git a/include/ui/console.h b/include/ui/console.h
index a4a49ffc64..47c39ed405 100644
--- a/include/ui/console.h
+++ b/include/ui/console.h
@@ -161,6 +161,7 @@ typedef struct QEMUCursor {
 uint16_twidth, height;
 int hot_x, hot_y;
 int refcount;
+int visible;
 uint32_tdata[];
} QEMUCursor;
diff --git a/ui/gtk.c b/ui/gtk.c
index 810d7fc796..12a76de570 100644
--- a/ui/gtk.c
+++ b/ui/gtk.c
@@ -478,6 +478,11 @@ static void gd_cursor_define(DisplayChangeListener *dcl,
 return;
 }
+if(!c->visible) {
+gdk_window_set_cursor(gtk_widget_get_window(vc->gfx.drawing_area), 
NULL);
+return;
+}
+
 pixbuf = gdk_pixbuf_new_from_data((guchar *)(c->data),
   GDK_COLORSPACE_RGB, true, 8,
   c->width, c->height, c->width * 4,
--
2.33.1

Re: [PULL 00/39] tcg and linux-user patch queue

2024-02-23 Thread Richard Henderson


On 2/23/24 03:45, Peter Maydell wrote:

bsd-user fails to compile:
https://gitlab.com/qemu-project/qemu/-/jobs/6241616724

../bsd-user/main.c:379:30: error: use of undeclared identifier 'arg';
did you mean 'argv'?
if (qemu_strtoui(arg, NULL, 10, ) || size != want) {
  ^~~


Grr.  I think it is An Error that make vm-build-freebsd does not test this.
Alex?


r~

Re: [RFC PATCH v3 21/21] hw/arm/virt: Add FEAT_GICv3_NMI feature support in virt GIC

2024-02-23 Thread Richard Henderson


On 2/23/24 00:32, Jinjie Ruan via wrote:

A PE that implements FEAT_NMI and FEAT_GICv3 also implements
FEAT_GICv3_NMI. A PE that does not implement FEAT_NMI, does not implement
FEAT_GICv3_NMI

So included support FEAT_GICv3_NMI feature as part of virt platform
GIC initialization if FEAT_NMI and FEAT_GICv3 supported.

Signed-off-by: Jinjie Ruan
---


Reviewed-by: Richard Henderson 

r~

Re: [RFC PATCH v3 19/21] hw/intc/arm_gicv3: Report the NMI interrupt in gicv3_cpuif_update()

2024-02-23 Thread Richard Henderson


On 2/23/24 00:32, Jinjie Ruan via wrote:

In CPU Interface, if the IRQ has the superpriority property, report
NMI to the corresponding PE.

Signed-off-by: Jinjie Ruan 
---
v3:
- Remove handling nmi_is_irq flag.
---
  hw/intc/arm_gicv3_cpuif.c | 6 +-
  1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/hw/intc/arm_gicv3_cpuif.c b/hw/intc/arm_gicv3_cpuif.c
index f5bf8df32b..3ffeb9543b 100644
--- a/hw/intc/arm_gicv3_cpuif.c
+++ b/hw/intc/arm_gicv3_cpuif.c
@@ -931,6 +931,7 @@ void gicv3_cpuif_update(GICv3CPUState *cs)
  /* Tell the CPU about its highest priority pending interrupt */
  int irqlevel = 0;
  int fiqlevel = 0;
+int nmilevel = 0;
  ARMCPU *cpu = ARM_CPU(cs->cpu);
  CPUARMState *env = >env;
  
@@ -967,7 +968,9 @@ void gicv3_cpuif_update(GICv3CPUState *cs)

  g_assert_not_reached();
  }
  
-if (isfiq) {

+if (cs->hppi.superprio) {
+nmilevel = 1;
+} else if (isfiq) {
  fiqlevel = 1;
  } else {
  irqlevel = 1;


NMI only applies to group 1, per Table 4-6.  Because group 0 always produces fiq, I think 
you need to swap the ordering of the IFs.



r~

Re: [PATCH 3/3] linux-user: Rewrite target_shmat

2024-02-23 Thread Richard Henderson


On 2/23/24 01:35, Ilya Leoshkevich wrote:

On Thu, Feb 22, 2024 at 05:03:09PM -1000, Richard Henderson wrote:

Handle combined host and guest alignment requirements.
Handle host and guest page size differences.
Handle SHM_EXEC.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/115
Signed-off-by: Richard Henderson 
---
  linux-user/mmap.c | 146 ++
  1 file changed, 110 insertions(+), 36 deletions(-)


[...]


-/* find out the length of the shared memory segment */
+/*
+ * Because we can't use host shmat() unless the address is sufficiently
+ * aligned for the host, we'll need to check both.
+ * TODO: Could be fixed with softmmu.
+ */


Are there any plans to introduce softmmu to qemu-user?


Long-term ones, yes.  There are *so* many places for which emulation doesn't quite work 
unless the host and guest page size match.  There are projects like asan which don't work 
unless the guest has a *very* specific memory layout, which often conflicts with the host.




r~

Re: [RFC PATCH v3 00/21] target/arm: Implement FEAT_NMI and FEAT_GICv3_NMI

2024-02-23 Thread Richard Henderson

On 2/23/24 00:32, Jinjie Ruan via wrote:

This patch set implements FEAT_NMI and FEAT_GICv3_NMI for armv8. These
introduce support for a new category of interrupts in the architecture
which we can use to provide NMI like functionality.

There are two modes for using this FEAT_NMI. When PSTATE.ALLINT or
PSTATE.SP & SCTLR_ELx.SCTLR_SPINTMASK is set, any entry to ELx causes all
interrupts including those with superpriority to be masked on entry to ELn
until the mask is explicitly removed by software or hardware. PSTATE.ALLINT
can be managed by software using the new register control ALLINT.ALLINT.
Independent controls are provided for this feature at each EL, usage at EL1
should not disrupt EL2 or EL3.

I have tested it with the following linux patches which try to support
FEAT_NMI in linux kernel:

https://lore.kernel.org/linux-arm-kernel/Y4sH5qX5bK9xfEBp@lpieralisi/T/#mb4ba4a2c045bf72c10c2202c1dd1b82d3240dc88

In the test, SGI, PPI and SPI interrupts can all be set to have super priority
to be converted to a hardware NMI interrupt. The SGI is tested with kernel
IPI as NMI framework, and the PPI interrupt is tested with "perf top" command
with hardware NMI enabled, and the PPI interrupt is tested with a custom
test module, in which NMI interrupts can be received and transmitted normally.

+-+
| Distributor |
+-+
SPI | NMI | NMI
\ /\ /
++ +---+
| Redist | | Redist|
++ +---+
SGI | NMI PPI | NMI
\ / \ /
+-+ +---+
|CPU Interface| ... | CPU Interface |
+-+ +---+
| NMI | NMI
\ / \ /
+-+ +-+
| PE | | PE |
+-+ +-+

Changes in v3:
- Remove the FIQ NMI.
- Adjust the patches Sequence.
- Reomve the patch "Set pstate.ALLINT in arm_cpu_reset_hold".
- Check whether support FEAT_NMI and FEAT_GICv3 for FEAT_GICv3_NMI.
- With CPU_INTERRUPT_NMI, both CPSR_I and ISR_IS must be set.
- Not include NMI logic when FEAT_NMI or SCTLR_ELx.NMI not enabled.
- Refator nmi mask in arm_excp_unmasked().
- Add VNMI definitions, add HCRX_VINMI and HCRX_VFNMI support in HCRX_EL2.
- Add Reviewed-by and Acked-by.
- Update the commit message.

1 2 3 >

1 - 100 of 290 matches

Mail list logo