[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Minios-devel] [UNIKRAFT PATCH] plat/kvm: Fix current thread retrieval in interrupt context on x86_64
Hey Costin,
thanks a lot for the patch. I have a couple of questions which I need
for my understanding. I put those inline...
Thanks,
Simon
On 11.06.19 21:33, Costin Lupu wrote:
Commit 017fffd5 introduced support for setting the current thread pointer on top
of interrupt stacks in order to retrieve the current thread in interrupt context
as well. Unfortunately, the wrong stack was picked for KVM platform. This patch
fixes that and sets the thread on cpu_intr_stack instead.
cpu_intr_stack was resized to STACK_SIZE because this is a mandatory condition
when saving threads on top. However, given that it also needs a STACK_SIZE
alignment, a new section was created for it, .intrstack, in order to avoid
breaking the entire binary image layout. Without this new section, the entire
.text section would have a STACK_SIZE alignment (i.e. 64KB) and this would imply
that the multiboot header, which is included in .text section, would also be
moved at an address higher than STACK_SIZE, even though it must stay in the
first 8KB of the binary.
Maybe, this is a stupid question: What if we take the pre-allocated
bootstack also for the interrupts? As soon as we enable scheduling the
bootstack is not used anymore by any thread. Except the case where we do
not have scheduling, the bootstack will be still used during life time,
would this be an issue?
Signed-off-by: Costin Lupu <costin.lupu@xxxxxxxxx>
---
plat/kvm/Makefile.uk | 1 +
plat/kvm/memory.c | 6 ------
plat/kvm/x86/link64.lds.S | 10 ++++++++++
plat/kvm/x86/memory.c | 44 ++++++++++++++++++++++++++++++++++++++++++++
plat/kvm/x86/traps.c | 5 ++++-
5 files changed, 59 insertions(+), 7 deletions(-)
create mode 100644 plat/kvm/x86/memory.c
diff --git a/plat/kvm/Makefile.uk b/plat/kvm/Makefile.uk
index 71c4c419..8eb162d4 100644
--- a/plat/kvm/Makefile.uk
+++ b/plat/kvm/Makefile.uk
@@ -46,6 +46,7 @@ LIBKVMPLAT_SRCS-$(CONFIG_ARCH_X86_64) +=
$(LIBKVMPLAT_BASE)/x86/lcpu.c
LIBKVMPLAT_SRCS-$(CONFIG_ARCH_X86_64) += $(LIBKVMPLAT_BASE)/x86/intctrl.c
LIBKVMPLAT_SRCS-$(CONFIG_ARCH_X86_64) += $(LIBKVMPLAT_BASE)/x86/tscclock.c
LIBKVMPLAT_SRCS-$(CONFIG_ARCH_X86_64) += $(LIBKVMPLAT_BASE)/x86/time.c
+LIBKVMPLAT_SRCS-$(CONFIG_ARCH_X86_64) += $(LIBKVMPLAT_BASE)/x86/memory.c|x86
ifeq ($(findstring y,$(CONFIG_KVM_KERNEL_VGA_CONSOLE)
$(CONFIG_KVM_DEBUG_VGA_CONSOLE)),y)
LIBKVMPLAT_SRCS-$(CONFIG_ARCH_X86_64) += $(LIBKVMPLAT_BASE)/x86/vga_console.c
endif
diff --git a/plat/kvm/memory.c b/plat/kvm/memory.c
index e96bad2c..7f2fb46a 100644
--- a/plat/kvm/memory.c
+++ b/plat/kvm/memory.c
@@ -178,9 +178,3 @@ int _ukplat_mem_mappings_init(void)
{
return 0;
}
-
-void ukplat_stack_set_current_thread(void *thread_addr)
-{
- *((unsigned long *) _libkvmplat_cfg.bstack.end) =
- (unsigned long) thread_addr;
-}
diff --git a/plat/kvm/x86/link64.lds.S b/plat/kvm/x86/link64.lds.S
index 362ba3e6..6103fc2d 100644
--- a/plat/kvm/x86/link64.lds.S
+++ b/plat/kvm/x86/link64.lds.S
@@ -99,6 +99,16 @@ SECTIONS
. = ALIGN(__PAGE_SIZE);
}
+ /* We keep the interrupt stack on a different section
+ * given that it may have a big alignment and it would
+ * change the entire binary layout
+ */
+ .intrstack :
+ {
+ *(.intrstack)
+ . = ALIGN(__PAGE_SIZE);
+ }
+
Would every platform need to do this?
_end = .;
.comment 0 : { *(.comment) }
diff --git a/plat/kvm/x86/memory.c b/plat/kvm/x86/memory.c
new file mode 100644
index 00000000..b8c7c7e7
--- /dev/null
+++ b/plat/kvm/x86/memory.c
@@ -0,0 +1,44 @@
+/* SPDX-License-Identifier: BSD-3-Clause */
+/*
+ * Authors: Costin Lupu <costin.lupu@xxxxxxxxx>
+ *
+ * Copyright (c) 2019, University Politehnica of Bucharest. All rights
reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ * 3. Neither the name of the copyright holder nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ *
+ * THIS HEADER MAY NOT BE EXTRACTED OR MODIFIED IN ANY WAY.
+ */
+
+#include <uk/plat/memory.h>
+
+
+extern char cpu_intr_stack[];
+
+void ukplat_stack_set_current_thread(void *thread_addr)
+{
+ *((unsigned long *) cpu_intr_stack) =
+ (unsigned long) thread_addr;
In general, instead of using the stack top, we could save the current
thread always on the stack bottom. Those bytes would be reserved and
would only be popped on a misbehaving program. Would that avoid the
issue that you are facing? Maybe we could get away from the problem that
the stacks have to have a fixed size for just getting the current value...
+}
diff --git a/plat/kvm/x86/traps.c b/plat/kvm/x86/traps.c
index 27ef6d93..fe1dd5a4 100644
--- a/plat/kvm/x86/traps.c
+++ b/plat/kvm/x86/traps.c
@@ -25,7 +25,9 @@
*/
#include <string.h>
+#include <uk/essentials.h>
#include <uk/arch/lcpu.h>
+#include <uk/plat/config.h>
#include <x86/desc.h>
#include <kvm-x86/traps.h>
@@ -59,7 +61,8 @@ static void gdt_init(void)
static struct tss64 cpu_tss;
-static char cpu_intr_stack[4096]; /* IST1 */
+ __section(".intrstack") __align(STACK_SIZE)
+char cpu_intr_stack[STACK_SIZE]; /* IST1 */
You removed the stack actually from the .bss section (instead of text)
and moved it to an own section. What if you keep it on the .bss but with
the bigger size? I expect this should not be a problem for the text section.
static char cpu_trap_stack[4096]; /* IST2 */
static char cpu_nmi_stack[4096]; /* IST3 */
_______________________________________________
Minios-devel mailing list
Minios-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/minios-devel
|
