* [PATCH v10 00/12] Arm cache coloring
@ 2024-11-19 14:13 Carlo Nonato
2024-11-19 14:13 ` [PATCH v10 01/12] xen/common: add cache coloring common code Carlo Nonato
` (11 more replies)
0 siblings, 12 replies; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Andrew Cooper,
Jan Beulich, Julien Grall, Stefano Stabellini, Bertrand Marquis,
Michal Orzel, Volodymyr Babchuk, Anthony PERARD, Juergen Gross
Shared caches in multi-core CPU architectures represent a problem for
predictability of memory access latency. This jeopardizes applicability
of many Arm platform in real-time critical and mixed-criticality
scenarios. We introduce support for cache partitioning with page
coloring, a transparent software technique that enables isolation
between domains and Xen, and thus avoids cache interference.
When creating a domain, a simple syntax (e.g. `0-3` or `4-11`) allows
the user to define assignments of cache partitions ids, called colors,
where assigning different colors guarantees no mutual eviction on cache
will ever happen. This instructs the Xen memory allocator to provide
the i-th color assignee only with pages that maps to color i, i.e. that
are indexed in the i-th cache partition.
The proposed implementation supports the dom0less feature.
The proposed implementation doesn't support the static-mem feature.
The solution has been tested in several scenarios, including Xilinx Zynq
MPSoCs.
Carlo Nonato (11):
xen/common: add cache coloring common code
xen/arm: add initial support for LLC coloring on arm64
xen/arm: permit non direct-mapped Dom0 construction
xen/arm: add Dom0 cache coloring support
xen: extend domctl interface for cache coloring
tools: add support for cache coloring configuration
xen/arm: add support for cache coloring configuration via device-tree
xen/page_alloc: introduce preserved page flags macro
xen: add cache coloring allocator for domains
xen/arm: make consider_modules() available for xen relocation
xen/arm: add cache coloring support for Xen
Luca Miccio (1):
xen/arm: add Xen cache colors command line parameter
SUPPORT.md | 7 +
docs/index.rst | 1 +
docs/man/xl.cfg.5.pod.in | 6 +
docs/misc/arm/device-tree/booting.txt | 4 +
docs/misc/cache-coloring.rst | 246 +++++++++++++++
docs/misc/xen-command-line.pandoc | 70 +++++
tools/include/libxl.h | 5 +
tools/include/xenctrl.h | 9 +
tools/libs/ctrl/xc_domain.c | 35 +++
tools/libs/light/libxl_create.c | 18 ++
tools/libs/light/libxl_types.idl | 1 +
tools/xl/xl_parse.c | 38 ++-
xen/arch/arm/Kconfig | 1 +
xen/arch/arm/Makefile | 1 +
xen/arch/arm/alternative.c | 30 +-
xen/arch/arm/arm32/mmu/mm.c | 95 +-----
xen/arch/arm/arm64/mmu/head.S | 58 +++-
xen/arch/arm/arm64/mmu/mm.c | 28 +-
xen/arch/arm/dom0less-build.c | 60 +---
xen/arch/arm/domain_build.c | 106 ++++++-
xen/arch/arm/include/asm/domain_build.h | 1 +
xen/arch/arm/include/asm/mm.h | 5 +
xen/arch/arm/include/asm/mmu/layout.h | 3 +
xen/arch/arm/include/asm/processor.h | 16 +
xen/arch/arm/include/asm/setup.h | 3 +
xen/arch/arm/llc-coloring.c | 141 +++++++++
xen/arch/arm/mmu/setup.c | 195 +++++++++++-
xen/arch/arm/setup.c | 13 +-
xen/common/Kconfig | 29 ++
xen/common/Makefile | 1 +
xen/common/domain.c | 3 +
xen/common/domctl.c | 10 +
xen/common/keyhandler.c | 3 +
xen/common/llc-coloring.c | 382 ++++++++++++++++++++++++
xen/common/page_alloc.c | 208 ++++++++++++-
xen/include/public/domctl.h | 9 +
xen/include/xen/llc-coloring.h | 64 ++++
xen/include/xen/sched.h | 5 +
xen/include/xen/xmalloc.h | 12 +
39 files changed, 1749 insertions(+), 173 deletions(-)
create mode 100644 docs/misc/cache-coloring.rst
create mode 100644 xen/arch/arm/llc-coloring.c
create mode 100644 xen/common/llc-coloring.c
create mode 100644 xen/include/xen/llc-coloring.h
--
2.43.0
^ permalink raw reply [flat|nested] 36+ messages in thread
* [PATCH v10 01/12] xen/common: add cache coloring common code
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-11-27 10:48 ` Michal Orzel
2024-11-28 10:39 ` Jan Beulich
2024-11-19 14:13 ` [PATCH v10 02/12] xen/arm: add initial support for LLC coloring on arm64 Carlo Nonato
` (10 subsequent siblings)
11 siblings, 2 replies; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Andrew Cooper,
Jan Beulich, Julien Grall, Stefano Stabellini
Last Level Cache (LLC) coloring allows to partition the cache in smaller
chunks called cache colors.
Since not all architectures can actually implement it, add a HAS_LLC_COLORING
Kconfig option.
LLC_COLORS_ORDER Kconfig option has a range maximum of 10 (2^10 = 1024)
because that's the number of colors that fit in a 4 KiB page when integers
are 4 bytes long.
LLC colors are a property of the domain, so struct domain has to be extended.
Based on original work from: Luca Miccio <lucmiccio@gmail.com>
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
---
v10:
- fixed commit message to use LLC_COLORS_ORDER
- added documentation to index.rst
- moved check on CONFIG_NUMA in arch/arm/Kconfig (next patch)
- fixed copyright line
- fixed array type for colors parameter in print_colors()
- added check on (way_size & ~PAGE_MASK)
v9:
- dropped _MAX_ from CONFIG_MAX_LLC_COLORS_ORDER
v8:
- minor documentation fixes
- "llc-coloring=on" is inferred from "llc-nr-ways" and "llc-size" usage
- turned CONFIG_NR_LLC_COLORS to CONFIG_MAX_LLC_COLORS_ORDER, base-2 exponent
- moved Kconfig options to common/Kconfig
- don't crash if computed max_nr_colors is too large
v7:
- SUPPORT.md changes added to this patch
- extended documentation to better address applicability of cache coloring
- "llc-nr-ways" and "llc-size" params introduced in favor of "llc-way-size"
- moved dump_llc_coloring_info() call in 'm' keyhandler (pagealloc_info())
v6:
- moved almost all code in common
- moved documentation in this patch
- reintroduced range for CONFIG_NR_LLC_COLORS
- reintroduced some stub functions to reduce the number of checks on
llc_coloring_enabled
- moved domain_llc_coloring_free() in same patch where allocation happens
- turned "d->llc_colors" to pointer-to-const
- llc_coloring_init() now returns void and panics if errors are found
v5:
- used - instead of _ for filenames
- removed domain_create_llc_colored()
- removed stub functions
- coloring domain fields are now #ifdef protected
v4:
- Kconfig options moved to xen/arch
- removed range for CONFIG_NR_LLC_COLORS
- added "llc_coloring_enabled" global to later implement the boot-time
switch
- added domain_create_llc_colored() to be able to pass colors
- added is_domain_llc_colored() macro
---
SUPPORT.md | 7 ++
docs/index.rst | 1 +
docs/misc/cache-coloring.rst | 116 ++++++++++++++++++++++++++++++
docs/misc/xen-command-line.pandoc | 37 ++++++++++
xen/common/Kconfig | 21 ++++++
xen/common/Makefile | 1 +
xen/common/keyhandler.c | 3 +
xen/common/llc-coloring.c | 115 +++++++++++++++++++++++++++++
xen/common/page_alloc.c | 3 +
xen/include/xen/llc-coloring.h | 37 ++++++++++
xen/include/xen/sched.h | 5 ++
11 files changed, 346 insertions(+)
create mode 100644 docs/misc/cache-coloring.rst
create mode 100644 xen/common/llc-coloring.c
create mode 100644 xen/include/xen/llc-coloring.h
diff --git a/SUPPORT.md b/SUPPORT.md
index 82239d0294..998faf5635 100644
--- a/SUPPORT.md
+++ b/SUPPORT.md
@@ -401,6 +401,13 @@ by maintaining multiple physical to machine (p2m) memory mappings.
Status, x86 HVM: Tech Preview
Status, ARM: Tech Preview
+### Cache coloring
+
+Allows to reserve Last Level Cache (LLC) partitions for Dom0, DomUs and Xen
+itself.
+
+ Status, Arm64: Experimental
+
## Resource Management
### CPU Pools
diff --git a/docs/index.rst b/docs/index.rst
index ab051a0f38..fad4462bbb 100644
--- a/docs/index.rst
+++ b/docs/index.rst
@@ -65,6 +65,7 @@ Documents in need of some rearranging.
designs/launch/hyperlaunch-devicetree
misc/xen-makefiles/makefiles
misra/index
+ misc/cache-coloring
Miscellanea
diff --git a/docs/misc/cache-coloring.rst b/docs/misc/cache-coloring.rst
new file mode 100644
index 0000000000..0fe3830c40
--- /dev/null
+++ b/docs/misc/cache-coloring.rst
@@ -0,0 +1,116 @@
+Xen cache coloring user guide
+=============================
+
+The cache coloring support in Xen allows to reserve Last Level Cache (LLC)
+partitions for Dom0, DomUs and Xen itself. Currently only ARM64 is supported.
+Cache coloring realizes per-set cache partitioning in software and is applicable
+to shared LLCs as implemented in Cortex-A53, Cortex-A72 and similar CPUs.
+
+To compile LLC coloring support set ``CONFIG_LLC_COLORING=y``.
+
+If needed, change the maximum number of colors with
+``CONFIG_LLC_COLORS_ORDER=<n>``.
+
+Runtime configuration is done via `Command line parameters`_.
+
+Background
+**********
+
+Cache hierarchy of a modern multi-core CPU typically has first levels dedicated
+to each core (hence using multiple cache units), while the last level is shared
+among all of them. Such configuration implies that memory operations on one
+core (e.g. running a DomU) are able to generate interference on another core
+(e.g. hosting another DomU). Cache coloring realizes per-set cache-partitioning
+in software and mitigates this, guaranteeing more predictable performances for
+memory accesses.
+Software-based cache coloring is particularly useful in those situations where
+no hardware mechanisms (e.g., DSU-based way partitioning) are available to
+partition caches. This is the case for e.g., Cortex-A53, A57 and A72 CPUs that
+feature a L2 LLC cache shared among all cores.
+
+The key concept underlying cache coloring is a fragmentation of the memory
+space into a set of sub-spaces called colors that are mapped to disjoint cache
+partitions. Technically, the whole memory space is first divided into a number
+of subsequent regions. Then each region is in turn divided into a number of
+subsequent sub-colors. The generic i-th color is then obtained by all the
+i-th sub-colors in each region.
+
+::
+
+ Region j Region j+1
+ ..................... ............
+ . . .
+ . .
+ _ _ _______________ _ _____________________ _ _
+ | | | | | | |
+ | c_0 | c_1 | | c_n | c_0 | c_1 |
+ _ _ _|_____|_____|_ _ _|_____|_____|_____|_ _ _
+ : :
+ : :... ... .
+ : color 0
+ :........................... ... .
+ :
+ . . ..................................:
+
+How colors are actually defined depends on the function that maps memory to
+cache lines. In case of physically-indexed, physically-tagged caches with linear
+mapping, the set index is found by extracting some contiguous bits from the
+physical address. This allows colors to be defined as shown in figure: they
+appear in memory as subsequent blocks of equal size and repeats themselves after
+``n`` different colors, where ``n`` is the total number of colors.
+
+If some kind of bit shuffling appears in the mapping function, then colors
+assume a different layout in memory. Those kind of caches aren't supported by
+the current implementation.
+
+**Note**: Finding the exact cache mapping function can be a really difficult
+task since it's not always documented in the CPU manual. As said Cortex-A53, A57
+and A72 are known to work with the current implementation.
+
+How to compute the number of colors
+###################################
+
+Given the linear mapping from physical memory to cache lines for granted, the
+number of available colors for a specific platform is computed using three
+parameters:
+
+- the size of the LLC.
+- the number of the LLC ways.
+- the page size used by Xen.
+
+The first two parameters can be found in the processor manual, while the third
+one is the minimum mapping granularity. Dividing the cache size by the number of
+its ways we obtain the size of a way. Dividing this number by the page size,
+the number of total cache colors is found. So for example an Arm Cortex-A53
+with a 16-ways associative 1 MiB LLC can isolate up to 16 colors when pages are
+4 KiB in size.
+
+Effective colors assignment
+###########################
+
+When assigning colors, if one wants to avoid cache interference between two
+domains, different colors needs to be used for their memory.
+
+Command line parameters
+***********************
+
+Specific documentation is available at `docs/misc/xen-command-line.pandoc`.
+
++----------------------+-------------------------------+
+| **Parameter** | **Description** |
++----------------------+-------------------------------+
+| ``llc-coloring`` | Enable coloring at runtime |
++----------------------+-------------------------------+
+| ``llc-size`` | Set the LLC size |
++----------------------+-------------------------------+
+| ``llc-nr-ways`` | Set the LLC number of ways |
++----------------------+-------------------------------+
+
+Auto-probing of LLC specs
+#########################
+
+LLC size and number of ways are probed automatically by default.
+
+LLC specs can be manually set via the above command line parameters. This
+bypasses any auto-probing and it's used to overcome failing situations, such as
+flawed probing logic, or for debugging/testing purposes.
diff --git a/docs/misc/xen-command-line.pandoc b/docs/misc/xen-command-line.pandoc
index 293dbc1a95..abd8dae96f 100644
--- a/docs/misc/xen-command-line.pandoc
+++ b/docs/misc/xen-command-line.pandoc
@@ -1708,6 +1708,43 @@ This option is intended for debugging purposes only. Enable MSR_DEBUGCTL.LBR
in hypervisor context to be able to dump the Last Interrupt/Exception To/From
record with other registers.
+### llc-coloring (arm64)
+> `= <boolean>`
+
+> Default: `false`
+
+Flag to enable or disable LLC coloring support at runtime. This option is
+available only when `CONFIG_LLC_COLORING` is enabled. See the general
+cache coloring documentation for more info.
+
+### llc-nr-ways (arm64)
+> `= <integer>`
+
+> Default: `Obtained from hardware`
+
+Specify the number of ways of the Last Level Cache. This option is available
+only when `CONFIG_LLC_COLORING` is enabled. LLC size and number of ways are used
+to find the number of supported cache colors. By default the value is
+automatically computed by probing the hardware, but in case of specific needs,
+it can be manually set. Those include failing probing and debugging/testing
+purposes so that it's possible to emulate platforms with different number of
+supported colors. If set, also "llc-size" must be set, otherwise the default
+will be used. Note that using both options implies "llc-coloring=on".
+
+### llc-size (arm64)
+> `= <size>`
+
+> Default: `Obtained from hardware`
+
+Specify the size of the Last Level Cache. This option is available only when
+`CONFIG_LLC_COLORING` is enabled. LLC size and number of ways are used to find
+the number of supported cache colors. By default the value is automatically
+computed by probing the hardware, but in case of specific needs, it can be
+manually set. Those include failing probing and debugging/testing purposes so
+that it's possible to emulate platforms with different number of supported
+colors. If set, also "llc-nr-ways" must be set, otherwise the default will be
+used. Note that using both options implies "llc-coloring=on".
+
### lock-depth-size
> `= <integer>`
diff --git a/xen/common/Kconfig b/xen/common/Kconfig
index 90268d9249..b4ec6893be 100644
--- a/xen/common/Kconfig
+++ b/xen/common/Kconfig
@@ -71,6 +71,9 @@ config HAS_IOPORTS
config HAS_KEXEC
bool
+config HAS_LLC_COLORING
+ bool
+
config HAS_PIRQ
bool
@@ -516,4 +519,22 @@ config TRACEBUFFER
to be collected at run time for debugging or performance analysis.
Memory and execution overhead when not active is minimal.
+config LLC_COLORING
+ bool "Last Level Cache (LLC) coloring" if EXPERT
+ depends on HAS_LLC_COLORING
+
+config LLC_COLORS_ORDER
+ int "Maximum number of LLC colors (base-2 exponent)"
+ range 1 10
+ default 7
+ depends on LLC_COLORING
+ help
+ Controls the build-time size of various arrays associated with LLC
+ coloring. The value is a base-2 exponent. Refer to cache coloring
+ documentation for how to compute the number of colors supported by the
+ platform. This is only an upper bound. The runtime value is autocomputed
+ or manually set via cmdline parameters.
+ The default value corresponds to an 8 MiB 16-ways LLC, which should be
+ more than what's needed in the general case.
+
endmenu
diff --git a/xen/common/Makefile b/xen/common/Makefile
index b279b09bfb..cba3b32733 100644
--- a/xen/common/Makefile
+++ b/xen/common/Makefile
@@ -25,6 +25,7 @@ obj-y += keyhandler.o
obj-$(CONFIG_KEXEC) += kexec.o
obj-$(CONFIG_KEXEC) += kimage.o
obj-$(CONFIG_LIVEPATCH) += livepatch.o livepatch_elf.o
+obj-$(CONFIG_LLC_COLORING) += llc-coloring.o
obj-$(CONFIG_MEM_ACCESS) += mem_access.o
obj-y += memory.o
obj-y += multicall.o
diff --git a/xen/common/keyhandler.c b/xen/common/keyhandler.c
index 6da291b34e..6ea54838d4 100644
--- a/xen/common/keyhandler.c
+++ b/xen/common/keyhandler.c
@@ -5,6 +5,7 @@
#include <asm/regs.h>
#include <xen/delay.h>
#include <xen/keyhandler.h>
+#include <xen/llc-coloring.h>
#include <xen/param.h>
#include <xen/sections.h>
#include <xen/shutdown.h>
@@ -304,6 +305,8 @@ static void cf_check dump_domains(unsigned char key)
arch_dump_domain_info(d);
+ domain_dump_llc_colors(d);
+
rangeset_domain_printk(d);
dump_pageframe_info(d);
diff --git a/xen/common/llc-coloring.c b/xen/common/llc-coloring.c
new file mode 100644
index 0000000000..29b75e0e0d
--- /dev/null
+++ b/xen/common/llc-coloring.c
@@ -0,0 +1,115 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Last Level Cache (LLC) coloring common code
+ *
+ * Copyright (C) 2024, Advanced Micro Devices, Inc.
+ * Copyright (C) 2024, Minerva Systems SRL
+ */
+#include <xen/keyhandler.h>
+#include <xen/llc-coloring.h>
+#include <xen/param.h>
+
+#define NR_LLC_COLORS (1U << CONFIG_LLC_COLORS_ORDER)
+
+static bool __ro_after_init llc_coloring_enabled;
+boolean_param("llc-coloring", llc_coloring_enabled);
+
+static unsigned int __initdata llc_size;
+size_param("llc-size", llc_size);
+static unsigned int __initdata llc_nr_ways;
+integer_param("llc-nr-ways", llc_nr_ways);
+/* Number of colors available in the LLC */
+static unsigned int __ro_after_init max_nr_colors;
+
+static void print_colors(const unsigned int colors[], unsigned int num_colors)
+{
+ unsigned int i;
+
+ printk("{ ");
+ for ( i = 0; i < num_colors; i++ )
+ {
+ unsigned int start = colors[i], end = start;
+
+ printk("%u", start);
+
+ for ( ; i < num_colors - 1 && end + 1 == colors[i + 1]; i++, end++ )
+ ;
+
+ if ( start != end )
+ printk("-%u", end);
+
+ if ( i < num_colors - 1 )
+ printk(", ");
+ }
+ printk(" }\n");
+}
+
+void __init llc_coloring_init(void)
+{
+ unsigned int way_size;
+
+ if ( llc_size && llc_nr_ways )
+ {
+ llc_coloring_enabled = true;
+ way_size = llc_size / llc_nr_ways;
+ }
+ else if ( !llc_coloring_enabled )
+ return;
+ else
+ {
+ way_size = get_llc_way_size();
+ if ( !way_size )
+ panic("LLC probing failed and 'llc-size' or 'llc-nr-ways' missing\n");
+ }
+
+ if ( way_size & ~PAGE_MASK )
+ panic("LLC way size must be a multiple of PAGE_SIZE\n");
+
+ /*
+ * The maximum number of colors must be a power of 2 in order to correctly
+ * map them to bits of an address.
+ */
+ max_nr_colors = way_size >> PAGE_SHIFT;
+
+ if ( max_nr_colors & (max_nr_colors - 1) )
+ panic("Number of LLC colors (%u) isn't a power of 2\n", max_nr_colors);
+
+ if ( max_nr_colors > NR_LLC_COLORS )
+ {
+ printk(XENLOG_WARNING
+ "Number of LLC colors (%u) too big. Using configured max %u\n",
+ max_nr_colors, NR_LLC_COLORS);
+ max_nr_colors = NR_LLC_COLORS;
+ } else if ( max_nr_colors < 2 )
+ panic("Number of LLC colors %u < 2\n", max_nr_colors);
+
+ arch_llc_coloring_init();
+}
+
+void dump_llc_coloring_info(void)
+{
+ if ( !llc_coloring_enabled )
+ return;
+
+ printk("LLC coloring info:\n");
+ printk(" Number of LLC colors supported: %u\n", max_nr_colors);
+}
+
+void domain_dump_llc_colors(const struct domain *d)
+{
+ if ( !llc_coloring_enabled )
+ return;
+
+ printk("%u LLC colors: ", d->num_llc_colors);
+ print_colors(d->llc_colors, d->num_llc_colors);
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
index 33c8c917d9..7b911b5ed9 100644
--- a/xen/common/page_alloc.c
+++ b/xen/common/page_alloc.c
@@ -126,6 +126,7 @@
#include <xen/irq.h>
#include <xen/keyhandler.h>
#include <xen/lib.h>
+#include <xen/llc-coloring.h>
#include <xen/mm.h>
#include <xen/nodemask.h>
#include <xen/numa.h>
@@ -2626,6 +2627,8 @@ static void cf_check pagealloc_info(unsigned char key)
}
printk(" Dom heap: %lukB free\n", total << (PAGE_SHIFT-10));
+
+ dump_llc_coloring_info();
}
static __init int cf_check pagealloc_keyhandler_init(void)
diff --git a/xen/include/xen/llc-coloring.h b/xen/include/xen/llc-coloring.h
new file mode 100644
index 0000000000..6ccf4613c2
--- /dev/null
+++ b/xen/include/xen/llc-coloring.h
@@ -0,0 +1,37 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Last Level Cache (LLC) coloring common header
+ *
+ * Copyright (C) 2024, Advanced Micro Devices, Inc.
+ * Copyright (C) 2024, Minerva Systems SRL
+ */
+#ifndef __COLORING_H__
+#define __COLORING_H__
+
+#include <xen/sched.h>
+#include <public/domctl.h>
+
+#ifdef CONFIG_LLC_COLORING
+void llc_coloring_init(void);
+void dump_llc_coloring_info(void);
+void domain_dump_llc_colors(const struct domain *d);
+#else
+static inline void llc_coloring_init(void) {}
+static inline void dump_llc_coloring_info(void) {}
+static inline void domain_dump_llc_colors(const struct domain *d) {}
+#endif
+
+unsigned int get_llc_way_size(void);
+void arch_llc_coloring_init(void);
+
+#endif /* __COLORING_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/xen/sched.h b/xen/include/xen/sched.h
index 90666576c2..c0e49cd1e7 100644
--- a/xen/include/xen/sched.h
+++ b/xen/include/xen/sched.h
@@ -637,6 +637,11 @@ struct domain
/* Holding CDF_* constant. Internal flags for domain creation. */
unsigned int cdf;
+
+#ifdef CONFIG_LLC_COLORING
+ unsigned int num_llc_colors;
+ const unsigned int *llc_colors;
+#endif
};
static inline struct page_list_head *page_to_list(
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* [PATCH v10 02/12] xen/arm: add initial support for LLC coloring on arm64
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
2024-11-19 14:13 ` [PATCH v10 01/12] xen/common: add cache coloring common code Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-11-27 14:32 ` Michal Orzel
2024-11-28 10:42 ` Jan Beulich
2024-11-19 14:13 ` [PATCH v10 03/12] xen/arm: permit non direct-mapped Dom0 construction Carlo Nonato
` (9 subsequent siblings)
11 siblings, 2 replies; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Andrew Cooper,
Jan Beulich, Julien Grall, Stefano Stabellini, Bertrand Marquis,
Michal Orzel, Volodymyr Babchuk
LLC coloring needs to know the last level cache layout in order to make the
best use of it. This can be probed by inspecting the CLIDR_EL1 register,
so the Last Level is defined as the last level visible by this register.
Note that this excludes system caches in some platforms.
Static memory allocation and cache coloring are incompatible because static
memory can't be guaranteed to use only colors assigned to the domain.
Panic during DomUs creation when both are enabled.
Based on original work from: Luca Miccio <lucmiccio@gmail.com>
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
---
v10:
- moved CONFIG_NUMA check in arch/arm/Kconfig
v9:
- no changes
v8:
- no changes
v7:
- only minor changes
v6:
- get_llc_way_size() now checks for at least separate I/D caches
v5:
- used - instead of _ for filenames
- moved static-mem check in this patch
- moved dom0 colors parsing in next patch
- moved color allocation and configuration in next patch
- moved check_colors() in next patch
- colors are now printed in short form
v4:
- added "llc-coloring" cmdline option for the boot-time switch
- dom0 colors are now checked during domain init as for any other domain
- fixed processor.h masks bit width
- check for overflow in parse_color_config()
- check_colors() now checks also that colors are sorted and unique
---
docs/misc/cache-coloring.rst | 14 +++++
xen/arch/arm/Kconfig | 1 +
xen/arch/arm/Makefile | 1 +
xen/arch/arm/dom0less-build.c | 6 +++
xen/arch/arm/include/asm/processor.h | 16 ++++++
xen/arch/arm/llc-coloring.c | 78 ++++++++++++++++++++++++++++
xen/arch/arm/setup.c | 3 ++
xen/common/llc-coloring.c | 2 +-
xen/include/xen/llc-coloring.h | 4 ++
9 files changed, 124 insertions(+), 1 deletion(-)
create mode 100644 xen/arch/arm/llc-coloring.c
diff --git a/docs/misc/cache-coloring.rst b/docs/misc/cache-coloring.rst
index 0fe3830c40..b608284e9b 100644
--- a/docs/misc/cache-coloring.rst
+++ b/docs/misc/cache-coloring.rst
@@ -111,6 +111,20 @@ Auto-probing of LLC specs
LLC size and number of ways are probed automatically by default.
+In the Arm implementation, this is done by inspecting the CLIDR_EL1 register.
+This means that other system caches that aren't visible there are ignored.
+
LLC specs can be manually set via the above command line parameters. This
bypasses any auto-probing and it's used to overcome failing situations, such as
flawed probing logic, or for debugging/testing purposes.
+
+Known issues and limitations
+****************************
+
+"xen,static-mem" isn't supported when coloring is enabled
+#########################################################
+
+In the domain configuration, "xen,static-mem" allows memory to be statically
+allocated to the domain. This isn't possible when LLC coloring is enabled,
+because that memory can't be guaranteed to use only colors assigned to the
+domain.
diff --git a/xen/arch/arm/Kconfig b/xen/arch/arm/Kconfig
index 15b2e4a227..7f0712e674 100644
--- a/xen/arch/arm/Kconfig
+++ b/xen/arch/arm/Kconfig
@@ -8,6 +8,7 @@ config ARM_64
depends on !ARM_32
select 64BIT
select HAS_FAST_MULTIPLY
+ select HAS_LLC_COLORING if !NUMA
config ARM
def_bool y
diff --git a/xen/arch/arm/Makefile b/xen/arch/arm/Makefile
index e4ad1ce851..ccbfc61f88 100644
--- a/xen/arch/arm/Makefile
+++ b/xen/arch/arm/Makefile
@@ -35,6 +35,7 @@ obj-$(CONFIG_IOREQ_SERVER) += ioreq.o
obj-y += irq.o
obj-y += kernel.init.o
obj-$(CONFIG_LIVEPATCH) += livepatch.o
+obj-$(CONFIG_LLC_COLORING) += llc-coloring.o
obj-$(CONFIG_MEM_ACCESS) += mem_access.o
obj-y += mm.o
obj-y += monitor.o
diff --git a/xen/arch/arm/dom0less-build.c b/xen/arch/arm/dom0less-build.c
index f328a044e9..d93a85434e 100644
--- a/xen/arch/arm/dom0less-build.c
+++ b/xen/arch/arm/dom0less-build.c
@@ -5,6 +5,7 @@
#include <xen/grant_table.h>
#include <xen/iocap.h>
#include <xen/libfdt/libfdt.h>
+#include <xen/llc-coloring.h>
#include <xen/sched.h>
#include <xen/serial.h>
#include <xen/sizes.h>
@@ -890,7 +891,12 @@ void __init create_domUs(void)
panic("No more domain IDs available\n");
if ( dt_find_property(node, "xen,static-mem", NULL) )
+ {
+ if ( llc_coloring_enabled )
+ panic("LLC coloring and static memory are incompatible\n");
+
flags |= CDF_staticmem;
+ }
if ( dt_property_read_bool(node, "direct-map") )
{
diff --git a/xen/arch/arm/include/asm/processor.h b/xen/arch/arm/include/asm/processor.h
index 8e02410465..ef33ea198c 100644
--- a/xen/arch/arm/include/asm/processor.h
+++ b/xen/arch/arm/include/asm/processor.h
@@ -18,6 +18,22 @@
#define CTR_IDC_SHIFT 28
#define CTR_DIC_SHIFT 29
+/* CCSIDR Current Cache Size ID Register */
+#define CCSIDR_LINESIZE_MASK _AC(0x7, UL)
+#define CCSIDR_NUMSETS_SHIFT 13
+#define CCSIDR_NUMSETS_MASK _AC(0x3fff, UL)
+#define CCSIDR_NUMSETS_SHIFT_FEAT_CCIDX 32
+#define CCSIDR_NUMSETS_MASK_FEAT_CCIDX _AC(0xffffff, UL)
+
+/* CSSELR Cache Size Selection Register */
+#define CSSELR_LEVEL_MASK _AC(0x7, UL)
+#define CSSELR_LEVEL_SHIFT 1
+
+/* CLIDR Cache Level ID Register */
+#define CLIDR_CTYPEn_SHIFT(n) (3 * ((n) - 1))
+#define CLIDR_CTYPEn_MASK _AC(0x7, UL)
+#define CLIDR_CTYPEn_LEVELS 7
+
#define ICACHE_POLICY_VPIPT 0
#define ICACHE_POLICY_AIVIVT 1
#define ICACHE_POLICY_VIPT 2
diff --git a/xen/arch/arm/llc-coloring.c b/xen/arch/arm/llc-coloring.c
new file mode 100644
index 0000000000..a7e0907816
--- /dev/null
+++ b/xen/arch/arm/llc-coloring.c
@@ -0,0 +1,78 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Last Level Cache (LLC) coloring support for ARM
+ *
+ * Copyright (C) 2024, Advanced Micro Devices, Inc.
+ * Copyright (C) 2024, Minerva Systems SRL
+ */
+#include <xen/llc-coloring.h>
+#include <xen/types.h>
+
+#include <asm/processor.h>
+#include <asm/sysregs.h>
+
+/* Return the LLC way size by probing the hardware */
+unsigned int __init get_llc_way_size(void)
+{
+ register_t ccsidr_el1;
+ register_t clidr_el1 = READ_SYSREG(CLIDR_EL1);
+ register_t csselr_el1 = READ_SYSREG(CSSELR_EL1);
+ register_t id_aa64mmfr2_el1 = READ_SYSREG(ID_AA64MMFR2_EL1);
+ uint32_t ccsidr_numsets_shift = CCSIDR_NUMSETS_SHIFT;
+ uint32_t ccsidr_numsets_mask = CCSIDR_NUMSETS_MASK;
+ unsigned int n, line_size, num_sets;
+
+ for ( n = CLIDR_CTYPEn_LEVELS; n != 0; n-- )
+ {
+ uint8_t ctype_n = (clidr_el1 >> CLIDR_CTYPEn_SHIFT(n)) &
+ CLIDR_CTYPEn_MASK;
+
+ /* Unified cache (see Arm ARM DDI 0487J.a D19.2.27) */
+ if ( ctype_n == 0b100 )
+ break;
+ }
+
+ if ( n == 0 )
+ return 0;
+
+ WRITE_SYSREG((n - 1) << CSSELR_LEVEL_SHIFT, CSSELR_EL1);
+ isb();
+
+ ccsidr_el1 = READ_SYSREG(CCSIDR_EL1);
+
+ /* Arm ARM: (Log2(Number of bytes in cache line)) - 4 */
+ line_size = 1U << ((ccsidr_el1 & CCSIDR_LINESIZE_MASK) + 4);
+
+ /* If FEAT_CCIDX is enabled, CCSIDR_EL1 has a different bit layout */
+ if ( (id_aa64mmfr2_el1 >> ID_AA64MMFR2_CCIDX_SHIFT) & 0x7 )
+ {
+ ccsidr_numsets_shift = CCSIDR_NUMSETS_SHIFT_FEAT_CCIDX;
+ ccsidr_numsets_mask = CCSIDR_NUMSETS_MASK_FEAT_CCIDX;
+ }
+
+ /* Arm ARM: (Number of sets in cache) - 1 */
+ num_sets = ((ccsidr_el1 >> ccsidr_numsets_shift) & ccsidr_numsets_mask) + 1;
+
+ printk(XENLOG_INFO "LLC found: L%u (line size: %u bytes, sets num: %u)\n",
+ n, line_size, num_sets);
+
+ /* Restore value in CSSELR_EL1 */
+ WRITE_SYSREG(csselr_el1, CSSELR_EL1);
+ isb();
+
+ return line_size * num_sets;
+}
+
+void __init arch_llc_coloring_init(void)
+{
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/arch/arm/setup.c b/xen/arch/arm/setup.c
index 71ebaa77ca..84fecaabea 100644
--- a/xen/arch/arm/setup.c
+++ b/xen/arch/arm/setup.c
@@ -12,6 +12,7 @@
#include <xen/device_tree.h>
#include <xen/domain_page.h>
#include <xen/grant_table.h>
+#include <xen/llc-coloring.h>
#include <xen/types.h>
#include <xen/string.h>
#include <xen/serial.h>
@@ -326,6 +327,8 @@ void asmlinkage __init start_xen(unsigned long fdt_paddr)
printk("Command line: %s\n", cmdline);
cmdline_parse(cmdline);
+ llc_coloring_init();
+
setup_mm();
vm_init();
diff --git a/xen/common/llc-coloring.c b/xen/common/llc-coloring.c
index 29b75e0e0d..45b001b105 100644
--- a/xen/common/llc-coloring.c
+++ b/xen/common/llc-coloring.c
@@ -11,7 +11,7 @@
#define NR_LLC_COLORS (1U << CONFIG_LLC_COLORS_ORDER)
-static bool __ro_after_init llc_coloring_enabled;
+bool __ro_after_init llc_coloring_enabled;
boolean_param("llc-coloring", llc_coloring_enabled);
static unsigned int __initdata llc_size;
diff --git a/xen/include/xen/llc-coloring.h b/xen/include/xen/llc-coloring.h
index 6ccf4613c2..ad2525e153 100644
--- a/xen/include/xen/llc-coloring.h
+++ b/xen/include/xen/llc-coloring.h
@@ -12,10 +12,14 @@
#include <public/domctl.h>
#ifdef CONFIG_LLC_COLORING
+extern bool llc_coloring_enabled;
+
void llc_coloring_init(void);
void dump_llc_coloring_info(void);
void domain_dump_llc_colors(const struct domain *d);
#else
+#define llc_coloring_enabled false
+
static inline void llc_coloring_init(void) {}
static inline void dump_llc_coloring_info(void) {}
static inline void domain_dump_llc_colors(const struct domain *d) {}
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* [PATCH v10 03/12] xen/arm: permit non direct-mapped Dom0 construction
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
2024-11-19 14:13 ` [PATCH v10 01/12] xen/common: add cache coloring common code Carlo Nonato
2024-11-19 14:13 ` [PATCH v10 02/12] xen/arm: add initial support for LLC coloring on arm64 Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-11-28 10:34 ` Michal Orzel
2024-11-19 14:13 ` [PATCH v10 04/12] xen/arm: add Dom0 cache coloring support Carlo Nonato
` (8 subsequent siblings)
11 siblings, 1 reply; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Stefano Stabellini,
Julien Grall, Bertrand Marquis, Michal Orzel, Volodymyr Babchuk
Cache coloring requires Dom0 not to be direct-mapped because of its non
contiguous mapping nature, so allocate_memory() is needed in this case.
8d2c3ab18cc1 ("arm/dom0less: put dom0less feature code in a separate module")
moved allocate_memory() in dom0less_build.c. In order to use it
in Dom0 construction bring it back to domain_build.c and declare it in
domain_build.h.
Take the opportunity to adapt the implementation of allocate_memory() so
that it uses the host layout when called on the hwdom, via
find_unallocated_memory().
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
---
v10:
- fixed a compilation bug that happened when dom0less support was disabled
v9:
- no changes
v8:
- patch adapted to new changes to allocate_memory()
v7:
- allocate_memory() now uses the host layout when called on the hwdom
v6:
- new patch
---
xen/arch/arm/dom0less-build.c | 44 ------------
xen/arch/arm/domain_build.c | 96 +++++++++++++++++++++++--
xen/arch/arm/include/asm/domain_build.h | 1 +
3 files changed, 93 insertions(+), 48 deletions(-)
diff --git a/xen/arch/arm/dom0less-build.c b/xen/arch/arm/dom0less-build.c
index d93a85434e..67b1503647 100644
--- a/xen/arch/arm/dom0less-build.c
+++ b/xen/arch/arm/dom0less-build.c
@@ -49,50 +49,6 @@ bool __init is_dom0less_mode(void)
return ( !dom0found && domUfound );
}
-static void __init allocate_memory(struct domain *d, struct kernel_info *kinfo)
-{
- struct membanks *mem = kernel_info_get_mem(kinfo);
- unsigned int i;
- paddr_t bank_size;
-
- printk(XENLOG_INFO "Allocating mappings totalling %ldMB for %pd:\n",
- /* Don't want format this as PRIpaddr (16 digit hex) */
- (unsigned long)(kinfo->unassigned_mem >> 20), d);
-
- mem->nr_banks = 0;
- bank_size = MIN(GUEST_RAM0_SIZE, kinfo->unassigned_mem);
- if ( !allocate_bank_memory(kinfo, gaddr_to_gfn(GUEST_RAM0_BASE),
- bank_size) )
- goto fail;
-
- bank_size = MIN(GUEST_RAM1_SIZE, kinfo->unassigned_mem);
- if ( !allocate_bank_memory(kinfo, gaddr_to_gfn(GUEST_RAM1_BASE),
- bank_size) )
- goto fail;
-
- if ( kinfo->unassigned_mem )
- goto fail;
-
- for( i = 0; i < mem->nr_banks; i++ )
- {
- printk(XENLOG_INFO "%pd BANK[%d] %#"PRIpaddr"-%#"PRIpaddr" (%ldMB)\n",
- d,
- i,
- mem->bank[i].start,
- mem->bank[i].start + mem->bank[i].size,
- /* Don't want format this as PRIpaddr (16 digit hex) */
- (unsigned long)(mem->bank[i].size >> 20));
- }
-
- return;
-
-fail:
- panic("Failed to allocate requested domain memory."
- /* Don't want format this as PRIpaddr (16 digit hex) */
- " %ldKB unallocated. Fix the VMs configurations.\n",
- (unsigned long)kinfo->unassigned_mem >> 10);
-}
-
#ifdef CONFIG_VGICV2
static int __init make_gicv2_domU_node(struct kernel_info *kinfo)
{
diff --git a/xen/arch/arm/domain_build.c b/xen/arch/arm/domain_build.c
index 2c30792de8..a95376dcdc 100644
--- a/xen/arch/arm/domain_build.c
+++ b/xen/arch/arm/domain_build.c
@@ -416,7 +416,6 @@ static void __init allocate_memory_11(struct domain *d,
}
}
-#ifdef CONFIG_DOM0LESS_BOOT
bool __init allocate_domheap_memory(struct domain *d, paddr_t tot_size,
alloc_domheap_mem_cb cb, void *extra)
{
@@ -508,7 +507,6 @@ bool __init allocate_bank_memory(struct kernel_info *kinfo, gfn_t sgfn,
return true;
}
-#endif
/*
* When PCI passthrough is available we want to keep the
@@ -1003,6 +1001,93 @@ out:
return res;
}
+void __init allocate_memory(struct domain *d, struct kernel_info *kinfo)
+{
+ struct membanks *mem = kernel_info_get_mem(kinfo);
+ unsigned int i, nr_banks = 2;
+ paddr_t bank_start, bank_size;
+ struct membanks *hwdom_ext_regions = NULL;
+
+ printk(XENLOG_INFO "Allocating mappings totalling %ldMB for %pd:\n",
+ /* Don't want format this as PRIpaddr (16 digit hex) */
+ (unsigned long)(kinfo->unassigned_mem >> 20), d);
+
+ mem->nr_banks = 0;
+ /*
+ * Use host memory layout for hwdom. Only case for this is when LLC coloring
+ * is enabled.
+ */
+ if ( is_hardware_domain(d) )
+ {
+ ASSERT(llc_coloring_enabled);
+
+ hwdom_ext_regions = xzalloc_flex_struct(struct membanks, bank,
+ NR_MEM_BANKS);
+ if ( !hwdom_ext_regions )
+ goto fail;
+ hwdom_ext_regions->max_banks = NR_MEM_BANKS;
+
+ if ( find_unallocated_memory(kinfo, hwdom_ext_regions) )
+ goto fail;
+
+ nr_banks = hwdom_ext_regions->nr_banks;
+ }
+
+ for ( i = 0; kinfo->unassigned_mem > 0 && nr_banks > 0; i++, nr_banks-- )
+ {
+ if ( is_hardware_domain(d) )
+ {
+ bank_start = hwdom_ext_regions->bank[i].start;
+ bank_size = hwdom_ext_regions->bank[i].size;
+
+ if ( bank_size < min_t(paddr_t, kinfo->unassigned_mem, MB(128)) )
+ continue;
+ }
+ else
+ {
+ if ( i == 0 )
+ {
+ bank_start = GUEST_RAM0_BASE;
+ bank_size = GUEST_RAM0_SIZE;
+ }
+ else if ( i == 1 )
+ {
+ bank_start = GUEST_RAM1_BASE;
+ bank_size = GUEST_RAM1_SIZE;
+ }
+ else
+ goto fail;
+ }
+
+ bank_size = MIN(bank_size, kinfo->unassigned_mem);
+ if ( !allocate_bank_memory(kinfo, gaddr_to_gfn(bank_start), bank_size) )
+ goto fail;
+ }
+
+ if ( kinfo->unassigned_mem )
+ goto fail;
+
+ for( i = 0; i < mem->nr_banks; i++ )
+ {
+ printk(XENLOG_INFO "%pd BANK[%d] %#"PRIpaddr"-%#"PRIpaddr" (%ldMB)\n",
+ d,
+ i,
+ mem->bank[i].start,
+ mem->bank[i].start + mem->bank[i].size,
+ /* Don't want format this as PRIpaddr (16 digit hex) */
+ (unsigned long)(mem->bank[i].size >> 20));
+ }
+
+ xfree(hwdom_ext_regions);
+ return;
+
+fail:
+ panic("Failed to allocate requested domain memory."
+ /* Don't want format this as PRIpaddr (16 digit hex) */
+ " %ldKB unallocated. Fix the VMs configurations.\n",
+ (unsigned long)kinfo->unassigned_mem >> 10);
+}
+
static int __init handle_pci_range(const struct dt_device_node *dev,
uint64_t addr, uint64_t len, void *data)
{
@@ -1223,7 +1308,7 @@ int __init make_hypervisor_node(struct domain *d,
ext_regions->max_banks = NR_MEM_BANKS;
- if ( is_domain_direct_mapped(d) )
+ if ( domain_use_host_layout(d) )
{
if ( !is_iommu_enabled(d) )
res = find_unallocated_memory(kinfo, ext_regions);
@@ -2152,7 +2237,10 @@ static int __init construct_dom0(struct domain *d)
/* type must be set before allocate_memory */
d->arch.type = kinfo.type;
#endif
- allocate_memory_11(d, &kinfo);
+ if ( is_domain_direct_mapped(d) )
+ allocate_memory_11(d, &kinfo);
+ else
+ allocate_memory(d, &kinfo);
find_gnttab_region(d, &kinfo);
rc = process_shm_chosen(d, &kinfo);
diff --git a/xen/arch/arm/include/asm/domain_build.h b/xen/arch/arm/include/asm/domain_build.h
index e712afbc7f..5d77af2e8b 100644
--- a/xen/arch/arm/include/asm/domain_build.h
+++ b/xen/arch/arm/include/asm/domain_build.h
@@ -11,6 +11,7 @@ bool allocate_domheap_memory(struct domain *d, paddr_t tot_size,
alloc_domheap_mem_cb cb, void *extra);
bool allocate_bank_memory(struct kernel_info *kinfo, gfn_t sgfn,
paddr_t tot_size);
+void allocate_memory(struct domain *d, struct kernel_info *kinfo);
int construct_domain(struct domain *d, struct kernel_info *kinfo);
int domain_fdt_begin_node(void *fdt, const char *name, uint64_t unit);
int make_chosen_node(const struct kernel_info *kinfo);
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* [PATCH v10 04/12] xen/arm: add Dom0 cache coloring support
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
` (2 preceding siblings ...)
2024-11-19 14:13 ` [PATCH v10 03/12] xen/arm: permit non direct-mapped Dom0 construction Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-11-28 11:27 ` Michal Orzel
2024-11-19 14:13 ` [PATCH v10 05/12] xen: extend domctl interface for cache coloring Carlo Nonato
` (7 subsequent siblings)
11 siblings, 1 reply; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Andrew Cooper,
Jan Beulich, Julien Grall, Stefano Stabellini, Bertrand Marquis,
Michal Orzel, Volodymyr Babchuk
Add a command line parameter to allow the user to set the coloring
configuration for Dom0.
A common configuration syntax for cache colors is introduced and
documented.
Take the opportunity to also add:
- default configuration notion.
- function to check well-formed configurations.
Direct mapping Dom0 isn't possible when coloring is enabled, so
CDF_directmap flag is removed when creating it.
Based on original work from: Luca Miccio <lucmiccio@gmail.com>
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
---
v10:
- fixed array type for colors parameter in check_colors()
v9:
- moved domain_llc_coloring_free() in next patch cause it's never used for dom0
v8:
- added bound check on dom0_num_colors
- default colors array set just once
v7:
- parse_color_config() doesn't accept leading/trailing commas anymore
- removed alloc_colors() helper
v6:
- moved domain_llc_coloring_free() in this patch
- removed domain_alloc_colors() in favor of a more explicit allocation
- parse_color_config() now accepts the size of the array to be filled
- allocate_memory() moved in another patch
v5:
- Carlo Nonato as the new author
- moved dom0 colors parsing (parse_colors()) in this patch
- added dom0_set_llc_colors() to set dom0 colors after creation
- moved color allocation and checking in this patch
- error handling when allocating color arrays
- FIXME: copy pasted allocate_memory() cause it got moved
v4:
- dom0 colors are dynamically allocated as for any other domain
(colors are duplicated in dom0_colors and in the new array, but logic
is simpler)
---
docs/misc/cache-coloring.rst | 29 ++++++++
docs/misc/xen-command-line.pandoc | 9 +++
xen/arch/arm/domain_build.c | 10 ++-
xen/common/llc-coloring.c | 119 +++++++++++++++++++++++++++++-
xen/include/xen/llc-coloring.h | 1 +
5 files changed, 166 insertions(+), 2 deletions(-)
diff --git a/docs/misc/cache-coloring.rst b/docs/misc/cache-coloring.rst
index b608284e9b..c5fb33996c 100644
--- a/docs/misc/cache-coloring.rst
+++ b/docs/misc/cache-coloring.rst
@@ -105,6 +105,35 @@ Specific documentation is available at `docs/misc/xen-command-line.pandoc`.
+----------------------+-------------------------------+
| ``llc-nr-ways`` | Set the LLC number of ways |
+----------------------+-------------------------------+
+| ``dom0-llc-colors`` | Dom0 color configuration |
++----------------------+-------------------------------+
+
+Colors selection format
+***********************
+
+Regardless of the memory pool that has to be colored (Xen, Dom0/DomUs),
+the color selection can be expressed using the same syntax. In particular a
+comma-separated list of colors or ranges of colors is used.
+Ranges are hyphen-separated intervals (such as `0-4`) and are inclusive on both
+sides.
+
+Note that:
+
+- no spaces are allowed between values.
+- no overlapping ranges or duplicated colors are allowed.
+- values must be written in ascending order.
+
+Examples:
+
++-------------------+-----------------------------+
+| **Configuration** | **Actual selection** |
++-------------------+-----------------------------+
+| 1-2,5-8 | [1, 2, 5, 6, 7, 8] |
++-------------------+-----------------------------+
+| 4-8,10,11,12 | [4, 5, 6, 7, 8, 10, 11, 12] |
++-------------------+-----------------------------+
+| 0 | [0] |
++-------------------+-----------------------------+
Auto-probing of LLC specs
#########################
diff --git a/docs/misc/xen-command-line.pandoc b/docs/misc/xen-command-line.pandoc
index abd8dae96f..bfdc8b0002 100644
--- a/docs/misc/xen-command-line.pandoc
+++ b/docs/misc/xen-command-line.pandoc
@@ -963,6 +963,15 @@ Controls for the dom0 IOMMU setup.
Specify a list of IO ports to be excluded from dom0 access.
+### dom0-llc-colors (arm64)
+> `= List of [ <integer> | <integer>-<integer> ]`
+
+> Default: `All available LLC colors`
+
+Specify dom0 LLC color configuration. This option is available only when
+`CONFIG_LLC_COLORING` is enabled. If the parameter is not set, all available
+colors are used.
+
### dom0_max_vcpus
Either:
diff --git a/xen/arch/arm/domain_build.c b/xen/arch/arm/domain_build.c
index a95376dcdc..01aa9016f5 100644
--- a/xen/arch/arm/domain_build.c
+++ b/xen/arch/arm/domain_build.c
@@ -2,6 +2,7 @@
#include <xen/init.h>
#include <xen/compile.h>
#include <xen/lib.h>
+#include <xen/llc-coloring.h>
#include <xen/mm.h>
#include <xen/param.h>
#include <xen/domain_page.h>
@@ -2284,6 +2285,7 @@ void __init create_dom0(void)
.max_maptrack_frames = -1,
.grant_opts = XEN_DOMCTL_GRANT_version(opt_gnttab_max_version),
};
+ unsigned int flags = CDF_privileged;
int rc;
/* The vGIC for DOM0 is exactly emulating the hardware GIC */
@@ -2311,10 +2313,16 @@ void __init create_dom0(void)
panic("SVE vector length error\n");
}
- dom0 = domain_create(0, &dom0_cfg, CDF_privileged | CDF_directmap);
+ if ( !llc_coloring_enabled )
+ flags |= CDF_directmap;
+
+ dom0 = domain_create(0, &dom0_cfg, flags);
if ( IS_ERR(dom0) )
panic("Error creating domain 0 (rc = %ld)\n", PTR_ERR(dom0));
+ if ( llc_coloring_enabled && (rc = dom0_set_llc_colors(dom0)) )
+ panic("Error initializing LLC coloring for domain 0 (rc = %d)", rc);
+
if ( alloc_dom0_vcpu0(dom0) == NULL )
panic("Error creating domain 0 vcpu0\n");
diff --git a/xen/common/llc-coloring.c b/xen/common/llc-coloring.c
index 45b001b105..740b5b9e4f 100644
--- a/xen/common/llc-coloring.c
+++ b/xen/common/llc-coloring.c
@@ -20,6 +20,66 @@ static unsigned int __initdata llc_nr_ways;
integer_param("llc-nr-ways", llc_nr_ways);
/* Number of colors available in the LLC */
static unsigned int __ro_after_init max_nr_colors;
+/* Default coloring configuration */
+static unsigned int __ro_after_init default_colors[NR_LLC_COLORS];
+
+static unsigned int __initdata dom0_colors[NR_LLC_COLORS];
+static unsigned int __initdata dom0_num_colors;
+
+/*
+ * Parse the coloring configuration given in the buf string, following the
+ * syntax below.
+ *
+ * COLOR_CONFIGURATION ::= COLOR | RANGE,...,COLOR | RANGE
+ * RANGE ::= COLOR-COLOR
+ *
+ * Example: "0,2-6,15-16" represents the set of colors: 0,2,3,4,5,6,15,16.
+ */
+static int __init parse_color_config(const char *buf, unsigned int colors[],
+ unsigned int max_num_colors,
+ unsigned int *num_colors)
+{
+ const char *s = buf;
+
+ *num_colors = 0;
+
+ while ( *s != '\0' )
+ {
+ unsigned int color, start, end;
+
+ start = simple_strtoul(s, &s, 0);
+
+ if ( *s == '-' ) /* Range */
+ {
+ s++;
+ end = simple_strtoul(s, &s, 0);
+ }
+ else /* Single value */
+ end = start;
+
+ if ( start > end || (end - start) > (UINT_MAX - *num_colors) ||
+ (*num_colors + (end - start)) >= max_num_colors )
+ return -EINVAL;
+
+ /* Colors are range checked in check_colors() */
+ for ( color = start; color <= end; color++ )
+ colors[(*num_colors)++] = color;
+
+ if ( *s == ',' )
+ s++;
+ else if ( *s != '\0' )
+ break;
+ }
+
+ return *s ? -EINVAL : 0;
+}
+
+static int __init parse_dom0_colors(const char *s)
+{
+ return parse_color_config(s, dom0_colors, ARRAY_SIZE(dom0_colors),
+ &dom0_num_colors);
+}
+custom_param("dom0-llc-colors", parse_dom0_colors);
static void print_colors(const unsigned int colors[], unsigned int num_colors)
{
@@ -44,9 +104,26 @@ static void print_colors(const unsigned int colors[], unsigned int num_colors)
printk(" }\n");
}
+static bool __init check_colors(const unsigned int colors[],
+ unsigned int num_colors)
+{
+ unsigned int i;
+
+ for ( i = 0; i < num_colors; i++ )
+ {
+ if ( colors[i] >= max_nr_colors )
+ {
+ printk(XENLOG_ERR "LLC color %u >= %u\n", colors[i], max_nr_colors);
+ return false;
+ }
+ }
+
+ return true;
+}
+
void __init llc_coloring_init(void)
{
- unsigned int way_size;
+ unsigned int way_size, i;
if ( llc_size && llc_nr_ways )
{
@@ -83,6 +160,9 @@ void __init llc_coloring_init(void)
} else if ( max_nr_colors < 2 )
panic("Number of LLC colors %u < 2\n", max_nr_colors);
+ for ( i = 0; i < max_nr_colors; i++ )
+ default_colors[i] = i;
+
arch_llc_coloring_init();
}
@@ -104,6 +184,43 @@ void domain_dump_llc_colors(const struct domain *d)
print_colors(d->llc_colors, d->num_llc_colors);
}
+static void __init domain_set_default_colors(struct domain *d)
+{
+ printk(XENLOG_WARNING
+ "LLC color config not found for %pd, using all colors\n", d);
+
+ d->llc_colors = default_colors;
+ d->num_llc_colors = max_nr_colors;
+}
+
+int __init dom0_set_llc_colors(struct domain *d)
+{
+ typeof(*dom0_colors) *colors;
+
+ if ( !dom0_num_colors )
+ {
+ domain_set_default_colors(d);
+ return 0;
+ }
+
+ if ( dom0_num_colors > max_nr_colors ||
+ !check_colors(dom0_colors, dom0_num_colors) )
+ {
+ printk(XENLOG_ERR "%pd: bad LLC color config\n", d);
+ return -EINVAL;
+ }
+
+ colors = xmalloc_array(typeof(*dom0_colors), dom0_num_colors);
+ if ( !colors )
+ return -ENOMEM;
+
+ memcpy(colors, dom0_colors, sizeof(*colors) * dom0_num_colors);
+ d->llc_colors = colors;
+ d->num_llc_colors = dom0_num_colors;
+
+ return 0;
+}
+
/*
* Local variables:
* mode: C
diff --git a/xen/include/xen/llc-coloring.h b/xen/include/xen/llc-coloring.h
index ad2525e153..9cf49f574e 100644
--- a/xen/include/xen/llc-coloring.h
+++ b/xen/include/xen/llc-coloring.h
@@ -27,6 +27,7 @@ static inline void domain_dump_llc_colors(const struct domain *d) {}
unsigned int get_llc_way_size(void);
void arch_llc_coloring_init(void);
+int dom0_set_llc_colors(struct domain *d);
#endif /* __COLORING_H__ */
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* [PATCH v10 05/12] xen: extend domctl interface for cache coloring
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
` (3 preceding siblings ...)
2024-11-19 14:13 ` [PATCH v10 04/12] xen/arm: add Dom0 cache coloring support Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-11-19 14:13 ` [PATCH v10 06/12] tools: add support for cache coloring configuration Carlo Nonato
` (6 subsequent siblings)
11 siblings, 0 replies; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Andrew Cooper,
Jan Beulich, Julien Grall, Stefano Stabellini
Add a new domctl hypercall to allow the user to set LLC coloring
configurations. Colors can be set only once, just after domain creation,
since recoloring isn't supported.
Based on original work from: Luca Miccio <lucmiccio@gmail.com>
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
---
v10:
- no changes
v9:
- minor printk message changes
- moved domain_llc_coloring_free() in this patch
v8:
- fixed memory leak on error path of domain_set_llc_colors()
v7:
- -EOPNOTSUPP returned in case of hypercall called without llc_coloring_enabled
- domain_set_llc_colors_domctl() renamed to domain_set_llc_colors()
- added padding and input bound checks to domain_set_llc_colors()
- removed alloc_colors() helper usage from domain_set_llc_colors()
v6:
- reverted the XEN_DOMCTL_INTERFACE_VERSION bump
- reverted to uint32 for the guest handle
- explicit padding added to the domctl struct
- rewrote domain_set_llc_colors_domctl() to be more explicit
v5:
- added a new hypercall to set colors
- uint for the guest handle
v4:
- updated XEN_DOMCTL_INTERFACE_VERSION
---
xen/common/domain.c | 3 ++
xen/common/domctl.c | 10 +++++++
xen/common/llc-coloring.c | 55 ++++++++++++++++++++++++++++++++--
xen/include/public/domctl.h | 9 ++++++
xen/include/xen/llc-coloring.h | 4 +++
5 files changed, 78 insertions(+), 3 deletions(-)
diff --git a/xen/common/domain.c b/xen/common/domain.c
index 92263a4fbd..842a23751a 100644
--- a/xen/common/domain.c
+++ b/xen/common/domain.c
@@ -34,6 +34,7 @@
#include <xen/xenoprof.h>
#include <xen/irq.h>
#include <xen/argo.h>
+#include <xen/llc-coloring.h>
#include <asm/p2m.h>
#include <asm/processor.h>
#include <public/sched.h>
@@ -1276,6 +1277,8 @@ void domain_destroy(struct domain *d)
{
BUG_ON(!d->is_dying);
+ domain_llc_coloring_free(d);
+
/* May be already destroyed, or get_domain() can race us. */
if ( atomic_cmpxchg(&d->refcnt, 0, DOMAIN_DESTROYED) != 0 )
return;
diff --git a/xen/common/domctl.c b/xen/common/domctl.c
index ea16b75910..6387dddbcd 100644
--- a/xen/common/domctl.c
+++ b/xen/common/domctl.c
@@ -8,6 +8,7 @@
#include <xen/types.h>
#include <xen/lib.h>
+#include <xen/llc-coloring.h>
#include <xen/err.h>
#include <xen/mm.h>
#include <xen/sched.h>
@@ -866,6 +867,15 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
__HYPERVISOR_domctl, "h", u_domctl);
break;
+ case XEN_DOMCTL_set_llc_colors:
+ if ( op->u.set_llc_colors.pad )
+ ret = -EINVAL;
+ else if ( llc_coloring_enabled )
+ ret = domain_set_llc_colors(d, &op->u.set_llc_colors);
+ else
+ ret = -EOPNOTSUPP;
+ break;
+
default:
ret = arch_do_domctl(op, d, u_domctl);
break;
diff --git a/xen/common/llc-coloring.c b/xen/common/llc-coloring.c
index 740b5b9e4f..3cfd0ce78c 100644
--- a/xen/common/llc-coloring.c
+++ b/xen/common/llc-coloring.c
@@ -5,6 +5,7 @@
* Copyright (C) 2024, Advanced Micro Devices, Inc.
* Copyright (C) 2024, Minerva Systems SRL
*/
+#include <xen/guest_access.h>
#include <xen/keyhandler.h>
#include <xen/llc-coloring.h>
#include <xen/param.h>
@@ -104,8 +105,7 @@ static void print_colors(const unsigned int colors[], unsigned int num_colors)
printk(" }\n");
}
-static bool __init check_colors(const unsigned int colors[],
- unsigned int num_colors)
+static bool check_colors(const unsigned int colors[], unsigned int num_colors)
{
unsigned int i;
@@ -184,7 +184,7 @@ void domain_dump_llc_colors(const struct domain *d)
print_colors(d->llc_colors, d->num_llc_colors);
}
-static void __init domain_set_default_colors(struct domain *d)
+static void domain_set_default_colors(struct domain *d)
{
printk(XENLOG_WARNING
"LLC color config not found for %pd, using all colors\n", d);
@@ -221,6 +221,55 @@ int __init dom0_set_llc_colors(struct domain *d)
return 0;
}
+int domain_set_llc_colors(struct domain *d,
+ const struct xen_domctl_set_llc_colors *config)
+{
+ unsigned int *colors;
+
+ if ( d->num_llc_colors )
+ return -EEXIST;
+
+ if ( !config->num_llc_colors )
+ {
+ domain_set_default_colors(d);
+ return 0;
+ }
+
+ if ( config->num_llc_colors > max_nr_colors )
+ return -EINVAL;
+
+ colors = xmalloc_array(unsigned int, config->num_llc_colors);
+ if ( !colors )
+ return -ENOMEM;
+
+ if ( copy_from_guest(colors, config->llc_colors, config->num_llc_colors) )
+ {
+ xfree(colors);
+ return -EFAULT;
+ }
+
+ if ( !check_colors(colors, config->num_llc_colors) )
+ {
+ printk(XENLOG_ERR "%pd: bad LLC color config\n", d);
+ xfree(colors);
+ return -EINVAL;
+ }
+
+ d->llc_colors = colors;
+ d->num_llc_colors = config->num_llc_colors;
+
+ return 0;
+}
+
+void domain_llc_coloring_free(struct domain *d)
+{
+ if ( !llc_coloring_enabled || d->llc_colors == default_colors )
+ return;
+
+ /* free pointer-to-const using __va(__pa()) */
+ xfree(__va(__pa(d->llc_colors)));
+}
+
/*
* Local variables:
* mode: C
diff --git a/xen/include/public/domctl.h b/xen/include/public/domctl.h
index 353f831e40..e2d392d1e5 100644
--- a/xen/include/public/domctl.h
+++ b/xen/include/public/domctl.h
@@ -1236,6 +1236,13 @@ struct xen_domctl_dt_overlay {
};
#endif
+struct xen_domctl_set_llc_colors {
+ /* IN LLC coloring parameters */
+ uint32_t num_llc_colors;
+ uint32_t pad;
+ XEN_GUEST_HANDLE_64(uint32) llc_colors;
+};
+
struct xen_domctl {
uint32_t cmd;
#define XEN_DOMCTL_createdomain 1
@@ -1325,6 +1332,7 @@ struct xen_domctl {
#define XEN_DOMCTL_set_paging_mempool_size 86
#define XEN_DOMCTL_dt_overlay 87
#define XEN_DOMCTL_gsi_permission 88
+#define XEN_DOMCTL_set_llc_colors 89
#define XEN_DOMCTL_gdbsx_guestmemio 1000
#define XEN_DOMCTL_gdbsx_pausevcpu 1001
#define XEN_DOMCTL_gdbsx_unpausevcpu 1002
@@ -1391,6 +1399,7 @@ struct xen_domctl {
#if defined(__arm__) || defined(__aarch64__)
struct xen_domctl_dt_overlay dt_overlay;
#endif
+ struct xen_domctl_set_llc_colors set_llc_colors;
uint8_t pad[128];
} u;
};
diff --git a/xen/include/xen/llc-coloring.h b/xen/include/xen/llc-coloring.h
index 9cf49f574e..657f4d0a10 100644
--- a/xen/include/xen/llc-coloring.h
+++ b/xen/include/xen/llc-coloring.h
@@ -17,17 +17,21 @@ extern bool llc_coloring_enabled;
void llc_coloring_init(void);
void dump_llc_coloring_info(void);
void domain_dump_llc_colors(const struct domain *d);
+void domain_llc_coloring_free(struct domain *d);
#else
#define llc_coloring_enabled false
static inline void llc_coloring_init(void) {}
static inline void dump_llc_coloring_info(void) {}
static inline void domain_dump_llc_colors(const struct domain *d) {}
+static inline void domain_llc_coloring_free(struct domain *d) {}
#endif
unsigned int get_llc_way_size(void);
void arch_llc_coloring_init(void);
int dom0_set_llc_colors(struct domain *d);
+int domain_set_llc_colors(struct domain *d,
+ const struct xen_domctl_set_llc_colors *config);
#endif /* __COLORING_H__ */
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* [PATCH v10 06/12] tools: add support for cache coloring configuration
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
` (4 preceding siblings ...)
2024-11-19 14:13 ` [PATCH v10 05/12] xen: extend domctl interface for cache coloring Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-11-20 16:25 ` Anthony PERARD
2024-11-20 16:48 ` Anthony PERARD
2024-11-19 14:13 ` [PATCH v10 07/12] xen/arm: add support for cache coloring configuration via device-tree Carlo Nonato
` (5 subsequent siblings)
11 siblings, 2 replies; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Anthony PERARD,
Juergen Gross
Add a new "llc_colors" parameter that defines the LLC color assignment for
a domain. The user can specify one or more color ranges using the same
syntax used everywhere else for color config described in the
documentation.
The parameter is defined as a list of strings that represent the color
ranges.
Documentation is also added.
Based on original work from: Luca Miccio <lucmiccio@gmail.com>
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
---
v10:
- no changes
v9:
- turned warning into error in case of coloring not enabled
v8:
- warn the user in case of coloring not supported at hypervisor level
v7:
- removed unneeded NULL check before xc_hypercall_buffer_free() in
xc_domain_set_llc_colors()
v6:
- no edits
v5:
- added LIBXL_HAVE_BUILDINFO_LLC_COLORS
- moved color configuration in xc_domain_set_llc_colors() cause of the new
hypercall
v4:
- removed overlapping color ranges checks during parsing
- moved hypercall buffer initialization in libxenctrl
---
docs/man/xl.cfg.5.pod.in | 6 +++++
tools/include/libxl.h | 5 +++++
tools/include/xenctrl.h | 9 ++++++++
tools/libs/ctrl/xc_domain.c | 35 +++++++++++++++++++++++++++++
tools/libs/light/libxl_create.c | 18 +++++++++++++++
tools/libs/light/libxl_types.idl | 1 +
tools/xl/xl_parse.c | 38 +++++++++++++++++++++++++++++++-
7 files changed, 111 insertions(+), 1 deletion(-)
diff --git a/docs/man/xl.cfg.5.pod.in b/docs/man/xl.cfg.5.pod.in
index ac3f88fd57..8e1422104e 100644
--- a/docs/man/xl.cfg.5.pod.in
+++ b/docs/man/xl.cfg.5.pod.in
@@ -3074,6 +3074,12 @@ raised.
=over 4
+=item B<llc_colors=[ "RANGE", "RANGE", ...]>
+
+Specify the Last Level Cache (LLC) color configuration for the guest.
+B<RANGE> can be either a single color value or a hypen-separated closed
+interval of colors (such as "0-4").
+
=item B<nr_spis="NR_SPIS">
An optional integer parameter specifying the number of SPIs (Shared
diff --git a/tools/include/libxl.h b/tools/include/libxl.h
index 8d32428ea9..f8fe4afd7d 100644
--- a/tools/include/libxl.h
+++ b/tools/include/libxl.h
@@ -1379,6 +1379,11 @@ void libxl_mac_copy(libxl_ctx *ctx, libxl_mac *dst, const libxl_mac *src);
*/
#define LIBXL_HAVE_BUILDINFO_HVM_SYSTEM_FIRMWARE
+/*
+ * The libxl_domain_build_info has the llc_colors array.
+ */
+#define LIBXL_HAVE_BUILDINFO_LLC_COLORS 1
+
/*
* ERROR_REMUS_XXX error code only exists from Xen 4.5, Xen 4.6 and it
* is changed to ERROR_CHECKPOINT_XXX in Xen 4.7
diff --git a/tools/include/xenctrl.h b/tools/include/xenctrl.h
index 29617585c5..aa2b4b04f3 100644
--- a/tools/include/xenctrl.h
+++ b/tools/include/xenctrl.h
@@ -2667,6 +2667,15 @@ int xc_livepatch_replace(xc_interface *xch, char *name, uint32_t timeout, uint32
int xc_domain_cacheflush(xc_interface *xch, uint32_t domid,
xen_pfn_t start_pfn, xen_pfn_t nr_pfns);
+/*
+ * Set LLC colors for a domain.
+ * It can only be used directly after domain creation. An attempt to use it
+ * afterwards will result in an error.
+ */
+int xc_domain_set_llc_colors(xc_interface *xch, uint32_t domid,
+ const unsigned int *llc_colors,
+ unsigned int num_llc_colors);
+
#if defined(__arm__) || defined(__aarch64__)
int xc_dt_overlay(xc_interface *xch, void *overlay_fdt,
uint32_t overlay_fdt_size, uint8_t overlay_op);
diff --git a/tools/libs/ctrl/xc_domain.c b/tools/libs/ctrl/xc_domain.c
index e3538ec0ba..4ed339e6e4 100644
--- a/tools/libs/ctrl/xc_domain.c
+++ b/tools/libs/ctrl/xc_domain.c
@@ -2195,6 +2195,41 @@ int xc_domain_soft_reset(xc_interface *xch,
domctl.domain = domid;
return do_domctl(xch, &domctl);
}
+
+int xc_domain_set_llc_colors(xc_interface *xch, uint32_t domid,
+ const unsigned int *llc_colors,
+ unsigned int num_llc_colors)
+{
+ struct xen_domctl domctl = {};
+ DECLARE_HYPERCALL_BUFFER(uint32_t, local);
+ int ret = -1;
+
+ if ( num_llc_colors )
+ {
+ size_t bytes = sizeof(uint32_t) * num_llc_colors;
+
+ local = xc_hypercall_buffer_alloc(xch, local, bytes);
+ if ( local == NULL )
+ {
+ PERROR("Could not allocate LLC colors for set_llc_colors");
+ ret = -ENOMEM;
+ goto out;
+ }
+ memcpy(local, llc_colors, bytes);
+ set_xen_guest_handle(domctl.u.set_llc_colors.llc_colors, local);
+ }
+
+ domctl.cmd = XEN_DOMCTL_set_llc_colors;
+ domctl.domain = domid;
+ domctl.u.set_llc_colors.num_llc_colors = num_llc_colors;
+
+ ret = do_domctl(xch, &domctl);
+
+out:
+ xc_hypercall_buffer_free(xch, local);
+
+ return ret;
+}
/*
* Local variables:
* mode: C
diff --git a/tools/libs/light/libxl_create.c b/tools/libs/light/libxl_create.c
index edeadd57ef..e03599ea99 100644
--- a/tools/libs/light/libxl_create.c
+++ b/tools/libs/light/libxl_create.c
@@ -747,6 +747,24 @@ int libxl__domain_make(libxl__gc *gc, libxl_domain_config *d_config,
/* A new domain now exists */
*domid = local_domid;
+ ret = xc_domain_set_llc_colors(ctx->xch, local_domid,
+ b_info->llc_colors,
+ b_info->num_llc_colors);
+ if (ret < 0) {
+ if (errno == EOPNOTSUPP) {
+ if (b_info->num_llc_colors > 0) {
+ LOGED(ERROR, local_domid,
+ "LLC coloring not enabled in the hypervisor");
+ rc = ERROR_FAIL;
+ goto out;
+ }
+ } else {
+ LOGED(ERROR, local_domid, "LLC colors allocation failed");
+ rc = ERROR_FAIL;
+ goto out;
+ }
+ }
+
rc = libxl__is_domid_recent(gc, local_domid, &recent);
if (rc)
goto out;
diff --git a/tools/libs/light/libxl_types.idl b/tools/libs/light/libxl_types.idl
index 4e65e6fda5..bd4b8721ff 100644
--- a/tools/libs/light/libxl_types.idl
+++ b/tools/libs/light/libxl_types.idl
@@ -616,6 +616,7 @@ libxl_domain_build_info = Struct("domain_build_info",[
("ioports", Array(libxl_ioport_range, "num_ioports")),
("irqs", Array(uint32, "num_irqs")),
("iomem", Array(libxl_iomem_range, "num_iomem")),
+ ("llc_colors", Array(uint32, "num_llc_colors")),
("claim_mode", libxl_defbool),
("event_channels", uint32),
("kernel", string),
diff --git a/tools/xl/xl_parse.c b/tools/xl/xl_parse.c
index e3a4800f6e..3d85be7dd4 100644
--- a/tools/xl/xl_parse.c
+++ b/tools/xl/xl_parse.c
@@ -1296,7 +1296,7 @@ void parse_config_data(const char *config_source,
XLU_ConfigList *cpus, *vbds, *nics, *pcis, *cvfbs, *cpuids, *vtpms,
*usbctrls, *usbdevs, *p9devs, *vdispls, *pvcallsifs_devs;
XLU_ConfigList *channels, *ioports, *irqs, *iomem, *viridian, *dtdevs,
- *mca_caps, *smbios;
+ *mca_caps, *smbios, *llc_colors;
int num_ioports, num_irqs, num_iomem, num_cpus, num_viridian, num_mca_caps;
int num_smbios;
int pci_power_mgmt = 0;
@@ -1304,6 +1304,7 @@ void parse_config_data(const char *config_source,
int pci_permissive = 0;
int pci_seize = 0;
int i, e;
+ int num_llc_colors;
char *kernel_basename;
libxl_domain_create_info *c_info = &d_config->c_info;
@@ -1447,6 +1448,41 @@ void parse_config_data(const char *config_source,
if (!xlu_cfg_get_long (config, "maxmem", &l, 0))
b_info->max_memkb = l * 1024;
+ if (!xlu_cfg_get_list(config, "llc_colors", &llc_colors, &num_llc_colors, 0)) {
+ int cur_index = 0;
+
+ b_info->num_llc_colors = 0;
+ for (i = 0; i < num_llc_colors; i++) {
+ uint32_t start = 0, end = 0, k;
+
+ buf = xlu_cfg_get_listitem(llc_colors, i);
+ if (!buf) {
+ fprintf(stderr,
+ "xl: Can't get element %d in LLC color list\n", i);
+ exit(1);
+ }
+
+ if (sscanf(buf, "%" SCNu32 "-%" SCNu32, &start, &end) != 2) {
+ if (sscanf(buf, "%" SCNu32, &start) != 1) {
+ fprintf(stderr, "xl: Invalid LLC color range: %s\n", buf);
+ exit(1);
+ }
+ end = start;
+ } else if (start > end) {
+ fprintf(stderr,
+ "xl: Start LLC color is greater than end: %s\n", buf);
+ exit(1);
+ }
+
+ b_info->num_llc_colors += (end - start) + 1;
+ b_info->llc_colors = (uint32_t *)realloc(b_info->llc_colors,
+ sizeof(*b_info->llc_colors) * b_info->num_llc_colors);
+
+ for (k = start; k <= end; k++)
+ b_info->llc_colors[cur_index++] = k;
+ }
+ }
+
if (!xlu_cfg_get_long (config, "vcpus", &l, 0)) {
vcpus = l;
if (libxl_cpu_bitmap_alloc(ctx, &b_info->avail_vcpus, l)) {
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* [PATCH v10 07/12] xen/arm: add support for cache coloring configuration via device-tree
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
` (5 preceding siblings ...)
2024-11-19 14:13 ` [PATCH v10 06/12] tools: add support for cache coloring configuration Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-11-28 12:00 ` Michal Orzel
2024-11-19 14:13 ` [PATCH v10 08/12] xen/page_alloc: introduce preserved page flags macro Carlo Nonato
` (4 subsequent siblings)
11 siblings, 1 reply; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Stefano Stabellini,
Julien Grall, Bertrand Marquis, Michal Orzel, Volodymyr Babchuk,
Andrew Cooper, Jan Beulich
Add the "llc-colors" Device Tree attribute to express DomUs and Dom0less
color configurations.
Based on original work from: Luca Miccio <lucmiccio@gmail.com>
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
Reviewed-by: Jan Beulich <jbeulich@suse.com> # non-Arm
---
v10:
- no changes
v9:
- use best-effort allocation in domain_set_llc_colors_from_str()
v8:
- fixed memory leak on error path of domain_set_llc_colors_from_str()
- realloc colors array after parsing from string to reduce memory usage
v7:
- removed alloc_colors() helper usage from domain_set_llc_colors_from_str()
v6:
- rewrote domain_set_llc_colors_from_str() to be more explicit
v5:
- static-mem check has been moved in a previous patch
- added domain_set_llc_colors_from_str() to set colors after domain creation
---
docs/misc/arm/device-tree/booting.txt | 4 +++
docs/misc/cache-coloring.rst | 48 +++++++++++++++++++++++++++
xen/arch/arm/dom0less-build.c | 10 ++++++
xen/common/llc-coloring.c | 41 +++++++++++++++++++++++
xen/include/xen/llc-coloring.h | 1 +
xen/include/xen/xmalloc.h | 12 +++++++
6 files changed, 116 insertions(+)
diff --git a/docs/misc/arm/device-tree/booting.txt b/docs/misc/arm/device-tree/booting.txt
index 3a04f5c57f..9085645433 100644
--- a/docs/misc/arm/device-tree/booting.txt
+++ b/docs/misc/arm/device-tree/booting.txt
@@ -162,6 +162,10 @@ with the following properties:
An integer specifying the number of vcpus to allocate to the guest.
+- llc-colors
+ A string specifying the LLC color configuration for the guest.
+ Refer to docs/misc/cache_coloring.rst for syntax.
+
- vpl011
An empty property to enable/disable a virtual pl011 for the guest to
diff --git a/docs/misc/cache-coloring.rst b/docs/misc/cache-coloring.rst
index c5fb33996c..c14b6ee2f4 100644
--- a/docs/misc/cache-coloring.rst
+++ b/docs/misc/cache-coloring.rst
@@ -12,6 +12,7 @@ If needed, change the maximum number of colors with
``CONFIG_LLC_COLORS_ORDER=<n>``.
Runtime configuration is done via `Command line parameters`_.
+For DomUs follow `DomUs configuration`_.
Background
**********
@@ -147,6 +148,53 @@ LLC specs can be manually set via the above command line parameters. This
bypasses any auto-probing and it's used to overcome failing situations, such as
flawed probing logic, or for debugging/testing purposes.
+DomUs configuration
+*******************
+
+DomUs colors can be set either in the ``xl`` configuration file (documentation
+at `docs/man/xl.cfg.pod.5.in`) or via Device Tree, also for Dom0less
+configurations (documentation at `docs/misc/arm/device-tree/booting.txt`) using
+the ``llc-colors`` option. For example:
+
+::
+
+ xen,xen-bootargs = "console=dtuart dtuart=serial0 dom0_mem=1G dom0_max_vcpus=1 sched=null llc-coloring=on dom0-llc-colors=2-6";
+ xen,dom0-bootargs "console=hvc0 earlycon=xen earlyprintk=xen root=/dev/ram0"
+
+ dom0 {
+ compatible = "xen,linux-zimage" "xen,multiboot-module";
+ reg = <0x0 0x1000000 0x0 15858176>;
+ };
+
+ dom0-ramdisk {
+ compatible = "xen,linux-initrd" "xen,multiboot-module";
+ reg = <0x0 0x2000000 0x0 20638062>;
+ };
+
+ domU0 {
+ #address-cells = <0x1>;
+ #size-cells = <0x1>;
+ compatible = "xen,domain";
+ memory = <0x0 0x40000>;
+ llc-colors = "4-8,10,11,12";
+ cpus = <0x1>;
+ vpl011 = <0x1>;
+
+ module@2000000 {
+ compatible = "multiboot,kernel", "multiboot,module";
+ reg = <0x2000000 0xffffff>;
+ bootargs = "console=ttyAMA0";
+ };
+
+ module@30000000 {
+ compatible = "multiboot,ramdisk", "multiboot,module";
+ reg = <0x3000000 0xffffff>;
+ };
+ };
+
+**Note:** If no color configuration is provided for a domain, the default one,
+which corresponds to all available colors is used instead.
+
Known issues and limitations
****************************
diff --git a/xen/arch/arm/dom0less-build.c b/xen/arch/arm/dom0less-build.c
index 67b1503647..49d1f14d65 100644
--- a/xen/arch/arm/dom0less-build.c
+++ b/xen/arch/arm/dom0less-build.c
@@ -817,6 +817,7 @@ void __init create_domUs(void)
bool iommu = false;
const struct dt_device_node *cpupool_node,
*chosen = dt_find_node_by_path("/chosen");
+ const char *llc_colors_str = NULL;
BUG_ON(chosen == NULL);
dt_for_each_child_node(chosen, node)
@@ -965,6 +966,10 @@ void __init create_domUs(void)
#endif
}
+ dt_property_read_string(node, "llc-colors", &llc_colors_str);
+ if ( !llc_coloring_enabled && llc_colors_str )
+ panic("'llc-colors' found, but LLC coloring is disabled\n");
+
/*
* The variable max_init_domid is initialized with zero, so here it's
* very important to use the pre-increment operator to call
@@ -975,6 +980,11 @@ void __init create_domUs(void)
panic("Error creating domain %s (rc = %ld)\n",
dt_node_name(node), PTR_ERR(d));
+ if ( llc_coloring_enabled &&
+ (rc = domain_set_llc_colors_from_str(d, llc_colors_str)) )
+ panic("Error initializing LLC coloring for domain %s (rc = %d)\n",
+ dt_node_name(node), rc);
+
d->is_console = true;
dt_device_set_used_by(node, d->domain_id);
diff --git a/xen/common/llc-coloring.c b/xen/common/llc-coloring.c
index 3cfd0ce78c..14e056fb10 100644
--- a/xen/common/llc-coloring.c
+++ b/xen/common/llc-coloring.c
@@ -5,6 +5,7 @@
* Copyright (C) 2024, Advanced Micro Devices, Inc.
* Copyright (C) 2024, Minerva Systems SRL
*/
+#include "xen/xmalloc.h"
#include <xen/guest_access.h>
#include <xen/keyhandler.h>
#include <xen/llc-coloring.h>
@@ -270,6 +271,46 @@ void domain_llc_coloring_free(struct domain *d)
xfree(__va(__pa(d->llc_colors)));
}
+int __init domain_set_llc_colors_from_str(struct domain *d, const char *str)
+{
+ int err;
+ unsigned int *colors, num_colors;
+
+ if ( !str )
+ {
+ domain_set_default_colors(d);
+ return 0;
+ }
+
+ colors = xmalloc_array(unsigned int, max_nr_colors);
+ if ( !colors )
+ return -ENOMEM;
+
+ err = parse_color_config(str, colors, max_nr_colors, &num_colors);
+ if ( err )
+ {
+ printk(XENLOG_ERR "Error parsing LLC color configuration");
+ xfree(colors);
+ return err;
+ }
+
+ if ( !check_colors(colors, num_colors) )
+ {
+ printk(XENLOG_ERR "%pd: bad LLC color config\n", d);
+ xfree(colors);
+ return -EINVAL;
+ }
+
+ /* Adjust the size cause it was initially set to max_nr_colors */
+ d->llc_colors = xrealloc_array(colors, num_colors);
+ if ( !d->llc_colors )
+ d->llc_colors = colors;
+
+ d->num_llc_colors = num_colors;
+
+ return 0;
+}
+
/*
* Local variables:
* mode: C
diff --git a/xen/include/xen/llc-coloring.h b/xen/include/xen/llc-coloring.h
index 657f4d0a10..76d3cbae0e 100644
--- a/xen/include/xen/llc-coloring.h
+++ b/xen/include/xen/llc-coloring.h
@@ -32,6 +32,7 @@ void arch_llc_coloring_init(void);
int dom0_set_llc_colors(struct domain *d);
int domain_set_llc_colors(struct domain *d,
const struct xen_domctl_set_llc_colors *config);
+int domain_set_llc_colors_from_str(struct domain *d, const char *str);
#endif /* __COLORING_H__ */
diff --git a/xen/include/xen/xmalloc.h b/xen/include/xen/xmalloc.h
index b903fa2e26..f0412fb4e0 100644
--- a/xen/include/xen/xmalloc.h
+++ b/xen/include/xen/xmalloc.h
@@ -37,6 +37,9 @@
((_type *)_xmalloc_array(sizeof(_type), __alignof__(_type), _num))
#define xzalloc_array(_type, _num) \
((_type *)_xzalloc_array(sizeof(_type), __alignof__(_type), _num))
+#define xrealloc_array(_ptr, _num) \
+ ((typeof(_ptr))_xrealloc_array(_ptr, sizeof(typeof(*(_ptr))), \
+ __alignof__(typeof(*(_ptr))), _num))
/* Allocate space for a structure with a flexible array of typed objects. */
#define xzalloc_flex_struct(type, field, nr) \
@@ -98,6 +101,15 @@ static inline void *_xzalloc_array(
return _xzalloc(size * num, align);
}
+static inline void *_xrealloc_array(
+ void *ptr, unsigned long size, unsigned long align, unsigned long num)
+{
+ /* Check for overflow. */
+ if ( size && num > UINT_MAX / size )
+ return NULL;
+ return _xrealloc(ptr, size * num, align);
+}
+
/*
* Pooled allocator interface.
*/
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* [PATCH v10 08/12] xen/page_alloc: introduce preserved page flags macro
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
` (6 preceding siblings ...)
2024-11-19 14:13 ` [PATCH v10 07/12] xen/arm: add support for cache coloring configuration via device-tree Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-11-28 11:04 ` Jan Beulich
2024-11-19 14:13 ` [PATCH v10 09/12] xen: add cache coloring allocator for domains Carlo Nonato
` (3 subsequent siblings)
11 siblings, 1 reply; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Andrew Cooper,
Jan Beulich, Julien Grall, Stefano Stabellini
PGC_static, PGC_extra and PGC_broken need to be preserved when assigning a
page. Define a new macro that groups those flags and use it instead of or'ing
every time.
To make preserved flags even more meaningful, they are kept also when
switching state in mark_page_free().
Enforce the removal of PGC_extra before freeing domain pages as this is
considered an error and can cause ASSERT violations.
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
---
v10:
- fixed commit message
v9:
- add PGC_broken to PGC_preserved
- clear PGC_extra in alloc_domheap_pages() only if MEMF_no_refcount is set
v8:
- fixed PGC_extra ASSERT fail in alloc_domheap_pages() by removing PGC_extra
before freeing
v7:
- PGC_preserved used also in mark_page_free()
v6:
- preserved_flags renamed to PGC_preserved
- PGC_preserved is used only in assign_pages()
v5:
- new patch
---
xen/common/page_alloc.c | 19 ++++++++++++++-----
1 file changed, 14 insertions(+), 5 deletions(-)
diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
index 7b911b5ed9..34cd473150 100644
--- a/xen/common/page_alloc.c
+++ b/xen/common/page_alloc.c
@@ -160,6 +160,7 @@
#endif
#define PGC_no_buddy_merge PGC_static
+#define PGC_preserved (PGC_extra | PGC_static | PGC_broken)
#ifndef PGT_TYPE_INFO_INITIALIZER
#define PGT_TYPE_INFO_INITIALIZER 0
@@ -1427,12 +1428,11 @@ static bool mark_page_free(struct page_info *pg, mfn_t mfn)
{
case PGC_state_inuse:
BUG_ON(pg->count_info & PGC_broken);
- pg->count_info = PGC_state_free;
+ pg->count_info = PGC_state_free | (pg->count_info & PGC_preserved);
break;
case PGC_state_offlining:
- pg->count_info = (pg->count_info & PGC_broken) |
- PGC_state_offlined;
+ pg->count_info = (pg->count_info & PGC_preserved) | PGC_state_offlined;
pg_offlined = true;
break;
@@ -2366,7 +2366,7 @@ int assign_pages(
for ( i = 0; i < nr; i++ )
{
- ASSERT(!(pg[i].count_info & ~(PGC_extra | PGC_static)));
+ ASSERT(!(pg[i].count_info & ~PGC_preserved));
if ( pg[i].count_info & PGC_extra )
extra_pages++;
}
@@ -2426,7 +2426,7 @@ int assign_pages(
page_set_owner(&pg[i], d);
smp_wmb(); /* Domain pointer must be visible before updating refcnt. */
pg[i].count_info =
- (pg[i].count_info & (PGC_extra | PGC_static)) | PGC_allocated | 1;
+ (pg[i].count_info & PGC_preserved) | PGC_allocated | 1;
page_list_add_tail(&pg[i], page_to_list(d, &pg[i]));
}
@@ -2485,6 +2485,14 @@ struct page_info *alloc_domheap_pages(
}
if ( assign_page(pg, order, d, memflags) )
{
+ if ( memflags & MEMF_no_refcount )
+ {
+ unsigned long i;
+
+ for ( i = 0; i < (1UL << order); i++ )
+ pg[i].count_info &= ~PGC_extra;
+ }
+
free_heap_pages(pg, order, memflags & MEMF_no_scrub);
return NULL;
}
@@ -2539,6 +2547,7 @@ void free_domheap_pages(struct page_info *pg, unsigned int order)
{
ASSERT(d->extra_pages);
d->extra_pages--;
+ pg[i].count_info &= ~PGC_extra;
}
}
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* [PATCH v10 09/12] xen: add cache coloring allocator for domains
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
` (7 preceding siblings ...)
2024-11-19 14:13 ` [PATCH v10 08/12] xen/page_alloc: introduce preserved page flags macro Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-11-28 11:43 ` Jan Beulich
2024-11-19 14:13 ` [PATCH v10 10/12] xen/arm: add Xen cache colors command line parameter Carlo Nonato
` (2 subsequent siblings)
11 siblings, 1 reply; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Andrew Cooper,
Jan Beulich, Julien Grall, Stefano Stabellini, Bertrand Marquis,
Michal Orzel, Volodymyr Babchuk
Add a new memory page allocator that implements the cache coloring mechanism.
The allocation algorithm enforces equal frequency distribution of cache
partitions, following the coloring configuration of a domain. This allows
for an even utilization of cache sets for every domain.
Pages are stored in a color-indexed array of lists. Those lists are filled
by a simple init function which computes the color of each page.
When a domain requests a page, the allocator extracts the page from the list
with the maximum number of free pages among those that the domain can access,
given its coloring configuration.
The allocator can only handle requests of order-0 pages. This allows for
easier implementation and since cache coloring targets only embedded systems,
it's assumed not to be a major problem.
The buddy allocator must coexist with the colored one because the Xen heap
isn't colored. For this reason a new Kconfig option and a command line
parameter are added to let the user set the amount of memory reserved for
the buddy allocator. Even when cache coloring is enabled, this memory
isn't managed by the colored allocator.
Colored heap information is dumped in the dump_heap() debug-key function.
Based on original work from: Luca Miccio <lucmiccio@gmail.com>
Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
---
v10:
- stated explicit dependency on CONFIG_LLC_COLORING for buddy-alloc-size
- fix for MISRA rule 20.7 parenthesis
v9:
- added ASSERT(order == 0) when freeing a colored page
- moved buddy_alloc_size initialization logic in Kconfig
v8:
- requests that uses MEMF_* flags that can't be served are now going to fail
- free_color_heap_page() is called directly from free_heap_pages()
v7:
- requests to alloc_color_heap_page() now fail if MEMF_bits is used
v6:
- colored allocator functions are now static
v5:
- Carlo Nonato as the new author
- the colored allocator balances color usage for each domain and it searches
linearly only in the number of colors (FIXME removed)
- addedd scrub functionality
- removed stub functions (still requires some macro definition)
- addr_to_color turned to mfn_to_color for easier operations
- removed BUG_ON in init_color_heap_pages() in favor of panic()
- only non empty page lists are logged in dump_color_heap()
v4:
- moved colored allocator code after buddy allocator because it now has
some dependencies on buddy functions
- buddy_alloc_size is now used only by the colored allocator
- fixed a bug that allowed the buddy to merge pages when they were colored
- free_color_heap_page() now calls mark_page_free()
- free_color_heap_page() uses of the frametable array for faster searches
- added FIXME comment for the linear search in free_color_heap_page()
- removed alloc_color_domheap_page() to let the colored allocator exploit
some more buddy allocator code
- alloc_color_heap_page() now allocs min address pages first
- reduced the mess in end_boot_allocator(): use the first loop for
init_color_heap_pages()
- fixed page_list_add_prev() (list.h) since it was doing the opposite of
what it was supposed to do
- fixed page_list_add_prev() (non list.h) to check also for next existence
- removed unused page_list_add_next()
- moved p2m code in another patch
---
docs/misc/cache-coloring.rst | 37 ++++++
docs/misc/xen-command-line.pandoc | 14 +++
xen/arch/arm/include/asm/mm.h | 5 +
xen/common/Kconfig | 8 ++
xen/common/llc-coloring.c | 13 +++
xen/common/page_alloc.c | 188 +++++++++++++++++++++++++++++-
xen/include/xen/llc-coloring.h | 4 +
7 files changed, 265 insertions(+), 4 deletions(-)
diff --git a/docs/misc/cache-coloring.rst b/docs/misc/cache-coloring.rst
index c14b6ee2f4..25f015c2de 100644
--- a/docs/misc/cache-coloring.rst
+++ b/docs/misc/cache-coloring.rst
@@ -11,6 +11,9 @@ To compile LLC coloring support set ``CONFIG_LLC_COLORING=y``.
If needed, change the maximum number of colors with
``CONFIG_LLC_COLORS_ORDER=<n>``.
+If needed, change the buddy allocator reserved size with
+``CONFIG_BUDDY_ALLOCATOR_SIZE=<n>``.
+
Runtime configuration is done via `Command line parameters`_.
For DomUs follow `DomUs configuration`_.
@@ -108,6 +111,8 @@ Specific documentation is available at `docs/misc/xen-command-line.pandoc`.
+----------------------+-------------------------------+
| ``dom0-llc-colors`` | Dom0 color configuration |
+----------------------+-------------------------------+
+| ``buddy-alloc-size`` | Buddy allocator reserved size |
++----------------------+-------------------------------+
Colors selection format
***********************
@@ -195,6 +200,17 @@ the ``llc-colors`` option. For example:
**Note:** If no color configuration is provided for a domain, the default one,
which corresponds to all available colors is used instead.
+Colored allocator and buddy allocator
+*************************************
+
+The colored allocator distributes pages based on color configurations of
+domains so that each domains only gets pages of its own colors.
+The colored allocator is meant as an alternative to the buddy allocator because
+its allocation policy is by definition incompatible with the generic one. Since
+the Xen heap is not colored yet, we need to support the coexistence of the two
+allocators and some memory must be left for the buddy one. Buddy memory
+reservation is configured via Kconfig or via command-line.
+
Known issues and limitations
****************************
@@ -205,3 +221,24 @@ In the domain configuration, "xen,static-mem" allows memory to be statically
allocated to the domain. This isn't possible when LLC coloring is enabled,
because that memory can't be guaranteed to use only colors assigned to the
domain.
+
+Cache coloring is intended only for embedded systems
+####################################################
+
+The current implementation aims to satisfy the need of predictability in
+embedded systems with small amount of memory to be managed in a colored way.
+Given that, some shortcuts are taken in the development. Expect worse
+performances on larger systems.
+
+Colored allocator can only make use of order-0 pages
+####################################################
+
+The cache coloring technique relies on memory mappings and on the smallest
+mapping granularity to achieve the maximum number of colors (cache partitions)
+possible. This granularity is what is normally called a page and, in Xen
+terminology, the order-0 page is the smallest one. The fairly simple
+colored allocator currently implemented, makes use only of such pages.
+It must be said that a more complex one could, in theory, adopt higher order
+pages if the colors selection contained adjacent colors. Two subsequent colors,
+for example, can be represented by an order-1 page, four colors correspond to
+an order-2 page, etc.
diff --git a/docs/misc/xen-command-line.pandoc b/docs/misc/xen-command-line.pandoc
index bfdc8b0002..3a70c49c05 100644
--- a/docs/misc/xen-command-line.pandoc
+++ b/docs/misc/xen-command-line.pandoc
@@ -270,6 +270,20 @@ and not running softirqs. Reduce this if softirqs are not being run frequently
enough. Setting this to a high value may cause boot failure, particularly if
the NMI watchdog is also enabled.
+### buddy-alloc-size (arm64)
+> `= <size>`
+
+> Default: `64M`
+
+Amount of memory reserved for the buddy allocator when colored allocator is
+active. This option is available only when `CONFIG_LLC_COLORING` is enabled.
+The colored allocator is meant as an alternative to the buddy allocator,
+because its allocation policy is by definition incompatible with the generic
+one. Since the Xen heap systems is not colored yet, we need to support the
+coexistence of the two allocators for now. This parameter, which is optional
+and for expert only, it's used to set the amount of memory reserved to the
+buddy allocator.
+
### cet
= List of [ shstk=<bool>, ibt=<bool> ]
diff --git a/xen/arch/arm/include/asm/mm.h b/xen/arch/arm/include/asm/mm.h
index 5abd4b0d1c..c1a5ac7bee 100644
--- a/xen/arch/arm/include/asm/mm.h
+++ b/xen/arch/arm/include/asm/mm.h
@@ -145,6 +145,11 @@ struct page_info
#else
#define PGC_static 0
#endif
+#ifdef CONFIG_LLC_COLORING
+/* Page is cache colored */
+#define _PGC_colored PG_shift(4)
+#define PGC_colored PG_mask(1, 4)
+#endif
/* ... */
/* Page is broken? */
#define _PGC_broken PG_shift(7)
diff --git a/xen/common/Kconfig b/xen/common/Kconfig
index b4ec6893be..bbd4738d4e 100644
--- a/xen/common/Kconfig
+++ b/xen/common/Kconfig
@@ -537,4 +537,12 @@ config LLC_COLORS_ORDER
The default value corresponds to an 8 MiB 16-ways LLC, which should be
more than what's needed in the general case.
+config BUDDY_ALLOCATOR_SIZE
+ int "Buddy allocator reserved memory size (MiB)" if LLC_COLORING
+ default "0" if !LLC_COLORING
+ default "64"
+ help
+ Amount of memory reserved for the buddy allocator to serve Xen heap,
+ working alongside the colored one.
+
endmenu
diff --git a/xen/common/llc-coloring.c b/xen/common/llc-coloring.c
index 14e056fb10..6ba9fb72ca 100644
--- a/xen/common/llc-coloring.c
+++ b/xen/common/llc-coloring.c
@@ -28,6 +28,9 @@ static unsigned int __ro_after_init default_colors[NR_LLC_COLORS];
static unsigned int __initdata dom0_colors[NR_LLC_COLORS];
static unsigned int __initdata dom0_num_colors;
+#define mfn_color_mask (max_nr_colors - 1)
+#define mfn_to_color(mfn) (mfn_x(mfn) & mfn_color_mask)
+
/*
* Parse the coloring configuration given in the buf string, following the
* syntax below.
@@ -311,6 +314,16 @@ int __init domain_set_llc_colors_from_str(struct domain *d, const char *str)
return 0;
}
+unsigned int page_to_llc_color(const struct page_info *pg)
+{
+ return mfn_to_color(page_to_mfn(pg));
+}
+
+unsigned int get_max_nr_llc_colors(void)
+{
+ return max_nr_colors;
+}
+
/*
* Local variables:
* mode: C
diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
index 34cd473150..27431a9bd2 100644
--- a/xen/common/page_alloc.c
+++ b/xen/common/page_alloc.c
@@ -159,8 +159,12 @@
#define PGC_static 0
#endif
-#define PGC_no_buddy_merge PGC_static
-#define PGC_preserved (PGC_extra | PGC_static | PGC_broken)
+#ifndef PGC_colored
+#define PGC_colored 0
+#endif
+
+#define PGC_no_buddy_merge (PGC_static | PGC_colored)
+#define PGC_preserved (PGC_extra | PGC_static | PGC_broken | PGC_colored)
#ifndef PGT_TYPE_INFO_INITIALIZER
#define PGT_TYPE_INFO_INITIALIZER 0
@@ -1457,6 +1461,8 @@ static bool mark_page_free(struct page_info *pg, mfn_t mfn)
return pg_offlined;
}
+static void free_color_heap_page(struct page_info *pg, bool need_scrub);
+
/* Free 2^@order set of pages. */
static void free_heap_pages(
struct page_info *pg, unsigned int order, bool need_scrub)
@@ -1481,6 +1487,15 @@ static void free_heap_pages(
pg[i].count_info |= PGC_need_scrub;
poison_one_page(&pg[i]);
}
+
+ if ( pg->count_info & PGC_colored )
+ {
+ ASSERT(order == 0);
+
+ free_color_heap_page(pg, need_scrub);
+ spin_unlock(&heap_lock);
+ return;
+ }
}
avail[node][zone] += 1 << order;
@@ -1945,6 +1960,155 @@ static unsigned long avail_heap_pages(
return free_pages;
}
+/*************************
+ * COLORED SIDE-ALLOCATOR
+ *
+ * Pages are grouped by LLC color in lists which are globally referred to as the
+ * color heap. Lists are populated in end_boot_allocator().
+ * After initialization there will be N lists where N is the number of
+ * available colors on the platform.
+ */
+static struct page_list_head *__ro_after_init _color_heap;
+#define color_heap(color) (&_color_heap[color])
+
+static unsigned long *__ro_after_init free_colored_pages;
+
+/* Memory required for buddy allocator to work with colored one */
+static unsigned long __initdata buddy_alloc_size =
+ MB(CONFIG_BUDDY_ALLOCATOR_SIZE);
+size_param("buddy-alloc-size", buddy_alloc_size);
+
+#ifdef CONFIG_LLC_COLORING
+#define domain_num_llc_colors(d) ((d)->num_llc_colors)
+#define domain_llc_color(d, i) ((d)->llc_colors[i])
+#else
+#define domain_num_llc_colors(d) 0
+#define domain_llc_color(d, i) 0
+#endif
+
+static void free_color_heap_page(struct page_info *pg, bool need_scrub)
+{
+ unsigned int color;
+
+ color = page_to_llc_color(pg);
+ free_colored_pages[color]++;
+ /*
+ * Head insertion allows re-using cache-hot pages in configurations without
+ * sharing of colors.
+ */
+ page_list_add(pg, color_heap(color));
+}
+
+static struct page_info *alloc_color_heap_page(unsigned int memflags,
+ const struct domain *d)
+{
+ struct page_info *pg = NULL;
+ unsigned int i, color = 0;
+ unsigned long max = 0;
+ bool need_tlbflush = false;
+ uint32_t tlbflush_timestamp = 0;
+ bool need_scrub;
+
+ if ( memflags & ~(MEMF_no_refcount | MEMF_no_owner | MEMF_no_tlbflush |
+ MEMF_no_icache_flush | MEMF_no_scrub) )
+ return NULL;
+
+ spin_lock(&heap_lock);
+
+ for ( i = 0; i < domain_num_llc_colors(d); i++ )
+ {
+ unsigned long free = free_colored_pages[domain_llc_color(d, i)];
+
+ if ( free > max )
+ {
+ color = domain_llc_color(d, i);
+ pg = page_list_first(color_heap(color));
+ max = free;
+ }
+ }
+
+ if ( !pg )
+ {
+ spin_unlock(&heap_lock);
+ return NULL;
+ }
+
+ need_scrub = pg->count_info & PGC_need_scrub;
+ pg->count_info = PGC_state_inuse | (pg->count_info & PGC_preserved);
+ free_colored_pages[color]--;
+ page_list_del(pg, color_heap(color));
+
+ if ( !(memflags & MEMF_no_tlbflush) )
+ accumulate_tlbflush(&need_tlbflush, pg, &tlbflush_timestamp);
+
+ init_free_page_fields(pg);
+
+ spin_unlock(&heap_lock);
+
+ if ( !(memflags & MEMF_no_scrub) )
+ {
+ if ( need_scrub )
+ scrub_one_page(pg);
+ else
+ check_one_page(pg);
+ }
+
+ if ( need_tlbflush )
+ filtered_flush_tlb_mask(tlbflush_timestamp);
+
+ flush_page_to_ram(mfn_x(page_to_mfn(pg)),
+ !(memflags & MEMF_no_icache_flush));
+
+ return pg;
+}
+
+static void __init init_color_heap_pages(struct page_info *pg,
+ unsigned long nr_pages)
+{
+ unsigned int i;
+ bool need_scrub = opt_bootscrub == BOOTSCRUB_IDLE;
+
+ if ( buddy_alloc_size >= PAGE_SIZE )
+ {
+ unsigned long buddy_pages = min(PFN_DOWN(buddy_alloc_size), nr_pages);
+
+ init_heap_pages(pg, buddy_pages);
+ nr_pages -= buddy_pages;
+ buddy_alloc_size -= buddy_pages << PAGE_SHIFT;
+ pg += buddy_pages;
+ }
+
+ if ( !_color_heap )
+ {
+ unsigned int max_nr_colors = get_max_nr_llc_colors();
+
+ _color_heap = xmalloc_array(struct page_list_head, max_nr_colors);
+ free_colored_pages = xzalloc_array(unsigned long, max_nr_colors);
+ if ( !_color_heap || !free_colored_pages )
+ panic("Can't allocate colored heap. Buddy reserved size is too low");
+
+ for ( i = 0; i < max_nr_colors; i++ )
+ INIT_PAGE_LIST_HEAD(color_heap(i));
+ }
+
+ for ( i = 0; i < nr_pages; i++ )
+ {
+ pg[i].count_info = PGC_colored;
+ free_color_heap_page(&pg[i], need_scrub);
+ }
+}
+
+static void dump_color_heap(void)
+{
+ unsigned int color;
+
+ printk("Dumping color heap info\n");
+ for ( color = 0; color < get_max_nr_llc_colors(); color++ )
+ if ( free_colored_pages[color] > 0 )
+ printk("Color heap[%u]: %lu pages\n",
+ color, free_colored_pages[color]);
+}
+
void __init end_boot_allocator(void)
{
unsigned int i;
@@ -1964,7 +2128,13 @@ void __init end_boot_allocator(void)
for ( i = nr_bootmem_regions; i-- > 0; )
{
struct bootmem_region *r = &bootmem_region_list[i];
- if ( r->s < r->e )
+
+ if ( r->s >= r->e )
+ continue;
+
+ if ( llc_coloring_enabled )
+ init_color_heap_pages(mfn_to_page(_mfn(r->s)), r->e - r->s);
+ else
init_heap_pages(mfn_to_page(_mfn(r->s)), r->e - r->s);
}
nr_bootmem_regions = 0;
@@ -2460,7 +2630,14 @@ struct page_info *alloc_domheap_pages(
if ( memflags & MEMF_no_owner )
memflags |= MEMF_no_refcount;
- if ( !dma_bitsize )
+ /* Only domains are supported for coloring */
+ if ( d && llc_coloring_enabled )
+ {
+ /* Colored allocation must be done on 0 order */
+ if ( order || (pg = alloc_color_heap_page(memflags, d)) == NULL )
+ return NULL;
+ }
+ else if ( !dma_bitsize )
memflags &= ~MEMF_no_dma;
else if ( (dma_zone = bits_to_zone(dma_bitsize)) < zone_hi )
pg = alloc_heap_pages(dma_zone + 1, zone_hi, order, memflags, d);
@@ -2686,6 +2863,9 @@ static void cf_check dump_heap(unsigned char key)
continue;
printk("Node %d has %lu unscrubbed pages\n", i, node_need_scrub[i]);
}
+
+ if ( llc_coloring_enabled )
+ dump_color_heap();
}
static __init int cf_check register_heap_trigger(void)
diff --git a/xen/include/xen/llc-coloring.h b/xen/include/xen/llc-coloring.h
index 76d3cbae0e..90daf214cd 100644
--- a/xen/include/xen/llc-coloring.h
+++ b/xen/include/xen/llc-coloring.h
@@ -34,6 +34,10 @@ int domain_set_llc_colors(struct domain *d,
const struct xen_domctl_set_llc_colors *config);
int domain_set_llc_colors_from_str(struct domain *d, const char *str);
+struct page_info;
+unsigned int page_to_llc_color(const struct page_info *pg);
+unsigned int get_max_nr_llc_colors(void);
+
#endif /* __COLORING_H__ */
/*
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* [PATCH v10 10/12] xen/arm: add Xen cache colors command line parameter
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
` (8 preceding siblings ...)
2024-11-19 14:13 ` [PATCH v10 09/12] xen: add cache coloring allocator for domains Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-11-19 14:13 ` [PATCH v10 11/12] xen/arm: make consider_modules() available for xen relocation Carlo Nonato
2024-11-19 14:13 ` [PATCH v10 12/12] xen/arm: add cache coloring support for Xen Carlo Nonato
11 siblings, 0 replies; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Luca Miccio, Andrew Cooper,
Jan Beulich, Julien Grall, Stefano Stabellini, Carlo Nonato
From: Luca Miccio <lucmiccio@gmail.com>
Add a new command line parameter to configure Xen cache colors.
These colors are dumped together with other coloring info.
Benchmarking the VM interrupt response time provides an estimation of
LLC usage by Xen's most latency-critical runtime task. Results on Arm
Cortex-A53 on Xilinx Zynq UltraScale+ XCZU9EG show that one color, which
reserves 64 KiB of L2, is enough to attain best responsiveness:
- Xen 1 color latency: 3.1 us
- Xen 2 color latency: 3.1 us
Since this is the most common target for Arm cache coloring, the default
amount of Xen colors is set to one.
More colors are instead very likely to be needed on processors whose L1
cache is physically-indexed and physically-tagged, such as Cortex-A57.
In such cases, coloring applies to L1 also, and there typically are two
distinct L1-colors. Therefore, reserving only one color for Xen would
senselessly partitions a cache memory that is already private, i.e.
underutilize it.
Signed-off-by: Luca Miccio <lucmiccio@gmail.com>
Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
---
v10:
- no changes
v9:
- no changes
v8:
- added bound check on xen_colors in llc_coloring_init()
v7:
- removed XEN_DEFAULT_COLOR
- XEN_DEFAULT_NUM_COLORS is now used in a for loop to set xen default colors
---
docs/misc/cache-coloring.rst | 2 ++
docs/misc/xen-command-line.pandoc | 10 ++++++++++
xen/common/llc-coloring.c | 31 ++++++++++++++++++++++++++++++-
3 files changed, 42 insertions(+), 1 deletion(-)
diff --git a/docs/misc/cache-coloring.rst b/docs/misc/cache-coloring.rst
index 25f015c2de..f967df7360 100644
--- a/docs/misc/cache-coloring.rst
+++ b/docs/misc/cache-coloring.rst
@@ -113,6 +113,8 @@ Specific documentation is available at `docs/misc/xen-command-line.pandoc`.
+----------------------+-------------------------------+
| ``buddy-alloc-size`` | Buddy allocator reserved size |
+----------------------+-------------------------------+
+| ``xen-llc-colors`` | Xen color configuration |
++----------------------+-------------------------------+
Colors selection format
***********************
diff --git a/docs/misc/xen-command-line.pandoc b/docs/misc/xen-command-line.pandoc
index 3a70c49c05..992e1f993e 100644
--- a/docs/misc/xen-command-line.pandoc
+++ b/docs/misc/xen-command-line.pandoc
@@ -2923,6 +2923,16 @@ mode.
**WARNING: `x2apic_phys` is deprecated and superseded by `x2apic-mode`.
The latter takes precedence if both are set.**
+### xen-llc-colors (arm64)
+> `= List of [ <integer> | <integer>-<integer> ]`
+
+> Default: `0: the lowermost color`
+
+Specify Xen LLC color configuration. This options is available only when
+`CONFIG_LLC_COLORING` is enabled.
+Two colors are most likely needed on platforms where private caches are
+physically indexed, e.g. the L1 instruction cache of the Arm Cortex-A57.
+
### xenheap_megabytes (arm32)
> `= <size>`
diff --git a/xen/common/llc-coloring.c b/xen/common/llc-coloring.c
index 6ba9fb72ca..6e7a652884 100644
--- a/xen/common/llc-coloring.c
+++ b/xen/common/llc-coloring.c
@@ -12,6 +12,7 @@
#include <xen/param.h>
#define NR_LLC_COLORS (1U << CONFIG_LLC_COLORS_ORDER)
+#define XEN_DEFAULT_NUM_COLORS 1
bool __ro_after_init llc_coloring_enabled;
boolean_param("llc-coloring", llc_coloring_enabled);
@@ -28,6 +29,9 @@ static unsigned int __ro_after_init default_colors[NR_LLC_COLORS];
static unsigned int __initdata dom0_colors[NR_LLC_COLORS];
static unsigned int __initdata dom0_num_colors;
+static unsigned int __ro_after_init xen_colors[NR_LLC_COLORS];
+static unsigned int __ro_after_init xen_num_colors;
+
#define mfn_color_mask (max_nr_colors - 1)
#define mfn_to_color(mfn) (mfn_x(mfn) & mfn_color_mask)
@@ -86,7 +90,14 @@ static int __init parse_dom0_colors(const char *s)
}
custom_param("dom0-llc-colors", parse_dom0_colors);
-static void print_colors(const unsigned int colors[], unsigned int num_colors)
+static int __init parse_xen_colors(const char *s)
+{
+ return parse_color_config(s, xen_colors, ARRAY_SIZE(xen_colors),
+ &xen_num_colors);
+}
+custom_param("xen-llc-colors", parse_xen_colors);
+
+static void print_colors(const unsigned int *colors, unsigned int num_colors)
{
unsigned int i;
@@ -167,6 +178,22 @@ void __init llc_coloring_init(void)
for ( i = 0; i < max_nr_colors; i++ )
default_colors[i] = i;
+ if ( !xen_num_colors )
+ {
+ unsigned int i;
+
+ xen_num_colors = MIN(XEN_DEFAULT_NUM_COLORS, max_nr_colors);
+
+ printk(XENLOG_WARNING
+ "Xen LLC color config not found. Using first %u colors\n",
+ xen_num_colors);
+ for ( i = 0; i < xen_num_colors; i++ )
+ xen_colors[i] = i;
+ }
+ else if ( xen_num_colors > max_nr_colors ||
+ !check_colors(xen_colors, xen_num_colors) )
+ panic("Bad LLC color config for Xen\n");
+
arch_llc_coloring_init();
}
@@ -177,6 +204,8 @@ void dump_llc_coloring_info(void)
printk("LLC coloring info:\n");
printk(" Number of LLC colors supported: %u\n", max_nr_colors);
+ printk(" Xen LLC colors (%u): ", xen_num_colors);
+ print_colors(xen_colors, xen_num_colors);
}
void domain_dump_llc_colors(const struct domain *d)
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* [PATCH v10 11/12] xen/arm: make consider_modules() available for xen relocation
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
` (9 preceding siblings ...)
2024-11-19 14:13 ` [PATCH v10 10/12] xen/arm: add Xen cache colors command line parameter Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-12-02 8:59 ` Michal Orzel
2024-11-19 14:13 ` [PATCH v10 12/12] xen/arm: add cache coloring support for Xen Carlo Nonato
11 siblings, 1 reply; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Stefano Stabellini,
Julien Grall, Bertrand Marquis, Michal Orzel, Volodymyr Babchuk
Cache coloring must physically relocate Xen in order to color the hypervisor
and consider_modules() is a key function that is needed to find a new
available physical address.
672d67f339c0 ("xen/arm: Split MMU-specific setup_mm() and related code out")
moved consider_modules() under arm32. Move it to mmu/setup.c and make it
non-static so that it can be used outside.
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
---
v10:
- no changes
v9:
- no changes
v8:
- patch adapted to new changes to consider_modules()
v7:
- moved consider_modules() to arm/mmu/setup.c
v6:
- new patch
---
xen/arch/arm/arm32/mmu/mm.c | 95 +------------------------------
xen/arch/arm/include/asm/setup.h | 3 +
xen/arch/arm/mmu/setup.c | 97 ++++++++++++++++++++++++++++++++
3 files changed, 101 insertions(+), 94 deletions(-)
diff --git a/xen/arch/arm/arm32/mmu/mm.c b/xen/arch/arm/arm32/mmu/mm.c
index 063611412b..c5fcd19291 100644
--- a/xen/arch/arm/arm32/mmu/mm.c
+++ b/xen/arch/arm/arm32/mmu/mm.c
@@ -9,6 +9,7 @@
#include <asm/fixmap.h>
#include <asm/static-memory.h>
#include <asm/static-shmem.h>
+#include <asm/setup.h>
static unsigned long opt_xenheap_megabytes __initdata;
integer_param("xenheap_megabytes", opt_xenheap_megabytes);
@@ -31,100 +32,6 @@ static void __init setup_directmap_mappings(unsigned long base_mfn,
directmap_virt_end = XENHEAP_VIRT_START + nr_mfns * PAGE_SIZE;
}
-/*
- * Returns the end address of the highest region in the range s..e
- * with required size and alignment that does not conflict with the
- * modules from first_mod to nr_modules.
- *
- * For non-recursive callers first_mod should normally be 0 (all
- * modules and Xen itself) or 1 (all modules but not Xen).
- */
-static paddr_t __init consider_modules(paddr_t s, paddr_t e,
- uint32_t size, paddr_t align,
- int first_mod)
-{
- const struct membanks *reserved_mem = bootinfo_get_reserved_mem();
-#ifdef CONFIG_STATIC_SHM
- const struct membanks *shmem = bootinfo_get_shmem();
-#endif
- const struct bootmodules *mi = &bootinfo.modules;
- int i;
- int nr;
-
- s = (s+align-1) & ~(align-1);
- e = e & ~(align-1);
-
- if ( s > e || e - s < size )
- return 0;
-
- /* First check the boot modules */
- for ( i = first_mod; i < mi->nr_mods; i++ )
- {
- paddr_t mod_s = mi->module[i].start;
- paddr_t mod_e = mod_s + mi->module[i].size;
-
- if ( s < mod_e && mod_s < e )
- {
- mod_e = consider_modules(mod_e, e, size, align, i+1);
- if ( mod_e )
- return mod_e;
-
- return consider_modules(s, mod_s, size, align, i+1);
- }
- }
-
- /*
- * i is the current bootmodule we are evaluating, across all
- * possible kinds of bootmodules.
- *
- * When retrieving the corresponding reserved-memory addresses, we
- * need to index the reserved_mem bank starting from 0, and only counting
- * the reserved-memory modules. Hence, we need to use i - nr.
- */
- nr = mi->nr_mods;
- for ( ; i - nr < reserved_mem->nr_banks; i++ )
- {
- paddr_t r_s = reserved_mem->bank[i - nr].start;
- paddr_t r_e = r_s + reserved_mem->bank[i - nr].size;
-
- if ( s < r_e && r_s < e )
- {
- r_e = consider_modules(r_e, e, size, align, i + 1);
- if ( r_e )
- return r_e;
-
- return consider_modules(s, r_s, size, align, i + 1);
- }
- }
-
-#ifdef CONFIG_STATIC_SHM
- nr += reserved_mem->nr_banks;
- for ( ; i - nr < shmem->nr_banks; i++ )
- {
- paddr_t r_s, r_e;
-
- r_s = shmem->bank[i - nr].start;
-
- /* Shared memory banks can contain INVALID_PADDR as start */
- if ( INVALID_PADDR == r_s )
- continue;
-
- r_e = r_s + shmem->bank[i - nr].size;
-
- if ( s < r_e && r_s < e )
- {
- r_e = consider_modules(r_e, e, size, align, i + 1);
- if ( r_e )
- return r_e;
-
- return consider_modules(s, r_s, size, align, i + 1);
- }
- }
-#endif
-
- return e;
-}
-
/*
* Find a contiguous region that fits in the static heap region with
* required size and alignment, and return the end address of the region
diff --git a/xen/arch/arm/include/asm/setup.h b/xen/arch/arm/include/asm/setup.h
index 64c227d171..0c560d141f 100644
--- a/xen/arch/arm/include/asm/setup.h
+++ b/xen/arch/arm/include/asm/setup.h
@@ -89,6 +89,9 @@ struct init_info
unsigned int cpuid;
};
+paddr_t consider_modules(paddr_t s, paddr_t e, uint32_t size, paddr_t align,
+ int first_mod);
+
#endif
/*
* Local variables:
diff --git a/xen/arch/arm/mmu/setup.c b/xen/arch/arm/mmu/setup.c
index 9664e85ee6..1cf62390e3 100644
--- a/xen/arch/arm/mmu/setup.c
+++ b/xen/arch/arm/mmu/setup.c
@@ -6,7 +6,10 @@
*/
#include <xen/init.h>
+#include <xen/lib.h>
#include <xen/libfdt/libfdt.h>
+#include <xen/libfdt/libfdt-xen.h>
+#include <xen/llc-coloring.h>
#include <xen/sections.h>
#include <xen/sizes.h>
#include <xen/vmap.h>
@@ -222,6 +225,100 @@ static void xen_pt_enforce_wnx(void)
flush_xen_tlb_local();
}
+/*
+ * Returns the end address of the highest region in the range s..e
+ * with required size and alignment that does not conflict with the
+ * modules from first_mod to nr_modules.
+ *
+ * For non-recursive callers first_mod should normally be 0 (all
+ * modules and Xen itself) or 1 (all modules but not Xen).
+ */
+paddr_t __init consider_modules(paddr_t s, paddr_t e,
+ uint32_t size, paddr_t align,
+ int first_mod)
+{
+ const struct membanks *reserved_mem = bootinfo_get_reserved_mem();
+#ifdef CONFIG_STATIC_SHM
+ const struct membanks *shmem = bootinfo_get_shmem();
+#endif
+ const struct bootmodules *mi = &bootinfo.modules;
+ int i;
+ int nr;
+
+ s = (s+align-1) & ~(align-1);
+ e = e & ~(align-1);
+
+ if ( s > e || e - s < size )
+ return 0;
+
+ /* First check the boot modules */
+ for ( i = first_mod; i < mi->nr_mods; i++ )
+ {
+ paddr_t mod_s = mi->module[i].start;
+ paddr_t mod_e = mod_s + mi->module[i].size;
+
+ if ( s < mod_e && mod_s < e )
+ {
+ mod_e = consider_modules(mod_e, e, size, align, i+1);
+ if ( mod_e )
+ return mod_e;
+
+ return consider_modules(s, mod_s, size, align, i+1);
+ }
+ }
+
+ /*
+ * i is the current bootmodule we are evaluating, across all
+ * possible kinds of bootmodules.
+ *
+ * When retrieving the corresponding reserved-memory addresses, we
+ * need to index the reserved_mem bank starting from 0, and only counting
+ * the reserved-memory modules. Hence, we need to use i - nr.
+ */
+ nr = mi->nr_mods;
+ for ( ; i - nr < reserved_mem->nr_banks; i++ )
+ {
+ paddr_t r_s = reserved_mem->bank[i - nr].start;
+ paddr_t r_e = r_s + reserved_mem->bank[i - nr].size;
+
+ if ( s < r_e && r_s < e )
+ {
+ r_e = consider_modules(r_e, e, size, align, i + 1);
+ if ( r_e )
+ return r_e;
+
+ return consider_modules(s, r_s, size, align, i + 1);
+ }
+ }
+
+#ifdef CONFIG_STATIC_SHM
+ nr += reserved_mem->nr_banks;
+ for ( ; i - nr < shmem->nr_banks; i++ )
+ {
+ paddr_t r_s, r_e;
+
+ r_s = shmem->bank[i - nr].start;
+
+ /* Shared memory banks can contain INVALID_PADDR as start */
+ if ( INVALID_PADDR == r_s )
+ continue;
+
+ r_e = r_s + shmem->bank[i - nr].size;
+
+ if ( s < r_e && r_s < e )
+ {
+ r_e = consider_modules(r_e, e, size, align, i + 1);
+ if ( r_e )
+ return r_e;
+
+ return consider_modules(s, r_s, size, align, i + 1);
+ }
+ }
+#endif
+
+ return e;
+}
+
/*
* Boot-time pagetable setup.
* Changes here may need matching changes in head.S
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* [PATCH v10 12/12] xen/arm: add cache coloring support for Xen
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
` (10 preceding siblings ...)
2024-11-19 14:13 ` [PATCH v10 11/12] xen/arm: make consider_modules() available for xen relocation Carlo Nonato
@ 2024-11-19 14:13 ` Carlo Nonato
2024-11-28 11:53 ` Jan Beulich
11 siblings, 1 reply; 36+ messages in thread
From: Carlo Nonato @ 2024-11-19 14:13 UTC (permalink / raw)
To: xen-devel
Cc: andrea.bastoni, marco.solieri, Carlo Nonato, Stefano Stabellini,
Julien Grall, Bertrand Marquis, Michal Orzel, Volodymyr Babchuk,
Andrew Cooper, Jan Beulich
Add the cache coloring support for Xen physical space.
Since Xen must be relocated to a new physical space, some relocation
functionalities must be brought back:
- the virtual address of the new space is taken from 0c18fb76323b
("xen/arm: Remove unused BOOT_RELOC_VIRT_START").
- relocate_xen() and get_xen_paddr() are taken from f60658c6ae47
("xen/arm: Stop relocating Xen").
setup_pagetables() must be adapted for coloring and for relocation. Runtime
page tables are used to map the colored space, but they are also linked in
boot tables so that the new space is temporarily available for relocation.
This implies that Xen protection must happen after the copy.
Finally, since the alternative framework needs to remap the Xen text and
inittext sections, this operation must be done in a coloring-aware way.
The function xen_remap_colored() is introduced for that.
Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
---
v10:
- no changes
v9:
- patch adapted to changes to setup_pagetables()
v8:
- moved xen_colored_map_size() to arm/llc-coloring.c
v7:
- added BUG_ON() checks to arch_llc_coloring_init() and
create_llc_coloring_mappings()
v6:
- squashed with BOOT_RELOC_VIRT_START patch
- consider_modules() moved in another patch
- removed psci and smpboot code because of new idmap work already handles that
- moved xen_remap_colored() in alternative.c since it's only used there
- removed xen_colored_temp[] in favor of xen_xenmap[] usage for mapping
- use of boot_module_find_by_kind() to remove the need of extra parameter in
setup_pagetables()
- moved get_xen_paddr() in arm/llc-coloring.c since it's only used there
v5:
- FIXME: consider_modules copy pasted since it got moved
v4:
- removed set_value_for_secondary() because it was wrongly cleaning cache
- relocate_xen() now calls switch_ttbr_id()
---
xen/arch/arm/alternative.c | 30 +++++++-
xen/arch/arm/arm64/mmu/head.S | 58 +++++++++++++++-
xen/arch/arm/arm64/mmu/mm.c | 28 ++++++--
xen/arch/arm/include/asm/mmu/layout.h | 3 +
xen/arch/arm/llc-coloring.c | 63 +++++++++++++++++
xen/arch/arm/mmu/setup.c | 98 +++++++++++++++++++++++----
xen/arch/arm/setup.c | 10 ++-
xen/common/llc-coloring.c | 18 +++++
xen/include/xen/llc-coloring.h | 13 ++++
9 files changed, 300 insertions(+), 21 deletions(-)
diff --git a/xen/arch/arm/alternative.c b/xen/arch/arm/alternative.c
index d99b507093..0fcf4e451d 100644
--- a/xen/arch/arm/alternative.c
+++ b/xen/arch/arm/alternative.c
@@ -9,6 +9,7 @@
#include <xen/init.h>
#include <xen/types.h>
#include <xen/kernel.h>
+#include <xen/llc-coloring.h>
#include <xen/mm.h>
#include <xen/vmap.h>
#include <xen/smp.h>
@@ -191,6 +192,27 @@ static int __apply_alternatives_multi_stop(void *xenmap)
return 0;
}
+static void __init *xen_remap_colored(mfn_t xen_mfn, paddr_t xen_size)
+{
+ unsigned int i;
+ void *xenmap;
+ mfn_t *xen_colored_mfns, mfn;
+
+ xen_colored_mfns = xmalloc_array(mfn_t, xen_size >> PAGE_SHIFT);
+ if ( !xen_colored_mfns )
+ panic("Can't allocate LLC colored MFNs\n");
+
+ for_each_xen_colored_mfn ( xen_mfn, mfn, i )
+ {
+ xen_colored_mfns[i] = mfn;
+ }
+
+ xenmap = vmap(xen_colored_mfns, xen_size >> PAGE_SHIFT);
+ xfree(xen_colored_mfns);
+
+ return xenmap;
+}
+
/*
* This function should only be called during boot and before CPU0 jump
* into the idle_loop.
@@ -209,8 +231,12 @@ void __init apply_alternatives_all(void)
* The text and inittext section are read-only. So re-map Xen to
* be able to patch the code.
*/
- xenmap = __vmap(&xen_mfn, 1U << xen_order, 1, 1, PAGE_HYPERVISOR,
- VMAP_DEFAULT);
+ if ( llc_coloring_enabled )
+ xenmap = xen_remap_colored(xen_mfn, xen_size);
+ else
+ xenmap = __vmap(&xen_mfn, 1U << xen_order, 1, 1, PAGE_HYPERVISOR,
+ VMAP_DEFAULT);
+
/* Re-mapping Xen is not expected to fail during boot. */
BUG_ON(!xenmap);
diff --git a/xen/arch/arm/arm64/mmu/head.S b/xen/arch/arm/arm64/mmu/head.S
index 665a51a337..a1fc9a82f1 100644
--- a/xen/arch/arm/arm64/mmu/head.S
+++ b/xen/arch/arm/arm64/mmu/head.S
@@ -428,6 +428,61 @@ FUNC_LOCAL(fail)
b 1b
END(fail)
+/*
+ * Copy Xen to new location and switch TTBR
+ * x0 ttbr
+ * x1 source address
+ * x2 destination address
+ * x3 length
+ *
+ * Source and destination must be word aligned, length is rounded up
+ * to a 16 byte boundary.
+ *
+ * MUST BE VERY CAREFUL when saving things to RAM over the copy
+ */
+ENTRY(relocate_xen)
+ /*
+ * Copy 16 bytes at a time using:
+ * x9: counter
+ * x10: data
+ * x11: data
+ * x12: source
+ * x13: destination
+ */
+ mov x9, x3
+ mov x12, x1
+ mov x13, x2
+
+1: ldp x10, x11, [x12], #16
+ stp x10, x11, [x13], #16
+
+ subs x9, x9, #16
+ bgt 1b
+
+ /*
+ * Flush destination from dcache using:
+ * x9: counter
+ * x10: step
+ * x11: vaddr
+ *
+ * This is to ensure data is visible to the instruction cache
+ */
+ dsb sy
+
+ mov x9, x3
+ ldr x10, =dcache_line_bytes /* x10 := step */
+ ldr x10, [x10]
+ mov x11, x2
+
+1: dc cvac, x11
+
+ add x11, x11, x10
+ subs x9, x9, x10
+ bgt 1b
+
+ /* No need for dsb/isb because they are alredy done in switch_ttbr_id */
+ b switch_ttbr_id
+
/*
* Switch TTBR
*
@@ -453,7 +508,8 @@ FUNC(switch_ttbr_id)
/*
* 5) Flush I-cache
- * This should not be necessary but it is kept for safety.
+ * This should not be necessary in the general case, but it's needed
+ * for cache coloring because code is relocated in that case.
*/
ic iallu
isb
diff --git a/xen/arch/arm/arm64/mmu/mm.c b/xen/arch/arm/arm64/mmu/mm.c
index 671eaadbc1..3732d5897e 100644
--- a/xen/arch/arm/arm64/mmu/mm.c
+++ b/xen/arch/arm/arm64/mmu/mm.c
@@ -1,6 +1,7 @@
/* SPDX-License-Identifier: GPL-2.0-only */
#include <xen/init.h>
+#include <xen/llc-coloring.h>
#include <xen/mm.h>
#include <xen/pfn.h>
@@ -138,27 +139,46 @@ void update_boot_mapping(bool enable)
}
extern void switch_ttbr_id(uint64_t ttbr);
+extern void relocate_xen(uint64_t ttbr, void *src, void *dst, size_t len);
typedef void (switch_ttbr_fn)(uint64_t ttbr);
+typedef void (relocate_xen_fn)(uint64_t ttbr, void *src, void *dst, size_t len);
void __init switch_ttbr(uint64_t ttbr)
{
- vaddr_t id_addr = virt_to_maddr(switch_ttbr_id);
- switch_ttbr_fn *fn = (switch_ttbr_fn *)id_addr;
+ vaddr_t vaddr, id_addr;
lpae_t pte;
+ if ( llc_coloring_enabled )
+ vaddr = (vaddr_t)relocate_xen;
+ else
+ vaddr = (vaddr_t)switch_ttbr_id;
+
+ id_addr = virt_to_maddr(vaddr);
+
/* Enable the identity mapping in the boot page tables */
update_identity_mapping(true);
/* Enable the identity mapping in the runtime page tables */
- pte = pte_of_xenaddr((vaddr_t)switch_ttbr_id);
+ pte = pte_of_xenaddr(vaddr);
pte.pt.table = 1;
pte.pt.xn = 0;
pte.pt.ro = 1;
write_pte(&xen_third_id[third_table_offset(id_addr)], pte);
/* Switch TTBR */
- fn(ttbr);
+ if ( llc_coloring_enabled )
+ {
+ relocate_xen_fn *fn = (relocate_xen_fn *)id_addr;
+
+ fn(ttbr, _start, (void *)BOOT_RELOC_VIRT_START, _end - _start);
+ }
+ else
+ {
+ switch_ttbr_fn *fn = (switch_ttbr_fn *)id_addr;
+
+ fn(ttbr);
+ }
/*
* Disable the identity mapping in the runtime page tables.
diff --git a/xen/arch/arm/include/asm/mmu/layout.h b/xen/arch/arm/include/asm/mmu/layout.h
index a3b546465b..19c0ec63a5 100644
--- a/xen/arch/arm/include/asm/mmu/layout.h
+++ b/xen/arch/arm/include/asm/mmu/layout.h
@@ -30,6 +30,7 @@
* 10M - 12M Fixmap: special-purpose 4K mapping slots
* 12M - 16M Early boot mapping of FDT
* 16M - 18M Livepatch vmap (if compiled in)
+ * 16M - 24M Cache-colored Xen text, data, bss (temporary, if compiled in)
*
* 1G - 2G VMAP: ioremap and early_ioremap
*
@@ -74,6 +75,8 @@
#define BOOT_FDT_VIRT_START (FIXMAP_VIRT_START + FIXMAP_VIRT_SIZE)
#define BOOT_FDT_VIRT_SIZE _AT(vaddr_t, MB(4))
+#define BOOT_RELOC_VIRT_START (BOOT_FDT_VIRT_START + BOOT_FDT_VIRT_SIZE)
+
#ifdef CONFIG_LIVEPATCH
#define LIVEPATCH_VMAP_START (BOOT_FDT_VIRT_START + BOOT_FDT_VIRT_SIZE)
#define LIVEPATCH_VMAP_SIZE _AT(vaddr_t, MB(2))
diff --git a/xen/arch/arm/llc-coloring.c b/xen/arch/arm/llc-coloring.c
index a7e0907816..18362d2317 100644
--- a/xen/arch/arm/llc-coloring.c
+++ b/xen/arch/arm/llc-coloring.c
@@ -10,6 +10,7 @@
#include <asm/processor.h>
#include <asm/sysregs.h>
+#include <asm/setup.h>
/* Return the LLC way size by probing the hardware */
unsigned int __init get_llc_way_size(void)
@@ -63,8 +64,70 @@ unsigned int __init get_llc_way_size(void)
return line_size * num_sets;
}
+/**
+ * get_xen_paddr - get physical address to relocate Xen to
+ *
+ * Xen is relocated to as near to the top of RAM as possible and
+ * aligned to a XEN_PADDR_ALIGN boundary.
+ */
+static paddr_t __init get_xen_paddr(paddr_t xen_size)
+{
+ const struct membanks *mem = bootinfo_get_mem();
+ paddr_t min_size, paddr = 0;
+ unsigned int i;
+
+ min_size = (xen_size + (XEN_PADDR_ALIGN-1)) & ~(XEN_PADDR_ALIGN-1);
+
+ /* Find the highest bank with enough space. */
+ for ( i = 0; i < mem->nr_banks; i++ )
+ {
+ const struct membank *bank = &mem->bank[i];
+ paddr_t s, e;
+
+ if ( bank->size >= min_size )
+ {
+ e = consider_modules(bank->start, bank->start + bank->size,
+ min_size, XEN_PADDR_ALIGN, 0);
+ if ( !e )
+ continue;
+
+#ifdef CONFIG_ARM_32
+ /* Xen must be under 4GB */
+ if ( e > GB(4) )
+ e = GB(4);
+ if ( e < bank->start )
+ continue;
+#endif
+
+ s = e - min_size;
+
+ if ( s > paddr )
+ paddr = s;
+ }
+ }
+
+ if ( !paddr )
+ panic("Not enough memory to relocate Xen\n");
+
+ printk("Placing Xen at 0x%"PRIpaddr"-0x%"PRIpaddr"\n",
+ paddr, paddr + min_size);
+
+ return paddr;
+}
+
+static paddr_t __init xen_colored_map_size(void)
+{
+ return ROUNDUP((_end - _start) * get_max_nr_llc_colors(), XEN_PADDR_ALIGN);
+}
+
void __init arch_llc_coloring_init(void)
{
+ struct bootmodule *xen_bootmodule = boot_module_find_by_kind(BOOTMOD_XEN);
+
+ BUG_ON(!xen_bootmodule);
+
+ xen_bootmodule->size = xen_colored_map_size();
+ xen_bootmodule->start = get_xen_paddr(xen_bootmodule->size);
}
/*
diff --git a/xen/arch/arm/mmu/setup.c b/xen/arch/arm/mmu/setup.c
index 1cf62390e3..bf6334dbd4 100644
--- a/xen/arch/arm/mmu/setup.c
+++ b/xen/arch/arm/mmu/setup.c
@@ -23,6 +23,9 @@
#undef virt_to_mfn
#define virt_to_mfn(va) _mfn(__virt_to_mfn(va))
+#define virt_to_reloc_virt(virt) \
+ (((vaddr_t)virt) - XEN_VIRT_START + BOOT_RELOC_VIRT_START)
+
/* Main runtime page tables */
/*
@@ -72,6 +75,7 @@ static void __init __maybe_unused build_assertions(void)
/* 2MB aligned regions */
BUILD_BUG_ON(XEN_VIRT_START & ~SECOND_MASK);
BUILD_BUG_ON(FIXMAP_ADDR(0) & ~SECOND_MASK);
+ BUILD_BUG_ON(BOOT_RELOC_VIRT_START & ~SECOND_MASK);
/* 1GB aligned regions */
#ifdef CONFIG_ARM_32
BUILD_BUG_ON(XENHEAP_VIRT_START & ~FIRST_MASK);
@@ -141,6 +145,9 @@ static void __init __maybe_unused build_assertions(void)
lpae_t __init pte_of_xenaddr(vaddr_t va)
{
+ if ( llc_coloring_enabled )
+ va = virt_to_reloc_virt(va);
+
return mfn_to_xen_entry(virt_to_mfn(va), MT_NORMAL);
}
@@ -319,9 +326,44 @@ paddr_t __init consider_modules(paddr_t s, paddr_t e,
return e;
}
+static void __init create_llc_coloring_mappings(void)
+{
+ lpae_t pte;
+ unsigned int i;
+ struct bootmodule *xen_bootmodule = boot_module_find_by_kind(BOOTMOD_XEN);
+ mfn_t start_mfn = maddr_to_mfn(xen_bootmodule->start), mfn;
+
+ for_each_xen_colored_mfn ( start_mfn, mfn, i )
+ {
+ pte = mfn_to_xen_entry(mfn, MT_NORMAL);
+ pte.pt.table = 1; /* level 3 mappings always have this bit set */
+ xen_xenmap[i] = pte;
+ }
+
+ for ( i = 0; i < XEN_NR_ENTRIES(2); i++ )
+ {
+ vaddr_t va = BOOT_RELOC_VIRT_START + (i << XEN_PT_LEVEL_SHIFT(2));
+
+ pte = mfn_to_xen_entry(virt_to_mfn(xen_xenmap +
+ i * XEN_PT_LPAE_ENTRIES),
+ MT_NORMAL);
+ pte.pt.table = 1;
+ write_pte(&boot_second[second_table_offset(va)], pte);
+ }
+}
+
/*
- * Boot-time pagetable setup.
+ * Boot-time pagetable setup with coloring support
* Changes here may need matching changes in head.S
+ *
+ * The cache coloring support consists of:
+ * - Create colored mapping that conforms to Xen color selection in xen_xenmap[]
+ * - Link the mapping in boot page tables using BOOT_RELOC_VIRT_START as vaddr
+ * - pte_of_xenaddr() takes care of translating addresses to the new space
+ * during runtime page tables creation
+ * - Relocate xen and update TTBR with the new address in the colored space
+ * (see switch_ttbr())
+ * - Protect the new space
*/
void __init setup_pagetables(void)
{
@@ -329,6 +371,9 @@ void __init setup_pagetables(void)
lpae_t pte, *p;
int i;
+ if ( llc_coloring_enabled )
+ create_llc_coloring_mappings();
+
arch_setup_page_tables();
#ifdef CONFIG_ARM_64
@@ -356,13 +401,7 @@ void __init setup_pagetables(void)
break;
pte = pte_of_xenaddr(va);
pte.pt.table = 1; /* third level mappings always have this bit set */
- if ( is_kernel_text(va) || is_kernel_inittext(va) )
- {
- pte.pt.xn = 0;
- pte.pt.ro = 1;
- }
- if ( is_kernel_rodata(va) )
- pte.pt.ro = 1;
+ pte.pt.xn = 0; /* Permissions will be enforced later. Allow execution */
xen_xenmap[i] = pte;
}
@@ -388,13 +427,48 @@ void __init setup_pagetables(void)
ttbr = virt_to_maddr(cpu0_pgtable);
#endif
- switch_ttbr(ttbr);
-
- xen_pt_enforce_wnx();
-
#ifdef CONFIG_ARM_32
per_cpu(xen_pgtable, 0) = cpu0_pgtable;
#endif
+
+ if ( llc_coloring_enabled )
+ ttbr = virt_to_maddr(virt_to_reloc_virt(THIS_CPU_PGTABLE));
+
+ switch_ttbr(ttbr);
+
+ /* Protect Xen */
+ for ( i = 0; i < XEN_NR_ENTRIES(3); i++ )
+ {
+ vaddr_t va = XEN_VIRT_START + (i << PAGE_SHIFT);
+ lpae_t *entry = xen_xenmap + i;
+
+ if ( !is_kernel(va) )
+ break;
+
+ pte = read_atomic(entry);
+
+ if ( is_kernel_text(va) || is_kernel_inittext(va) )
+ {
+ pte.pt.xn = 0;
+ pte.pt.ro = 1;
+ } else if ( is_kernel_rodata(va) ) {
+ pte.pt.ro = 1;
+ pte.pt.xn = 1;
+ } else {
+ pte.pt.xn = 1;
+ pte.pt.ro = 0;
+ }
+
+ write_pte(entry, pte);
+ }
+
+ /*
+ * We modified live page-tables. Ensure the TLBs are invalidated
+ * before setting enforcing the WnX permissions.
+ */
+ flush_xen_tlb_local();
+
+ xen_pt_enforce_wnx();
}
void *__init arch_vmap_virt_end(void)
diff --git a/xen/arch/arm/setup.c b/xen/arch/arm/setup.c
index 84fecaabea..3b3fb49fbd 100644
--- a/xen/arch/arm/setup.c
+++ b/xen/arch/arm/setup.c
@@ -304,8 +304,6 @@ void asmlinkage __init start_xen(unsigned long fdt_paddr)
/* Initialize traps early allow us to get backtrace when an error occurred */
init_traps();
- setup_pagetables();
-
smp_clear_cpu_maps();
device_tree_flattened = early_fdt_map(fdt_paddr);
@@ -329,6 +327,14 @@ void asmlinkage __init start_xen(unsigned long fdt_paddr)
llc_coloring_init();
+ /*
+ * Page tables must be setup after LLC coloring initialization because
+ * coloring info are required in order to create colored mappings
+ */
+ setup_pagetables();
+ /* Device-tree was mapped in boot page tables, remap it in the new tables */
+ device_tree_flattened = early_fdt_map(fdt_paddr);
+
setup_mm();
vm_init();
diff --git a/xen/common/llc-coloring.c b/xen/common/llc-coloring.c
index 6e7a652884..b80a47f8ad 100644
--- a/xen/common/llc-coloring.c
+++ b/xen/common/llc-coloring.c
@@ -34,6 +34,8 @@ static unsigned int __ro_after_init xen_num_colors;
#define mfn_color_mask (max_nr_colors - 1)
#define mfn_to_color(mfn) (mfn_x(mfn) & mfn_color_mask)
+#define get_mfn_with_color(mfn, color) \
+ (_mfn((mfn_x(mfn) & ~mfn_color_mask) | (color)))
/*
* Parse the coloring configuration given in the buf string, following the
@@ -353,6 +355,22 @@ unsigned int get_max_nr_llc_colors(void)
return max_nr_colors;
}
+mfn_t __init xen_colored_mfn(mfn_t mfn)
+{
+ unsigned int i, color = mfn_to_color(mfn);
+
+ for ( i = 0; i < xen_num_colors; i++ )
+ {
+ if ( color == xen_colors[i] )
+ return mfn;
+ else if ( color < xen_colors[i] )
+ return get_mfn_with_color(mfn, xen_colors[i]);
+ }
+
+ /* Jump to next color space (max_nr_colors mfns) and use the first color */
+ return get_mfn_with_color(mfn_add(mfn, max_nr_colors), xen_colors[0]);
+}
+
/*
* Local variables:
* mode: C
diff --git a/xen/include/xen/llc-coloring.h b/xen/include/xen/llc-coloring.h
index 90daf214cd..bf3e685d63 100644
--- a/xen/include/xen/llc-coloring.h
+++ b/xen/include/xen/llc-coloring.h
@@ -27,6 +27,17 @@ static inline void domain_dump_llc_colors(const struct domain *d) {}
static inline void domain_llc_coloring_free(struct domain *d) {}
#endif
+/**
+ * Iterate over each Xen mfn in the colored space.
+ * @start_mfn: the first mfn that needs to be colored.
+ * @mfn: the current mfn.
+ * @i: loop index.
+ */
+#define for_each_xen_colored_mfn(start_mfn, mfn, i) \
+ for ( i = 0, mfn = xen_colored_mfn(start_mfn); \
+ i < (_end - _start) >> PAGE_SHIFT; \
+ i++, mfn = xen_colored_mfn(mfn_add(mfn, 1)) )
+
unsigned int get_llc_way_size(void);
void arch_llc_coloring_init(void);
int dom0_set_llc_colors(struct domain *d);
@@ -38,6 +49,8 @@ struct page_info;
unsigned int page_to_llc_color(const struct page_info *pg);
unsigned int get_max_nr_llc_colors(void);
+mfn_t xen_colored_mfn(mfn_t mfn);
+
#endif /* __COLORING_H__ */
/*
--
2.43.0
^ permalink raw reply related [flat|nested] 36+ messages in thread
* Re: [PATCH v10 06/12] tools: add support for cache coloring configuration
2024-11-19 14:13 ` [PATCH v10 06/12] tools: add support for cache coloring configuration Carlo Nonato
@ 2024-11-20 16:25 ` Anthony PERARD
2024-11-21 11:48 ` Carlo Nonato
2024-11-20 16:48 ` Anthony PERARD
1 sibling, 1 reply; 36+ messages in thread
From: Anthony PERARD @ 2024-11-20 16:25 UTC (permalink / raw)
To: Carlo Nonato; +Cc: xen-devel, andrea.bastoni, marco.solieri, Juergen Gross
Hi Carlo,
On Tue, Nov 19, 2024 at 03:13:23PM +0100, Carlo Nonato wrote:
> diff --git a/tools/libs/ctrl/xc_domain.c b/tools/libs/ctrl/xc_domain.c
> index e3538ec0ba..4ed339e6e4 100644
> --- a/tools/libs/ctrl/xc_domain.c
> +++ b/tools/libs/ctrl/xc_domain.c
> @@ -2195,6 +2195,41 @@ int xc_domain_soft_reset(xc_interface *xch,
> domctl.domain = domid;
> return do_domctl(xch, &domctl);
> }
> +
> +int xc_domain_set_llc_colors(xc_interface *xch, uint32_t domid,
> + const unsigned int *llc_colors,
The hypercall seems to take a buffer of `uint32_t` for `llc_colors`, but
the parameter of the function is `unsigned int` instead, is there any
reason for them to be different?
> + unsigned int num_llc_colors)
> +{
> + struct xen_domctl domctl = {};
> + DECLARE_HYPERCALL_BUFFER(uint32_t, local);
> + int ret = -1;
> +
> + if ( num_llc_colors )
> + {
> + size_t bytes = sizeof(uint32_t) * num_llc_colors;
> +
> + local = xc_hypercall_buffer_alloc(xch, local, bytes);
> + if ( local == NULL )
> + {
> + PERROR("Could not allocate LLC colors for set_llc_colors");
> + ret = -ENOMEM;
I think we are supposed to return -1 in case of error, see:
https://elixir.bootlin.com/xen/v4.19.0/source/tools/include/xenctrl.h#L101
And there's nothing else to do, xc_hypercall_buffer_alloc() should
already have set `errno`, and PERROR() preserves it.
> + goto out;
> + }
Thanks,
--
Anthony Perard | Vates XCP-ng Developer
XCP-ng & Xen Orchestra - Vates solutions
web: https://vates.tech
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 06/12] tools: add support for cache coloring configuration
2024-11-19 14:13 ` [PATCH v10 06/12] tools: add support for cache coloring configuration Carlo Nonato
2024-11-20 16:25 ` Anthony PERARD
@ 2024-11-20 16:48 ` Anthony PERARD
1 sibling, 0 replies; 36+ messages in thread
From: Anthony PERARD @ 2024-11-20 16:48 UTC (permalink / raw)
To: Carlo Nonato; +Cc: xen-devel, andrea.bastoni, marco.solieri, Juergen Gross
On Tue, Nov 19, 2024 at 03:13:23PM +0100, Carlo Nonato wrote:
> tools/libs/light/libxl_types.idl | 1 +
An other thing I've completely forgot, with this IDL changed there will
be a need to regen some other files in the repo, namely:
tools/golang/xenlight/helpers.gen.go
tools/golang/xenlight/types.gen.go
Running this following command should be enough, even if you don't have
golang installed.
make -C tools/golang/xenlight helpers.gen.go types.gen.go
If you send an update of the patch series, it would be nice to include
the regenerated files.
Cheers,
--
Anthony Perard | Vates XCP-ng Developer
XCP-ng & Xen Orchestra - Vates solutions
web: https://vates.tech
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 06/12] tools: add support for cache coloring configuration
2024-11-20 16:25 ` Anthony PERARD
@ 2024-11-21 11:48 ` Carlo Nonato
0 siblings, 0 replies; 36+ messages in thread
From: Carlo Nonato @ 2024-11-21 11:48 UTC (permalink / raw)
To: Anthony PERARD; +Cc: xen-devel, andrea.bastoni, marco.solieri, Juergen Gross
Hi Anthony,
On Wed, Nov 20, 2024 at 5:25 PM Anthony PERARD
<anthony.perard@vates.tech> wrote:
>
> Hi Carlo,
>
> On Tue, Nov 19, 2024 at 03:13:23PM +0100, Carlo Nonato wrote:
> > diff --git a/tools/libs/ctrl/xc_domain.c b/tools/libs/ctrl/xc_domain.c
> > index e3538ec0ba..4ed339e6e4 100644
> > --- a/tools/libs/ctrl/xc_domain.c
> > +++ b/tools/libs/ctrl/xc_domain.c
> > @@ -2195,6 +2195,41 @@ int xc_domain_soft_reset(xc_interface *xch,
> > domctl.domain = domid;
> > return do_domctl(xch, &domctl);
> > }
> > +
> > +int xc_domain_set_llc_colors(xc_interface *xch, uint32_t domid,
> > + const unsigned int *llc_colors,
>
> The hypercall seems to take a buffer of `uint32_t` for `llc_colors`, but
> the parameter of the function is `unsigned int` instead, is there any
> reason for them to be different?
I don't remember anymore, but I think it should be uint32_t also for the
parameter.
> > + unsigned int num_llc_colors)
> > +{
> > + struct xen_domctl domctl = {};
> > + DECLARE_HYPERCALL_BUFFER(uint32_t, local);
> > + int ret = -1;
> > +
> > + if ( num_llc_colors )
> > + {
> > + size_t bytes = sizeof(uint32_t) * num_llc_colors;
> > +
> > + local = xc_hypercall_buffer_alloc(xch, local, bytes);
> > + if ( local == NULL )
> > + {
> > + PERROR("Could not allocate LLC colors for set_llc_colors");
> > + ret = -ENOMEM;
>
> I think we are supposed to return -1 in case of error, see:
> https://elixir.bootlin.com/xen/v4.19.0/source/tools/include/xenctrl.h#L101
>
> And there's nothing else to do, xc_hypercall_buffer_alloc() should
> already have set `errno`, and PERROR() preserves it.
You're right.
>> > + goto out;
> > + }
>
> Thanks,
>
> --
>
> Anthony Perard | Vates XCP-ng Developer
>
> XCP-ng & Xen Orchestra - Vates solutions
> web: https://vates.tech
(answering the other mail)
On Wed, Nov 20, 2024 at 5:48 PM Anthony PERARD
<anthony.perard@vates.tech> wrote:
>
> On Tue, Nov 19, 2024 at 03:13:23PM +0100, Carlo Nonato wrote:
> > tools/libs/light/libxl_types.idl | 1 +
>
> An other thing I've completely forgot, with this IDL changed there will
> be a need to regen some other files in the repo, namely:
> tools/golang/xenlight/helpers.gen.go
> tools/golang/xenlight/types.gen.go
>
> Running this following command should be enough, even if you don't have
> golang installed.
>
> make -C tools/golang/xenlight helpers.gen.go types.gen.go
>
> If you send an update of the patch series, it would be nice to include
> the regenerated files.
Yes, will do that.
> Cheers,
>
> --
>
> Anthony Perard | Vates XCP-ng Developer
>
> XCP-ng & Xen Orchestra - Vates solutions
>
> web: https://vates.tech
Thanks.
- Carlo
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 01/12] xen/common: add cache coloring common code
2024-11-19 14:13 ` [PATCH v10 01/12] xen/common: add cache coloring common code Carlo Nonato
@ 2024-11-27 10:48 ` Michal Orzel
2024-11-27 13:24 ` Carlo Nonato
2024-11-28 10:39 ` Jan Beulich
1 sibling, 1 reply; 36+ messages in thread
From: Michal Orzel @ 2024-11-27 10:48 UTC (permalink / raw)
To: Carlo Nonato, xen-devel
Cc: andrea.bastoni, marco.solieri, Andrew Cooper, Jan Beulich,
Julien Grall, Stefano Stabellini
On 19/11/2024 15:13, Carlo Nonato wrote:
>
>
> Last Level Cache (LLC) coloring allows to partition the cache in smaller
> chunks called cache colors.
>
> Since not all architectures can actually implement it, add a HAS_LLC_COLORING
> Kconfig option.
> LLC_COLORS_ORDER Kconfig option has a range maximum of 10 (2^10 = 1024)
> because that's the number of colors that fit in a 4 KiB page when integers
> are 4 bytes long.
>
> LLC colors are a property of the domain, so struct domain has to be extended.
>
> Based on original work from: Luca Miccio <lucmiccio@gmail.com>
>
> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
> ---
> v10:
> - fixed commit message to use LLC_COLORS_ORDER
> - added documentation to index.rst
> - moved check on CONFIG_NUMA in arch/arm/Kconfig (next patch)
> - fixed copyright line
> - fixed array type for colors parameter in print_colors()
> - added check on (way_size & ~PAGE_MASK)
> v9:
> - dropped _MAX_ from CONFIG_MAX_LLC_COLORS_ORDER
> v8:
> - minor documentation fixes
> - "llc-coloring=on" is inferred from "llc-nr-ways" and "llc-size" usage
> - turned CONFIG_NR_LLC_COLORS to CONFIG_MAX_LLC_COLORS_ORDER, base-2 exponent
> - moved Kconfig options to common/Kconfig
> - don't crash if computed max_nr_colors is too large
> v7:
> - SUPPORT.md changes added to this patch
> - extended documentation to better address applicability of cache coloring
> - "llc-nr-ways" and "llc-size" params introduced in favor of "llc-way-size"
> - moved dump_llc_coloring_info() call in 'm' keyhandler (pagealloc_info())
> v6:
> - moved almost all code in common
> - moved documentation in this patch
> - reintroduced range for CONFIG_NR_LLC_COLORS
> - reintroduced some stub functions to reduce the number of checks on
> llc_coloring_enabled
> - moved domain_llc_coloring_free() in same patch where allocation happens
> - turned "d->llc_colors" to pointer-to-const
> - llc_coloring_init() now returns void and panics if errors are found
> v5:
> - used - instead of _ for filenames
> - removed domain_create_llc_colored()
> - removed stub functions
> - coloring domain fields are now #ifdef protected
> v4:
> - Kconfig options moved to xen/arch
> - removed range for CONFIG_NR_LLC_COLORS
> - added "llc_coloring_enabled" global to later implement the boot-time
> switch
> - added domain_create_llc_colored() to be able to pass colors
> - added is_domain_llc_colored() macro
> ---
> SUPPORT.md | 7 ++
> docs/index.rst | 1 +
> docs/misc/cache-coloring.rst | 116 ++++++++++++++++++++++++++++++
> docs/misc/xen-command-line.pandoc | 37 ++++++++++
> xen/common/Kconfig | 21 ++++++
> xen/common/Makefile | 1 +
> xen/common/keyhandler.c | 3 +
> xen/common/llc-coloring.c | 115 +++++++++++++++++++++++++++++
> xen/common/page_alloc.c | 3 +
> xen/include/xen/llc-coloring.h | 37 ++++++++++
> xen/include/xen/sched.h | 5 ++
> 11 files changed, 346 insertions(+)
> create mode 100644 docs/misc/cache-coloring.rst
> create mode 100644 xen/common/llc-coloring.c
> create mode 100644 xen/include/xen/llc-coloring.h
>
> diff --git a/SUPPORT.md b/SUPPORT.md
> index 82239d0294..998faf5635 100644
> --- a/SUPPORT.md
> +++ b/SUPPORT.md
> @@ -401,6 +401,13 @@ by maintaining multiple physical to machine (p2m) memory mappings.
> Status, x86 HVM: Tech Preview
> Status, ARM: Tech Preview
>
> +### Cache coloring
> +
> +Allows to reserve Last Level Cache (LLC) partitions for Dom0, DomUs and Xen
> +itself.
> +
> + Status, Arm64: Experimental
> +
> ## Resource Management
>
> ### CPU Pools
> diff --git a/docs/index.rst b/docs/index.rst
> index ab051a0f38..fad4462bbb 100644
> --- a/docs/index.rst
> +++ b/docs/index.rst
> @@ -65,6 +65,7 @@ Documents in need of some rearranging.
> designs/launch/hyperlaunch-devicetree
> misc/xen-makefiles/makefiles
> misra/index
> + misc/cache-coloring
>
>
> Miscellanea
> diff --git a/docs/misc/cache-coloring.rst b/docs/misc/cache-coloring.rst
> new file mode 100644
> index 0000000000..0fe3830c40
> --- /dev/null
> +++ b/docs/misc/cache-coloring.rst
> @@ -0,0 +1,116 @@
You're missing SPDX tag at the top of the RST file:
.. SPDX-License-Identifier: CC-BY-4.0
> +Xen cache coloring user guide
> +=============================
> +
> +The cache coloring support in Xen allows to reserve Last Level Cache (LLC)
> +partitions for Dom0, DomUs and Xen itself. Currently only ARM64 is supported.
> +Cache coloring realizes per-set cache partitioning in software and is applicable
> +to shared LLCs as implemented in Cortex-A53, Cortex-A72 and similar CPUs.
> +
> +To compile LLC coloring support set ``CONFIG_LLC_COLORING=y``.
> +
> +If needed, change the maximum number of colors with
> +``CONFIG_LLC_COLORS_ORDER=<n>``.
> +
> +Runtime configuration is done via `Command line parameters`_.
> +
> +Background
> +**********
> +
> +Cache hierarchy of a modern multi-core CPU typically has first levels dedicated
> +to each core (hence using multiple cache units), while the last level is shared
> +among all of them. Such configuration implies that memory operations on one
> +core (e.g. running a DomU) are able to generate interference on another core
> +(e.g. hosting another DomU). Cache coloring realizes per-set cache-partitioning
> +in software and mitigates this, guaranteeing more predictable performances for
> +memory accesses.
> +Software-based cache coloring is particularly useful in those situations where
> +no hardware mechanisms (e.g., DSU-based way partitioning) are available to
> +partition caches. This is the case for e.g., Cortex-A53, A57 and A72 CPUs that
> +feature a L2 LLC cache shared among all cores.
> +
> +The key concept underlying cache coloring is a fragmentation of the memory
> +space into a set of sub-spaces called colors that are mapped to disjoint cache
> +partitions. Technically, the whole memory space is first divided into a number
> +of subsequent regions. Then each region is in turn divided into a number of
> +subsequent sub-colors. The generic i-th color is then obtained by all the
> +i-th sub-colors in each region.
> +
> +::
> +
> + Region j Region j+1
> + ..................... ............
> + . . .
> + . .
> + _ _ _______________ _ _____________________ _ _
> + | | | | | | |
> + | c_0 | c_1 | | c_n | c_0 | c_1 |
> + _ _ _|_____|_____|_ _ _|_____|_____|_____|_ _ _
> + : :
> + : :... ... .
> + : color 0
> + :........................... ... .
> + :
> + . . ..................................:
> +
> +How colors are actually defined depends on the function that maps memory to
> +cache lines. In case of physically-indexed, physically-tagged caches with linear
> +mapping, the set index is found by extracting some contiguous bits from the
> +physical address. This allows colors to be defined as shown in figure: they
> +appear in memory as subsequent blocks of equal size and repeats themselves after
> +``n`` different colors, where ``n`` is the total number of colors.
> +
> +If some kind of bit shuffling appears in the mapping function, then colors
> +assume a different layout in memory. Those kind of caches aren't supported by
> +the current implementation.
> +
> +**Note**: Finding the exact cache mapping function can be a really difficult
> +task since it's not always documented in the CPU manual. As said Cortex-A53, A57
> +and A72 are known to work with the current implementation.
> +
> +How to compute the number of colors
> +###################################
> +
> +Given the linear mapping from physical memory to cache lines for granted, the
> +number of available colors for a specific platform is computed using three
> +parameters:
> +
> +- the size of the LLC.
> +- the number of the LLC ways.
> +- the page size used by Xen.
> +
> +The first two parameters can be found in the processor manual, while the third
> +one is the minimum mapping granularity. Dividing the cache size by the number of
> +its ways we obtain the size of a way. Dividing this number by the page size,
> +the number of total cache colors is found. So for example an Arm Cortex-A53
> +with a 16-ways associative 1 MiB LLC can isolate up to 16 colors when pages are
> +4 KiB in size.
> +
> +Effective colors assignment
> +###########################
> +
> +When assigning colors, if one wants to avoid cache interference between two
> +domains, different colors needs to be used for their memory.
> +
> +Command line parameters
> +***********************
> +
> +Specific documentation is available at `docs/misc/xen-command-line.pandoc`.
> +
> ++----------------------+-------------------------------+
> +| **Parameter** | **Description** |
> ++----------------------+-------------------------------+
> +| ``llc-coloring`` | Enable coloring at runtime |
> ++----------------------+-------------------------------+
> +| ``llc-size`` | Set the LLC size |
> ++----------------------+-------------------------------+
> +| ``llc-nr-ways`` | Set the LLC number of ways |
> ++----------------------+-------------------------------+
> +
> +Auto-probing of LLC specs
> +#########################
> +
> +LLC size and number of ways are probed automatically by default.
> +
> +LLC specs can be manually set via the above command line parameters. This
> +bypasses any auto-probing and it's used to overcome failing situations, such as
> +flawed probing logic, or for debugging/testing purposes.
> diff --git a/docs/misc/xen-command-line.pandoc b/docs/misc/xen-command-line.pandoc
> index 293dbc1a95..abd8dae96f 100644
> --- a/docs/misc/xen-command-line.pandoc
> +++ b/docs/misc/xen-command-line.pandoc
> @@ -1708,6 +1708,43 @@ This option is intended for debugging purposes only. Enable MSR_DEBUGCTL.LBR
> in hypervisor context to be able to dump the Last Interrupt/Exception To/From
> record with other registers.
>
> +### llc-coloring (arm64)
> +> `= <boolean>`
> +
> +> Default: `false`
> +
> +Flag to enable or disable LLC coloring support at runtime. This option is
> +available only when `CONFIG_LLC_COLORING` is enabled. See the general
> +cache coloring documentation for more info.
> +
> +### llc-nr-ways (arm64)
> +> `= <integer>`
> +
> +> Default: `Obtained from hardware`
> +
> +Specify the number of ways of the Last Level Cache. This option is available
> +only when `CONFIG_LLC_COLORING` is enabled. LLC size and number of ways are used
> +to find the number of supported cache colors. By default the value is
> +automatically computed by probing the hardware, but in case of specific needs,
> +it can be manually set. Those include failing probing and debugging/testing
> +purposes so that it's possible to emulate platforms with different number of
> +supported colors. If set, also "llc-size" must be set, otherwise the default
> +will be used. Note that using both options implies "llc-coloring=on".
I can understand this decision, but ...
[...]
> + if ( llc_size && llc_nr_ways )
> + {
> + llc_coloring_enabled = true;
> + way_size = llc_size / llc_nr_ways;
> + }
> + else if ( !llc_coloring_enabled )
> + return;
the above code does not seem to be right. When debugging or in general it is useful to have on the cmdline:
llc-size=1M llc-nr-ways=16 llc-coloring=on
and be able to disable it by just switching between on/off in llc-coloring=. However, with your solution,
even if I specify llc-coloring=off, it will be enabled because I specified both llc-size and llc-nr-ways.
I think llc-coloring= should have a precedence.
> + else
> + {
> + way_size = get_llc_way_size();
> + if ( !way_size )
> + panic("LLC probing failed and 'llc-size' or 'llc-nr-ways' missing\n");
> + }
> +
> + if ( way_size & ~PAGE_MASK )
> + panic("LLC way size must be a multiple of PAGE_SIZE\n");
> +
> + /*
> + * The maximum number of colors must be a power of 2 in order to correctly
> + * map them to bits of an address.
> + */
> + max_nr_colors = way_size >> PAGE_SHIFT;
> +
> + if ( max_nr_colors & (max_nr_colors - 1) )
> + panic("Number of LLC colors (%u) isn't a power of 2\n", max_nr_colors);
> +
> + if ( max_nr_colors > NR_LLC_COLORS )
> + {
> + printk(XENLOG_WARNING
> + "Number of LLC colors (%u) too big. Using configured max %u\n",
> + max_nr_colors, NR_LLC_COLORS);
> + max_nr_colors = NR_LLC_COLORS;
> + } else if ( max_nr_colors < 2 )
CODING STYLE
> + panic("Number of LLC colors %u < 2\n", max_nr_colors);
> +
> + arch_llc_coloring_init();
> +}
> +
> +void dump_llc_coloring_info(void)
> +{
> + if ( !llc_coloring_enabled )
> + return;
> +
> + printk("LLC coloring info:\n");
> + printk(" Number of LLC colors supported: %u\n", max_nr_colors);
> +}
> +
> +void domain_dump_llc_colors(const struct domain *d)
> +{
> + if ( !llc_coloring_enabled )
> + return;
> +
> + printk("%u LLC colors: ", d->num_llc_colors);
> + print_colors(d->llc_colors, d->num_llc_colors);
> +}
> +
> +/*
> + * Local variables:
> + * mode: C
> + * c-file-style: "BSD"
> + * c-basic-offset: 4
> + * tab-width: 4
> + * indent-tabs-mode: nil
> + * End:
> + */
> diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
> index 33c8c917d9..7b911b5ed9 100644
> --- a/xen/common/page_alloc.c
> +++ b/xen/common/page_alloc.c
> @@ -126,6 +126,7 @@
> #include <xen/irq.h>
> #include <xen/keyhandler.h>
> #include <xen/lib.h>
> +#include <xen/llc-coloring.h>
> #include <xen/mm.h>
> #include <xen/nodemask.h>
> #include <xen/numa.h>
> @@ -2626,6 +2627,8 @@ static void cf_check pagealloc_info(unsigned char key)
> }
>
> printk(" Dom heap: %lukB free\n", total << (PAGE_SHIFT-10));
> +
> + dump_llc_coloring_info();
> }
>
> static __init int cf_check pagealloc_keyhandler_init(void)
> diff --git a/xen/include/xen/llc-coloring.h b/xen/include/xen/llc-coloring.h
> new file mode 100644
> index 0000000000..6ccf4613c2
> --- /dev/null
> +++ b/xen/include/xen/llc-coloring.h
> @@ -0,0 +1,37 @@
> +/* SPDX-License-Identifier: GPL-2.0-only */
> +/*
> + * Last Level Cache (LLC) coloring common header
> + *
> + * Copyright (C) 2024, Advanced Micro Devices, Inc.
> + * Copyright (C) 2024, Minerva Systems SRL
> + */
> +#ifndef __COLORING_H__
> +#define __COLORING_H__
> +
> +#include <xen/sched.h>
You could just forward declare struct domain unless you know you need sched.h for other things too in other patches.
> +#include <public/domctl.h>
Why is this here? It's not needed yet at this point.
> +
> +#ifdef CONFIG_LLC_COLORING
> +void llc_coloring_init(void);
> +void dump_llc_coloring_info(void);
> +void domain_dump_llc_colors(const struct domain *d);
> +#else
> +static inline void llc_coloring_init(void) {}
> +static inline void dump_llc_coloring_info(void) {}
> +static inline void domain_dump_llc_colors(const struct domain *d) {}
> +#endif
> +
> +unsigned int get_llc_way_size(void);
> +void arch_llc_coloring_init(void);
> +
> +#endif /* __COLORING_H__ */
> +
> +/*
> + * Local variables:
> + * mode: C
> + * c-file-style: "BSD"
> + * c-basic-offset: 4
> + * tab-width: 4
> + * indent-tabs-mode: nil
> + * End:
> + */
> diff --git a/xen/include/xen/sched.h b/xen/include/xen/sched.h
> index 90666576c2..c0e49cd1e7 100644
> --- a/xen/include/xen/sched.h
> +++ b/xen/include/xen/sched.h
> @@ -637,6 +637,11 @@ struct domain
>
> /* Holding CDF_* constant. Internal flags for domain creation. */
> unsigned int cdf;
> +
> +#ifdef CONFIG_LLC_COLORING
> + unsigned int num_llc_colors;
> + const unsigned int *llc_colors;
> +#endif
> };
>
> static inline struct page_list_head *page_to_list(
> --
> 2.43.0
>
Otherwise, with the remarks addressed:
Acked-by: Michal Orzel <michal.orzel@amd.com>
~Michal
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 01/12] xen/common: add cache coloring common code
2024-11-27 10:48 ` Michal Orzel
@ 2024-11-27 13:24 ` Carlo Nonato
2024-11-27 13:57 ` Michal Orzel
0 siblings, 1 reply; 36+ messages in thread
From: Carlo Nonato @ 2024-11-27 13:24 UTC (permalink / raw)
To: Michal Orzel
Cc: xen-devel, andrea.bastoni, marco.solieri, Andrew Cooper,
Jan Beulich, Julien Grall, Stefano Stabellini
Hi Michal,
On Wed, Nov 27, 2024 at 11:48 AM Michal Orzel <michal.orzel@amd.com> wrote:
> On 19/11/2024 15:13, Carlo Nonato wrote:
> >
> >
> > Last Level Cache (LLC) coloring allows to partition the cache in smaller
> > chunks called cache colors.
> >
> > Since not all architectures can actually implement it, add a HAS_LLC_COLORING
> > Kconfig option.
> > LLC_COLORS_ORDER Kconfig option has a range maximum of 10 (2^10 = 1024)
> > because that's the number of colors that fit in a 4 KiB page when integers
> > are 4 bytes long.
> >
> > LLC colors are a property of the domain, so struct domain has to be extended.
> >
> > Based on original work from: Luca Miccio <lucmiccio@gmail.com>
> >
> > Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> > Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
> > ---
> > v10:
> > - fixed commit message to use LLC_COLORS_ORDER
> > - added documentation to index.rst
> > - moved check on CONFIG_NUMA in arch/arm/Kconfig (next patch)
> > - fixed copyright line
> > - fixed array type for colors parameter in print_colors()
> > - added check on (way_size & ~PAGE_MASK)
> > v9:
> > - dropped _MAX_ from CONFIG_MAX_LLC_COLORS_ORDER
> > v8:
> > - minor documentation fixes
> > - "llc-coloring=on" is inferred from "llc-nr-ways" and "llc-size" usage
> > - turned CONFIG_NR_LLC_COLORS to CONFIG_MAX_LLC_COLORS_ORDER, base-2 exponent
> > - moved Kconfig options to common/Kconfig
> > - don't crash if computed max_nr_colors is too large
> > v7:
> > - SUPPORT.md changes added to this patch
> > - extended documentation to better address applicability of cache coloring
> > - "llc-nr-ways" and "llc-size" params introduced in favor of "llc-way-size"
> > - moved dump_llc_coloring_info() call in 'm' keyhandler (pagealloc_info())
> > v6:
> > - moved almost all code in common
> > - moved documentation in this patch
> > - reintroduced range for CONFIG_NR_LLC_COLORS
> > - reintroduced some stub functions to reduce the number of checks on
> > llc_coloring_enabled
> > - moved domain_llc_coloring_free() in same patch where allocation happens
> > - turned "d->llc_colors" to pointer-to-const
> > - llc_coloring_init() now returns void and panics if errors are found
> > v5:
> > - used - instead of _ for filenames
> > - removed domain_create_llc_colored()
> > - removed stub functions
> > - coloring domain fields are now #ifdef protected
> > v4:
> > - Kconfig options moved to xen/arch
> > - removed range for CONFIG_NR_LLC_COLORS
> > - added "llc_coloring_enabled" global to later implement the boot-time
> > switch
> > - added domain_create_llc_colored() to be able to pass colors
> > - added is_domain_llc_colored() macro
> > ---
> > SUPPORT.md | 7 ++
> > docs/index.rst | 1 +
> > docs/misc/cache-coloring.rst | 116 ++++++++++++++++++++++++++++++
> > docs/misc/xen-command-line.pandoc | 37 ++++++++++
> > xen/common/Kconfig | 21 ++++++
> > xen/common/Makefile | 1 +
> > xen/common/keyhandler.c | 3 +
> > xen/common/llc-coloring.c | 115 +++++++++++++++++++++++++++++
> > xen/common/page_alloc.c | 3 +
> > xen/include/xen/llc-coloring.h | 37 ++++++++++
> > xen/include/xen/sched.h | 5 ++
> > 11 files changed, 346 insertions(+)
> > create mode 100644 docs/misc/cache-coloring.rst
> > create mode 100644 xen/common/llc-coloring.c
> > create mode 100644 xen/include/xen/llc-coloring.h
> >
> > diff --git a/SUPPORT.md b/SUPPORT.md
> > index 82239d0294..998faf5635 100644
> > --- a/SUPPORT.md
> > +++ b/SUPPORT.md
> > @@ -401,6 +401,13 @@ by maintaining multiple physical to machine (p2m) memory mappings.
> > Status, x86 HVM: Tech Preview
> > Status, ARM: Tech Preview
> >
> > +### Cache coloring
> > +
> > +Allows to reserve Last Level Cache (LLC) partitions for Dom0, DomUs and Xen
> > +itself.
> > +
> > + Status, Arm64: Experimental
> > +
> > ## Resource Management
> >
> > ### CPU Pools
> > diff --git a/docs/index.rst b/docs/index.rst
> > index ab051a0f38..fad4462bbb 100644
> > --- a/docs/index.rst
> > +++ b/docs/index.rst
> > @@ -65,6 +65,7 @@ Documents in need of some rearranging.
> > designs/launch/hyperlaunch-devicetree
> > misc/xen-makefiles/makefiles
> > misra/index
> > + misc/cache-coloring
> >
> >
> > Miscellanea
> > diff --git a/docs/misc/cache-coloring.rst b/docs/misc/cache-coloring.rst
> > new file mode 100644
> > index 0000000000..0fe3830c40
> > --- /dev/null
> > +++ b/docs/misc/cache-coloring.rst
> > @@ -0,0 +1,116 @@
> You're missing SPDX tag at the top of the RST file:
> .. SPDX-License-Identifier: CC-BY-4.0
>
> > +Xen cache coloring user guide
> > +=============================
> > +
> > +The cache coloring support in Xen allows to reserve Last Level Cache (LLC)
> > +partitions for Dom0, DomUs and Xen itself. Currently only ARM64 is supported.
> > +Cache coloring realizes per-set cache partitioning in software and is applicable
> > +to shared LLCs as implemented in Cortex-A53, Cortex-A72 and similar CPUs.
> > +
> > +To compile LLC coloring support set ``CONFIG_LLC_COLORING=y``.
> > +
> > +If needed, change the maximum number of colors with
> > +``CONFIG_LLC_COLORS_ORDER=<n>``.
> > +
> > +Runtime configuration is done via `Command line parameters`_.
> > +
> > +Background
> > +**********
> > +
> > +Cache hierarchy of a modern multi-core CPU typically has first levels dedicated
> > +to each core (hence using multiple cache units), while the last level is shared
> > +among all of them. Such configuration implies that memory operations on one
> > +core (e.g. running a DomU) are able to generate interference on another core
> > +(e.g. hosting another DomU). Cache coloring realizes per-set cache-partitioning
> > +in software and mitigates this, guaranteeing more predictable performances for
> > +memory accesses.
> > +Software-based cache coloring is particularly useful in those situations where
> > +no hardware mechanisms (e.g., DSU-based way partitioning) are available to
> > +partition caches. This is the case for e.g., Cortex-A53, A57 and A72 CPUs that
> > +feature a L2 LLC cache shared among all cores.
> > +
> > +The key concept underlying cache coloring is a fragmentation of the memory
> > +space into a set of sub-spaces called colors that are mapped to disjoint cache
> > +partitions. Technically, the whole memory space is first divided into a number
> > +of subsequent regions. Then each region is in turn divided into a number of
> > +subsequent sub-colors. The generic i-th color is then obtained by all the
> > +i-th sub-colors in each region.
> > +
> > +::
> > +
> > + Region j Region j+1
> > + ..................... ............
> > + . . .
> > + . .
> > + _ _ _______________ _ _____________________ _ _
> > + | | | | | | |
> > + | c_0 | c_1 | | c_n | c_0 | c_1 |
> > + _ _ _|_____|_____|_ _ _|_____|_____|_____|_ _ _
> > + : :
> > + : :... ... .
> > + : color 0
> > + :........................... ... .
> > + :
> > + . . ..................................:
> > +
> > +How colors are actually defined depends on the function that maps memory to
> > +cache lines. In case of physically-indexed, physically-tagged caches with linear
> > +mapping, the set index is found by extracting some contiguous bits from the
> > +physical address. This allows colors to be defined as shown in figure: they
> > +appear in memory as subsequent blocks of equal size and repeats themselves after
> > +``n`` different colors, where ``n`` is the total number of colors.
> > +
> > +If some kind of bit shuffling appears in the mapping function, then colors
> > +assume a different layout in memory. Those kind of caches aren't supported by
> > +the current implementation.
> > +
> > +**Note**: Finding the exact cache mapping function can be a really difficult
> > +task since it's not always documented in the CPU manual. As said Cortex-A53, A57
> > +and A72 are known to work with the current implementation.
> > +
> > +How to compute the number of colors
> > +###################################
> > +
> > +Given the linear mapping from physical memory to cache lines for granted, the
> > +number of available colors for a specific platform is computed using three
> > +parameters:
> > +
> > +- the size of the LLC.
> > +- the number of the LLC ways.
> > +- the page size used by Xen.
> > +
> > +The first two parameters can be found in the processor manual, while the third
> > +one is the minimum mapping granularity. Dividing the cache size by the number of
> > +its ways we obtain the size of a way. Dividing this number by the page size,
> > +the number of total cache colors is found. So for example an Arm Cortex-A53
> > +with a 16-ways associative 1 MiB LLC can isolate up to 16 colors when pages are
> > +4 KiB in size.
> > +
> > +Effective colors assignment
> > +###########################
> > +
> > +When assigning colors, if one wants to avoid cache interference between two
> > +domains, different colors needs to be used for their memory.
> > +
> > +Command line parameters
> > +***********************
> > +
> > +Specific documentation is available at `docs/misc/xen-command-line.pandoc`.
> > +
> > ++----------------------+-------------------------------+
> > +| **Parameter** | **Description** |
> > ++----------------------+-------------------------------+
> > +| ``llc-coloring`` | Enable coloring at runtime |
> > ++----------------------+-------------------------------+
> > +| ``llc-size`` | Set the LLC size |
> > ++----------------------+-------------------------------+
> > +| ``llc-nr-ways`` | Set the LLC number of ways |
> > ++----------------------+-------------------------------+
> > +
> > +Auto-probing of LLC specs
> > +#########################
> > +
> > +LLC size and number of ways are probed automatically by default.
> > +
> > +LLC specs can be manually set via the above command line parameters. This
> > +bypasses any auto-probing and it's used to overcome failing situations, such as
> > +flawed probing logic, or for debugging/testing purposes.
> > diff --git a/docs/misc/xen-command-line.pandoc b/docs/misc/xen-command-line.pandoc
> > index 293dbc1a95..abd8dae96f 100644
> > --- a/docs/misc/xen-command-line.pandoc
> > +++ b/docs/misc/xen-command-line.pandoc
> > @@ -1708,6 +1708,43 @@ This option is intended for debugging purposes only. Enable MSR_DEBUGCTL.LBR
> > in hypervisor context to be able to dump the Last Interrupt/Exception To/From
> > record with other registers.
> >
> > +### llc-coloring (arm64)
> > +> `= <boolean>`
> > +
> > +> Default: `false`
> > +
> > +Flag to enable or disable LLC coloring support at runtime. This option is
> > +available only when `CONFIG_LLC_COLORING` is enabled. See the general
> > +cache coloring documentation for more info.
> > +
> > +### llc-nr-ways (arm64)
> > +> `= <integer>`
> > +
> > +> Default: `Obtained from hardware`
> > +
> > +Specify the number of ways of the Last Level Cache. This option is available
> > +only when `CONFIG_LLC_COLORING` is enabled. LLC size and number of ways are used
> > +to find the number of supported cache colors. By default the value is
> > +automatically computed by probing the hardware, but in case of specific needs,
> > +it can be manually set. Those include failing probing and debugging/testing
> > +purposes so that it's possible to emulate platforms with different number of
> > +supported colors. If set, also "llc-size" must be set, otherwise the default
> > +will be used. Note that using both options implies "llc-coloring=on".
> I can understand this decision, but ...
>
> [...]
>
> > + if ( llc_size && llc_nr_ways )
> > + {
> > + llc_coloring_enabled = true;
> > + way_size = llc_size / llc_nr_ways;
> > + }
> > + else if ( !llc_coloring_enabled )
> > + return;
> the above code does not seem to be right. When debugging or in general it is useful to have on the cmdline:
> llc-size=1M llc-nr-ways=16 llc-coloring=on
> and be able to disable it by just switching between on/off in llc-coloring=. However, with your solution,
> even if I specify llc-coloring=off, it will be enabled because I specified both llc-size and llc-nr-ways.
> I think llc-coloring= should have a precedence.
How do you differentiate from
llc-size=1M llc-nr-ways=16 llc-coloring=off
where llc coloring is disabled, and
llc-size=1M llc-nr-ways=16
where llc coloring is enabled? I mean, in both situations llc_coloring_enabled
is going to be set to false.
> > + else
> > + {
> > + way_size = get_llc_way_size();
> > + if ( !way_size )
> > + panic("LLC probing failed and 'llc-size' or 'llc-nr-ways' missing\n");
> > + }
> > +
> > + if ( way_size & ~PAGE_MASK )
> > + panic("LLC way size must be a multiple of PAGE_SIZE\n");
> > +
> > + /*
> > + * The maximum number of colors must be a power of 2 in order to correctly
> > + * map them to bits of an address.
> > + */
> > + max_nr_colors = way_size >> PAGE_SHIFT;
> > +
> > + if ( max_nr_colors & (max_nr_colors - 1) )
> > + panic("Number of LLC colors (%u) isn't a power of 2\n", max_nr_colors);
> > +
> > + if ( max_nr_colors > NR_LLC_COLORS )
> > + {
> > + printk(XENLOG_WARNING
> > + "Number of LLC colors (%u) too big. Using configured max %u\n",
> > + max_nr_colors, NR_LLC_COLORS);
> > + max_nr_colors = NR_LLC_COLORS;
> > + } else if ( max_nr_colors < 2 )
> CODING STYLE
>
> > + panic("Number of LLC colors %u < 2\n", max_nr_colors);
> > +
> > + arch_llc_coloring_init();
> > +}
> > +
> > +void dump_llc_coloring_info(void)
> > +{
> > + if ( !llc_coloring_enabled )
> > + return;
> > +
> > + printk("LLC coloring info:\n");
> > + printk(" Number of LLC colors supported: %u\n", max_nr_colors);
> > +}
> > +
> > +void domain_dump_llc_colors(const struct domain *d)
> > +{
> > + if ( !llc_coloring_enabled )
> > + return;
> > +
> > + printk("%u LLC colors: ", d->num_llc_colors);
> > + print_colors(d->llc_colors, d->num_llc_colors);
> > +}
> > +
> > +/*
> > + * Local variables:
> > + * mode: C
> > + * c-file-style: "BSD"
> > + * c-basic-offset: 4
> > + * tab-width: 4
> > + * indent-tabs-mode: nil
> > + * End:
> > + */
> > diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
> > index 33c8c917d9..7b911b5ed9 100644
> > --- a/xen/common/page_alloc.c
> > +++ b/xen/common/page_alloc.c
> > @@ -126,6 +126,7 @@
> > #include <xen/irq.h>
> > #include <xen/keyhandler.h>
> > #include <xen/lib.h>
> > +#include <xen/llc-coloring.h>
> > #include <xen/mm.h>
> > #include <xen/nodemask.h>
> > #include <xen/numa.h>
> > @@ -2626,6 +2627,8 @@ static void cf_check pagealloc_info(unsigned char key)
> > }
> >
> > printk(" Dom heap: %lukB free\n", total << (PAGE_SHIFT-10));
> > +
> > + dump_llc_coloring_info();
> > }
> >
> > static __init int cf_check pagealloc_keyhandler_init(void)
> > diff --git a/xen/include/xen/llc-coloring.h b/xen/include/xen/llc-coloring.h
> > new file mode 100644
> > index 0000000000..6ccf4613c2
> > --- /dev/null
> > +++ b/xen/include/xen/llc-coloring.h
> > @@ -0,0 +1,37 @@
> > +/* SPDX-License-Identifier: GPL-2.0-only */
> > +/*
> > + * Last Level Cache (LLC) coloring common header
> > + *
> > + * Copyright (C) 2024, Advanced Micro Devices, Inc.
> > + * Copyright (C) 2024, Minerva Systems SRL
> > + */
> > +#ifndef __COLORING_H__
> > +#define __COLORING_H__
> > +
> > +#include <xen/sched.h>
> You could just forward declare struct domain unless you know you need sched.h for other things too in other patches.
>
> > +#include <public/domctl.h>
> Why is this here? It's not needed yet at this point.
>
> > +
> > +#ifdef CONFIG_LLC_COLORING
> > +void llc_coloring_init(void);
> > +void dump_llc_coloring_info(void);
> > +void domain_dump_llc_colors(const struct domain *d);
> > +#else
> > +static inline void llc_coloring_init(void) {}
> > +static inline void dump_llc_coloring_info(void) {}
> > +static inline void domain_dump_llc_colors(const struct domain *d) {}
> > +#endif
> > +
> > +unsigned int get_llc_way_size(void);
> > +void arch_llc_coloring_init(void);
> > +
> > +#endif /* __COLORING_H__ */
> > +
> > +/*
> > + * Local variables:
> > + * mode: C
> > + * c-file-style: "BSD"
> > + * c-basic-offset: 4
> > + * tab-width: 4
> > + * indent-tabs-mode: nil
> > + * End:
> > + */
> > diff --git a/xen/include/xen/sched.h b/xen/include/xen/sched.h
> > index 90666576c2..c0e49cd1e7 100644
> > --- a/xen/include/xen/sched.h
> > +++ b/xen/include/xen/sched.h
> > @@ -637,6 +637,11 @@ struct domain
> >
> > /* Holding CDF_* constant. Internal flags for domain creation. */
> > unsigned int cdf;
> > +
> > +#ifdef CONFIG_LLC_COLORING
> > + unsigned int num_llc_colors;
> > + const unsigned int *llc_colors;
> > +#endif
> > };
> >
> > static inline struct page_list_head *page_to_list(
> > --
> > 2.43.0
> >
>
> Otherwise, with the remarks addressed:
> Acked-by: Michal Orzel <michal.orzel@amd.com>
>
> ~Michal
Thanks.
- Carlo
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 01/12] xen/common: add cache coloring common code
2024-11-27 13:24 ` Carlo Nonato
@ 2024-11-27 13:57 ` Michal Orzel
0 siblings, 0 replies; 36+ messages in thread
From: Michal Orzel @ 2024-11-27 13:57 UTC (permalink / raw)
To: Carlo Nonato
Cc: xen-devel, andrea.bastoni, marco.solieri, Andrew Cooper,
Jan Beulich, Julien Grall, Stefano Stabellini
On 27/11/2024 14:24, Carlo Nonato wrote:
>
>
> Hi Michal,
>
> On Wed, Nov 27, 2024 at 11:48 AM Michal Orzel <michal.orzel@amd.com> wrote:
>> On 19/11/2024 15:13, Carlo Nonato wrote:
>>>
>>>
>>> Last Level Cache (LLC) coloring allows to partition the cache in smaller
>>> chunks called cache colors.
>>>
>>> Since not all architectures can actually implement it, add a HAS_LLC_COLORING
>>> Kconfig option.
>>> LLC_COLORS_ORDER Kconfig option has a range maximum of 10 (2^10 = 1024)
>>> because that's the number of colors that fit in a 4 KiB page when integers
>>> are 4 bytes long.
>>>
>>> LLC colors are a property of the domain, so struct domain has to be extended.
>>>
>>> Based on original work from: Luca Miccio <lucmiccio@gmail.com>
>>>
>>> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
>>> Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
[...]
>>> +### llc-nr-ways (arm64)
>>> +> `= <integer>`
>>> +
>>> +> Default: `Obtained from hardware`
>>> +
>>> +Specify the number of ways of the Last Level Cache. This option is available
>>> +only when `CONFIG_LLC_COLORING` is enabled. LLC size and number of ways are used
>>> +to find the number of supported cache colors. By default the value is
>>> +automatically computed by probing the hardware, but in case of specific needs,
>>> +it can be manually set. Those include failing probing and debugging/testing
>>> +purposes so that it's possible to emulate platforms with different number of
>>> +supported colors. If set, also "llc-size" must be set, otherwise the default
>>> +will be used. Note that using both options implies "llc-coloring=on".
>> I can understand this decision, but ...
>>
>> [...]
>>
>>> + if ( llc_size && llc_nr_ways )
>>> + {
>>> + llc_coloring_enabled = true;
>>> + way_size = llc_size / llc_nr_ways;
>>> + }
>>> + else if ( !llc_coloring_enabled )
>>> + return;
>> the above code does not seem to be right. When debugging or in general it is useful to have on the cmdline:
>> llc-size=1M llc-nr-ways=16 llc-coloring=on
>> and be able to disable it by just switching between on/off in llc-coloring=. However, with your solution,
>> even if I specify llc-coloring=off, it will be enabled because I specified both llc-size and llc-nr-ways.
>> I think llc-coloring= should have a precedence.
>
> How do you differentiate from
> llc-size=1M llc-nr-ways=16 llc-coloring=off
> where llc coloring is disabled, and
> llc-size=1M llc-nr-ways=16
> where llc coloring is enabled? I mean, in both situations llc_coloring_enabled
> is going to be set to false.
I was thinking about the following:
diff --git a/xen/common/llc-coloring.c b/xen/common/llc-coloring.c
index 29b75e0e0d6a..b6684a6cf736 100644
--- a/xen/common/llc-coloring.c
+++ b/xen/common/llc-coloring.c
@@ -11,7 +11,12 @@
#define NR_LLC_COLORS (1U << CONFIG_LLC_COLORS_ORDER)
-static bool __ro_after_init llc_coloring_enabled;
+/*
+ * -1: not specified (disabled unless llc-size and llc-nr-ways present)
+ * 0: explicitly disabled through cmdline
+ * 1: explicitly enabled through cmdline
+ */
+static int8_t __ro_after_init llc_coloring_enabled = -1;
boolean_param("llc-coloring", llc_coloring_enabled);
static unsigned int __initdata llc_size;
@@ -48,7 +53,7 @@ void __init llc_coloring_init(void)
{
unsigned int way_size;
- if ( llc_size && llc_nr_ways )
+ if ( (llc_coloring_enabled < 0) && (llc_size && llc_nr_ways) )
{
llc_coloring_enabled = true;
way_size = llc_size / llc_nr_ways;
~Michal
^ permalink raw reply related [flat|nested] 36+ messages in thread
* Re: [PATCH v10 02/12] xen/arm: add initial support for LLC coloring on arm64
2024-11-19 14:13 ` [PATCH v10 02/12] xen/arm: add initial support for LLC coloring on arm64 Carlo Nonato
@ 2024-11-27 14:32 ` Michal Orzel
2024-11-28 10:42 ` Jan Beulich
1 sibling, 0 replies; 36+ messages in thread
From: Michal Orzel @ 2024-11-27 14:32 UTC (permalink / raw)
To: Carlo Nonato, xen-devel
Cc: andrea.bastoni, marco.solieri, Andrew Cooper, Jan Beulich,
Julien Grall, Stefano Stabellini, Bertrand Marquis,
Volodymyr Babchuk
On 19/11/2024 15:13, Carlo Nonato wrote:
>
>
> LLC coloring needs to know the last level cache layout in order to make the
> best use of it. This can be probed by inspecting the CLIDR_EL1 register,
> so the Last Level is defined as the last level visible by this register.
> Note that this excludes system caches in some platforms.
>
> Static memory allocation and cache coloring are incompatible because static
> memory can't be guaranteed to use only colors assigned to the domain.
> Panic during DomUs creation when both are enabled.
>
> Based on original work from: Luca Miccio <lucmiccio@gmail.com>
>
> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
> ---
> v10:
> - moved CONFIG_NUMA check in arch/arm/Kconfig
> v9:
> - no changes
> v8:
> - no changes
> v7:
> - only minor changes
> v6:
> - get_llc_way_size() now checks for at least separate I/D caches
> v5:
> - used - instead of _ for filenames
> - moved static-mem check in this patch
> - moved dom0 colors parsing in next patch
> - moved color allocation and configuration in next patch
> - moved check_colors() in next patch
> - colors are now printed in short form
> v4:
> - added "llc-coloring" cmdline option for the boot-time switch
> - dom0 colors are now checked during domain init as for any other domain
> - fixed processor.h masks bit width
> - check for overflow in parse_color_config()
> - check_colors() now checks also that colors are sorted and unique
> ---
> docs/misc/cache-coloring.rst | 14 +++++
> xen/arch/arm/Kconfig | 1 +
> xen/arch/arm/Makefile | 1 +
> xen/arch/arm/dom0less-build.c | 6 +++
> xen/arch/arm/include/asm/processor.h | 16 ++++++
> xen/arch/arm/llc-coloring.c | 78 ++++++++++++++++++++++++++++
> xen/arch/arm/setup.c | 3 ++
> xen/common/llc-coloring.c | 2 +-
> xen/include/xen/llc-coloring.h | 4 ++
> 9 files changed, 124 insertions(+), 1 deletion(-)
> create mode 100644 xen/arch/arm/llc-coloring.c
>
> diff --git a/docs/misc/cache-coloring.rst b/docs/misc/cache-coloring.rst
> index 0fe3830c40..b608284e9b 100644
> --- a/docs/misc/cache-coloring.rst
> +++ b/docs/misc/cache-coloring.rst
> @@ -111,6 +111,20 @@ Auto-probing of LLC specs
>
> LLC size and number of ways are probed automatically by default.
>
> +In the Arm implementation, this is done by inspecting the CLIDR_EL1 register.
> +This means that other system caches that aren't visible there are ignored.
> +
> LLC specs can be manually set via the above command line parameters. This
> bypasses any auto-probing and it's used to overcome failing situations, such as
> flawed probing logic, or for debugging/testing purposes.
> +
> +Known issues and limitations
> +****************************
> +
> +"xen,static-mem" isn't supported when coloring is enabled
> +#########################################################
> +
> +In the domain configuration, "xen,static-mem" allows memory to be statically
> +allocated to the domain. This isn't possible when LLC coloring is enabled,
> +because that memory can't be guaranteed to use only colors assigned to the
> +domain.
> diff --git a/xen/arch/arm/Kconfig b/xen/arch/arm/Kconfig
> index 15b2e4a227..7f0712e674 100644
> --- a/xen/arch/arm/Kconfig
> +++ b/xen/arch/arm/Kconfig
> @@ -8,6 +8,7 @@ config ARM_64
> depends on !ARM_32
> select 64BIT
> select HAS_FAST_MULTIPLY
> + select HAS_LLC_COLORING if !NUMA
>
> config ARM
> def_bool y
> diff --git a/xen/arch/arm/Makefile b/xen/arch/arm/Makefile
> index e4ad1ce851..ccbfc61f88 100644
> --- a/xen/arch/arm/Makefile
> +++ b/xen/arch/arm/Makefile
> @@ -35,6 +35,7 @@ obj-$(CONFIG_IOREQ_SERVER) += ioreq.o
> obj-y += irq.o
> obj-y += kernel.init.o
> obj-$(CONFIG_LIVEPATCH) += livepatch.o
> +obj-$(CONFIG_LLC_COLORING) += llc-coloring.o
> obj-$(CONFIG_MEM_ACCESS) += mem_access.o
> obj-y += mm.o
> obj-y += monitor.o
> diff --git a/xen/arch/arm/dom0less-build.c b/xen/arch/arm/dom0less-build.c
> index f328a044e9..d93a85434e 100644
> --- a/xen/arch/arm/dom0less-build.c
> +++ b/xen/arch/arm/dom0less-build.c
> @@ -5,6 +5,7 @@
> #include <xen/grant_table.h>
> #include <xen/iocap.h>
> #include <xen/libfdt/libfdt.h>
> +#include <xen/llc-coloring.h>
> #include <xen/sched.h>
> #include <xen/serial.h>
> #include <xen/sizes.h>
> @@ -890,7 +891,12 @@ void __init create_domUs(void)
> panic("No more domain IDs available\n");
>
> if ( dt_find_property(node, "xen,static-mem", NULL) )
> + {
> + if ( llc_coloring_enabled )
> + panic("LLC coloring and static memory are incompatible\n");
> +
> flags |= CDF_staticmem;
> + }
>
> if ( dt_property_read_bool(node, "direct-map") )
> {
> diff --git a/xen/arch/arm/include/asm/processor.h b/xen/arch/arm/include/asm/processor.h
> index 8e02410465..ef33ea198c 100644
> --- a/xen/arch/arm/include/asm/processor.h
> +++ b/xen/arch/arm/include/asm/processor.h
> @@ -18,6 +18,22 @@
> #define CTR_IDC_SHIFT 28
> #define CTR_DIC_SHIFT 29
>
> +/* CCSIDR Current Cache Size ID Register */
> +#define CCSIDR_LINESIZE_MASK _AC(0x7, UL)
> +#define CCSIDR_NUMSETS_SHIFT 13
> +#define CCSIDR_NUMSETS_MASK _AC(0x3fff, UL)
> +#define CCSIDR_NUMSETS_SHIFT_FEAT_CCIDX 32
> +#define CCSIDR_NUMSETS_MASK_FEAT_CCIDX _AC(0xffffff, UL)
> +
> +/* CSSELR Cache Size Selection Register */
> +#define CSSELR_LEVEL_MASK _AC(0x7, UL)
Seems unused. If so, please remove.
> +#define CSSELR_LEVEL_SHIFT 1
> +
> +/* CLIDR Cache Level ID Register */
> +#define CLIDR_CTYPEn_SHIFT(n) (3 * ((n) - 1))
> +#define CLIDR_CTYPEn_MASK _AC(0x7, UL)
> +#define CLIDR_CTYPEn_LEVELS 7
> +
> #define ICACHE_POLICY_VPIPT 0
> #define ICACHE_POLICY_AIVIVT 1
> #define ICACHE_POLICY_VIPT 2
> diff --git a/xen/arch/arm/llc-coloring.c b/xen/arch/arm/llc-coloring.c
> new file mode 100644
> index 0000000000..a7e0907816
> --- /dev/null
> +++ b/xen/arch/arm/llc-coloring.c
> @@ -0,0 +1,78 @@
> +/* SPDX-License-Identifier: GPL-2.0-only */
> +/*
> + * Last Level Cache (LLC) coloring support for ARM
> + *
> + * Copyright (C) 2024, Advanced Micro Devices, Inc.
> + * Copyright (C) 2024, Minerva Systems SRL
> + */
> +#include <xen/llc-coloring.h>
> +#include <xen/types.h>
> +
> +#include <asm/processor.h>
> +#include <asm/sysregs.h>
> +
> +/* Return the LLC way size by probing the hardware */
> +unsigned int __init get_llc_way_size(void)
> +{
> + register_t ccsidr_el1;
> + register_t clidr_el1 = READ_SYSREG(CLIDR_EL1);
> + register_t csselr_el1 = READ_SYSREG(CSSELR_EL1);
> + register_t id_aa64mmfr2_el1 = READ_SYSREG(ID_AA64MMFR2_EL1);
> + uint32_t ccsidr_numsets_shift = CCSIDR_NUMSETS_SHIFT;
> + uint32_t ccsidr_numsets_mask = CCSIDR_NUMSETS_MASK;
> + unsigned int n, line_size, num_sets;
> +
> + for ( n = CLIDR_CTYPEn_LEVELS; n != 0; n-- )
> + {
> + uint8_t ctype_n = (clidr_el1 >> CLIDR_CTYPEn_SHIFT(n)) &
> + CLIDR_CTYPEn_MASK;
NIT: align CLIDR_xxx to opening (
Other than that:
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
~Michal
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 03/12] xen/arm: permit non direct-mapped Dom0 construction
2024-11-19 14:13 ` [PATCH v10 03/12] xen/arm: permit non direct-mapped Dom0 construction Carlo Nonato
@ 2024-11-28 10:34 ` Michal Orzel
2024-11-29 9:06 ` Carlo Nonato
0 siblings, 1 reply; 36+ messages in thread
From: Michal Orzel @ 2024-11-28 10:34 UTC (permalink / raw)
To: Carlo Nonato, xen-devel
Cc: andrea.bastoni, marco.solieri, Stefano Stabellini, Julien Grall,
Bertrand Marquis, Volodymyr Babchuk
On 19/11/2024 15:13, Carlo Nonato wrote:
>
>
> Cache coloring requires Dom0 not to be direct-mapped because of its non
> contiguous mapping nature, so allocate_memory() is needed in this case.
> 8d2c3ab18cc1 ("arm/dom0less: put dom0less feature code in a separate module")
> moved allocate_memory() in dom0less_build.c. In order to use it
> in Dom0 construction bring it back to domain_build.c and declare it in
> domain_build.h.
>
> Take the opportunity to adapt the implementation of allocate_memory() so
> that it uses the host layout when called on the hwdom, via
> find_unallocated_memory().
>
> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> ---
> v10:
> - fixed a compilation bug that happened when dom0less support was disabled
> v9:
> - no changes
> v8:
> - patch adapted to new changes to allocate_memory()
> v7:
> - allocate_memory() now uses the host layout when called on the hwdom
> v6:
> - new patch
> ---
> xen/arch/arm/dom0less-build.c | 44 ------------
> xen/arch/arm/domain_build.c | 96 +++++++++++++++++++++++--
> xen/arch/arm/include/asm/domain_build.h | 1 +
> 3 files changed, 93 insertions(+), 48 deletions(-)
>
> diff --git a/xen/arch/arm/dom0less-build.c b/xen/arch/arm/dom0less-build.c
> index d93a85434e..67b1503647 100644
> --- a/xen/arch/arm/dom0less-build.c
> +++ b/xen/arch/arm/dom0less-build.c
> @@ -49,50 +49,6 @@ bool __init is_dom0less_mode(void)
> return ( !dom0found && domUfound );
> }
>
> -static void __init allocate_memory(struct domain *d, struct kernel_info *kinfo)
> -{
> - struct membanks *mem = kernel_info_get_mem(kinfo);
> - unsigned int i;
> - paddr_t bank_size;
> -
> - printk(XENLOG_INFO "Allocating mappings totalling %ldMB for %pd:\n",
> - /* Don't want format this as PRIpaddr (16 digit hex) */
> - (unsigned long)(kinfo->unassigned_mem >> 20), d);
> -
> - mem->nr_banks = 0;
> - bank_size = MIN(GUEST_RAM0_SIZE, kinfo->unassigned_mem);
> - if ( !allocate_bank_memory(kinfo, gaddr_to_gfn(GUEST_RAM0_BASE),
> - bank_size) )
> - goto fail;
> -
> - bank_size = MIN(GUEST_RAM1_SIZE, kinfo->unassigned_mem);
> - if ( !allocate_bank_memory(kinfo, gaddr_to_gfn(GUEST_RAM1_BASE),
> - bank_size) )
> - goto fail;
> -
> - if ( kinfo->unassigned_mem )
> - goto fail;
> -
> - for( i = 0; i < mem->nr_banks; i++ )
> - {
> - printk(XENLOG_INFO "%pd BANK[%d] %#"PRIpaddr"-%#"PRIpaddr" (%ldMB)\n",
> - d,
> - i,
> - mem->bank[i].start,
> - mem->bank[i].start + mem->bank[i].size,
> - /* Don't want format this as PRIpaddr (16 digit hex) */
> - (unsigned long)(mem->bank[i].size >> 20));
> - }
> -
> - return;
> -
> -fail:
> - panic("Failed to allocate requested domain memory."
> - /* Don't want format this as PRIpaddr (16 digit hex) */
> - " %ldKB unallocated. Fix the VMs configurations.\n",
> - (unsigned long)kinfo->unassigned_mem >> 10);
> -}
> -
> #ifdef CONFIG_VGICV2
> static int __init make_gicv2_domU_node(struct kernel_info *kinfo)
> {
> diff --git a/xen/arch/arm/domain_build.c b/xen/arch/arm/domain_build.c
> index 2c30792de8..a95376dcdc 100644
> --- a/xen/arch/arm/domain_build.c
> +++ b/xen/arch/arm/domain_build.c
> @@ -416,7 +416,6 @@ static void __init allocate_memory_11(struct domain *d,
> }
> }
>
> -#ifdef CONFIG_DOM0LESS_BOOT
> bool __init allocate_domheap_memory(struct domain *d, paddr_t tot_size,
> alloc_domheap_mem_cb cb, void *extra)
> {
> @@ -508,7 +507,6 @@ bool __init allocate_bank_memory(struct kernel_info *kinfo, gfn_t sgfn,
>
> return true;
> }
> -#endif
>
> /*
> * When PCI passthrough is available we want to keep the
> @@ -1003,6 +1001,93 @@ out:
> return res;
> }
>
> +void __init allocate_memory(struct domain *d, struct kernel_info *kinfo)
> +{
> + struct membanks *mem = kernel_info_get_mem(kinfo);
> + unsigned int i, nr_banks = 2;
Instead of hardcoding, shouldn't it be GUEST_RAM_BANKS?
Also, the second bank won't work with CONFIG_ARM_PA_BITS_32 which limits PA to 32bit.
> + paddr_t bank_start, bank_size;
> + struct membanks *hwdom_ext_regions = NULL;
AFAICT you search for free memory. Naming it as extended regions is not a good choice.
Instead, hwdom_free_mem?
> +
> + printk(XENLOG_INFO "Allocating mappings totalling %ldMB for %pd:\n",
> + /* Don't want format this as PRIpaddr (16 digit hex) */
> + (unsigned long)(kinfo->unassigned_mem >> 20), d);
> +
> + mem->nr_banks = 0;
> + /*
> + * Use host memory layout for hwdom. Only case for this is when LLC coloring
> + * is enabled.
> + */
> + if ( is_hardware_domain(d) )
> + {
> + ASSERT(llc_coloring_enabled);
> +
> + hwdom_ext_regions = xzalloc_flex_struct(struct membanks, bank,
> + NR_MEM_BANKS);
> + if ( !hwdom_ext_regions )
> + goto fail;
empty line here please
> + hwdom_ext_regions->max_banks = NR_MEM_BANKS;
> +
> + if ( find_unallocated_memory(kinfo, hwdom_ext_regions) )
If you reuse find_unallocated_memory for a purpose other than extended regions, I think
the description of this function should change as well as comments inside.
Today, the function gets all RAM and exclude dom0 RAM (in your case is 0 at this point, reserved memory,
static shmem and gnttab (in your case is 0 at this point). I think we cannot get away without
making this function more generic. Maybe it should take as a parameter struct membanks * array?
Also, the callback add_ext_regions() may not be suited for all purposes (i.e. it takes only banks
> 64MB into account). I know that there will be more use cases for a function that will return the
free memory for domains. As an example, today, for direct mapped domains we hardcode the gnttab region
(only dom0 is special cased). This should not be like that. We would need to find a free memory region
to expose as gnttab.
> + goto fail;
> +
> + nr_banks = hwdom_ext_regions->nr_banks;
> + }
> +
> + for ( i = 0; kinfo->unassigned_mem > 0 && nr_banks > 0; i++, nr_banks-- )
> + {
> + if ( is_hardware_domain(d) )
> + {
> + bank_start = hwdom_ext_regions->bank[i].start;
> + bank_size = hwdom_ext_regions->bank[i].size;
> +
> + if ( bank_size < min_t(paddr_t, kinfo->unassigned_mem, MB(128)) )
I would expect some explanation.
> + continue;
> + }
> + else
> + {
> + if ( i == 0 )
> + {
> + bank_start = GUEST_RAM0_BASE;
> + bank_size = GUEST_RAM0_SIZE;
> + }
> + else if ( i == 1 )
> + {
> + bank_start = GUEST_RAM1_BASE;
> + bank_size = GUEST_RAM1_SIZE;
> + }
> + else
> + goto fail;
This could be simplified:
const uint64_t bankbase[] = GUEST_RAM_BANK_BASES;
const uint64_t banksize[] = GUEST_RAM_BANK_SIZES;
if ( i >= GUEST_RAM_BANKS )
goto fail;
bank_start = bankbase[i];
bank_size = banksize[i];
This patch requires also opinion of other maintainers.
~Michal
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 01/12] xen/common: add cache coloring common code
2024-11-19 14:13 ` [PATCH v10 01/12] xen/common: add cache coloring common code Carlo Nonato
2024-11-27 10:48 ` Michal Orzel
@ 2024-11-28 10:39 ` Jan Beulich
1 sibling, 0 replies; 36+ messages in thread
From: Jan Beulich @ 2024-11-28 10:39 UTC (permalink / raw)
To: Carlo Nonato, xen-devel
Cc: andrea.bastoni, marco.solieri, Andrew Cooper, Julien Grall,
Stefano Stabellini
On 19.11.2024 15:13, Carlo Nonato wrote:
> --- /dev/null
> +++ b/xen/include/xen/llc-coloring.h
> @@ -0,0 +1,37 @@
> +/* SPDX-License-Identifier: GPL-2.0-only */
> +/*
> + * Last Level Cache (LLC) coloring common header
> + *
> + * Copyright (C) 2024, Advanced Micro Devices, Inc.
> + * Copyright (C) 2024, Minerva Systems SRL
> + */
> +#ifndef __COLORING_H__
> +#define __COLORING_H__
This certainly wants LLC in the identifier, and likely also XEN.
Jan
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 02/12] xen/arm: add initial support for LLC coloring on arm64
2024-11-19 14:13 ` [PATCH v10 02/12] xen/arm: add initial support for LLC coloring on arm64 Carlo Nonato
2024-11-27 14:32 ` Michal Orzel
@ 2024-11-28 10:42 ` Jan Beulich
1 sibling, 0 replies; 36+ messages in thread
From: Jan Beulich @ 2024-11-28 10:42 UTC (permalink / raw)
To: Carlo Nonato
Cc: andrea.bastoni, marco.solieri, Andrew Cooper, Julien Grall,
Stefano Stabellini, Bertrand Marquis, Michal Orzel,
Volodymyr Babchuk, xen-devel
On 19.11.2024 15:13, Carlo Nonato wrote:
> LLC coloring needs to know the last level cache layout in order to make the
> best use of it. This can be probed by inspecting the CLIDR_EL1 register,
> so the Last Level is defined as the last level visible by this register.
> Note that this excludes system caches in some platforms.
>
> Static memory allocation and cache coloring are incompatible because static
> memory can't be guaranteed to use only colors assigned to the domain.
> Panic during DomUs creation when both are enabled.
>
> Based on original work from: Luca Miccio <lucmiccio@gmail.com>
>
> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
Acked-by: Jan Beulich <jbeulich@suse.com>
(As this is inapplicable to Arm code, it's independent of Michal's comments.)
Jan
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 08/12] xen/page_alloc: introduce preserved page flags macro
2024-11-19 14:13 ` [PATCH v10 08/12] xen/page_alloc: introduce preserved page flags macro Carlo Nonato
@ 2024-11-28 11:04 ` Jan Beulich
2024-11-29 9:32 ` Carlo Nonato
0 siblings, 1 reply; 36+ messages in thread
From: Jan Beulich @ 2024-11-28 11:04 UTC (permalink / raw)
To: Carlo Nonato
Cc: andrea.bastoni, marco.solieri, Andrew Cooper, Julien Grall,
Stefano Stabellini, xen-devel
On 19.11.2024 15:13, Carlo Nonato wrote:
> PGC_static, PGC_extra and PGC_broken need to be preserved when assigning a
> page. Define a new macro that groups those flags and use it instead of or'ing
> every time.
>
> To make preserved flags even more meaningful, they are kept also when
> switching state in mark_page_free().
> Enforce the removal of PGC_extra before freeing domain pages as this is
> considered an error and can cause ASSERT violations.
>
> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> ---
> v10:
> - fixed commit message
> v9:
> - add PGC_broken to PGC_preserved
> - clear PGC_extra in alloc_domheap_pages() only if MEMF_no_refcount is set
> v8:
> - fixed PGC_extra ASSERT fail in alloc_domheap_pages() by removing PGC_extra
> before freeing
> v7:
> - PGC_preserved used also in mark_page_free()
> v6:
> - preserved_flags renamed to PGC_preserved
> - PGC_preserved is used only in assign_pages()
> v5:
> - new patch
> ---
> xen/common/page_alloc.c | 19 ++++++++++++++-----
> 1 file changed, 14 insertions(+), 5 deletions(-)
>
> diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
> index 7b911b5ed9..34cd473150 100644
> --- a/xen/common/page_alloc.c
> +++ b/xen/common/page_alloc.c
> @@ -160,6 +160,7 @@
> #endif
>
> #define PGC_no_buddy_merge PGC_static
> +#define PGC_preserved (PGC_extra | PGC_static | PGC_broken)
>
> #ifndef PGT_TYPE_INFO_INITIALIZER
> #define PGT_TYPE_INFO_INITIALIZER 0
> @@ -1427,12 +1428,11 @@ static bool mark_page_free(struct page_info *pg, mfn_t mfn)
> {
> case PGC_state_inuse:
> BUG_ON(pg->count_info & PGC_broken);
> - pg->count_info = PGC_state_free;
> + pg->count_info = PGC_state_free | (pg->count_info & PGC_preserved);
> break;
PGC_extra doesn't want preserving here. Since it's mark_page_free(), and
since PGC_extra is removed before freeing, the state change is apparently
fine. But an assertion may want adding, for documentation purposes if
nothing else.
Alternatively it may make sense to indeed exclude PGC_extra here. In fact
PGC_static doesn't need using here either, as unprepare_staticmem_pages()
will explicitly set it again anyway.
Hence I wonder whether the change here really is necessary (one will then
be needed in the next patch aiui, when PGC_colored is introduced). Which
would then eliminate the need for the final two hunks of the patch, I
think.
> case PGC_state_offlining:
> - pg->count_info = (pg->count_info & PGC_broken) |
> - PGC_state_offlined;
> + pg->count_info = (pg->count_info & PGC_preserved) | PGC_state_offlined;
> pg_offlined = true;
> break;
I'm similarly unconvinced that anything other than PGC_broken (and
subsequently perhaps PGC_colored) would need preserving here.
Jan
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 04/12] xen/arm: add Dom0 cache coloring support
2024-11-19 14:13 ` [PATCH v10 04/12] xen/arm: add Dom0 cache coloring support Carlo Nonato
@ 2024-11-28 11:27 ` Michal Orzel
0 siblings, 0 replies; 36+ messages in thread
From: Michal Orzel @ 2024-11-28 11:27 UTC (permalink / raw)
To: Carlo Nonato, xen-devel
Cc: andrea.bastoni, marco.solieri, Andrew Cooper, Jan Beulich,
Julien Grall, Stefano Stabellini, Bertrand Marquis,
Volodymyr Babchuk
On 19/11/2024 15:13, Carlo Nonato wrote:
>
>
> Add a command line parameter to allow the user to set the coloring
> configuration for Dom0.
> A common configuration syntax for cache colors is introduced and
> documented.
> Take the opportunity to also add:
> - default configuration notion.
> - function to check well-formed configurations.
>
> Direct mapping Dom0 isn't possible when coloring is enabled, so
> CDF_directmap flag is removed when creating it.
>
> Based on original work from: Luca Miccio <lucmiccio@gmail.com>
>
> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
> Reviewed-by: Jan Beulich <jbeulich@suse.com>
> ---
> v10:
> - fixed array type for colors parameter in check_colors()
> v9:
> - moved domain_llc_coloring_free() in next patch cause it's never used for dom0
> v8:
> - added bound check on dom0_num_colors
> - default colors array set just once
> v7:
> - parse_color_config() doesn't accept leading/trailing commas anymore
> - removed alloc_colors() helper
> v6:
> - moved domain_llc_coloring_free() in this patch
> - removed domain_alloc_colors() in favor of a more explicit allocation
> - parse_color_config() now accepts the size of the array to be filled
> - allocate_memory() moved in another patch
> v5:
> - Carlo Nonato as the new author
> - moved dom0 colors parsing (parse_colors()) in this patch
> - added dom0_set_llc_colors() to set dom0 colors after creation
> - moved color allocation and checking in this patch
> - error handling when allocating color arrays
> - FIXME: copy pasted allocate_memory() cause it got moved
> v4:
> - dom0 colors are dynamically allocated as for any other domain
> (colors are duplicated in dom0_colors and in the new array, but logic
> is simpler)
> ---
> docs/misc/cache-coloring.rst | 29 ++++++++
> docs/misc/xen-command-line.pandoc | 9 +++
> xen/arch/arm/domain_build.c | 10 ++-
> xen/common/llc-coloring.c | 119 +++++++++++++++++++++++++++++-
> xen/include/xen/llc-coloring.h | 1 +
> 5 files changed, 166 insertions(+), 2 deletions(-)
>
> diff --git a/docs/misc/cache-coloring.rst b/docs/misc/cache-coloring.rst
> index b608284e9b..c5fb33996c 100644
> --- a/docs/misc/cache-coloring.rst
> +++ b/docs/misc/cache-coloring.rst
> @@ -105,6 +105,35 @@ Specific documentation is available at `docs/misc/xen-command-line.pandoc`.
> +----------------------+-------------------------------+
> | ``llc-nr-ways`` | Set the LLC number of ways |
> +----------------------+-------------------------------+
> +| ``dom0-llc-colors`` | Dom0 color configuration |
> ++----------------------+-------------------------------+
> +
> +Colors selection format
> +***********************
> +
> +Regardless of the memory pool that has to be colored (Xen, Dom0/DomUs),
> +the color selection can be expressed using the same syntax. In particular a
> +comma-separated list of colors or ranges of colors is used.
> +Ranges are hyphen-separated intervals (such as `0-4`) and are inclusive on both
> +sides.
> +
> +Note that:
> +
> +- no spaces are allowed between values.
> +- no overlapping ranges or duplicated colors are allowed.
> +- values must be written in ascending order.
> +
> +Examples:
> +
> ++-------------------+-----------------------------+
> +| **Configuration** | **Actual selection** |
> ++-------------------+-----------------------------+
> +| 1-2,5-8 | [1, 2, 5, 6, 7, 8] |
> ++-------------------+-----------------------------+
> +| 4-8,10,11,12 | [4, 5, 6, 7, 8, 10, 11, 12] |
> ++-------------------+-----------------------------+
> +| 0 | [0] |
> ++-------------------+-----------------------------+
>
> Auto-probing of LLC specs
> #########################
> diff --git a/docs/misc/xen-command-line.pandoc b/docs/misc/xen-command-line.pandoc
> index abd8dae96f..bfdc8b0002 100644
> --- a/docs/misc/xen-command-line.pandoc
> +++ b/docs/misc/xen-command-line.pandoc
> @@ -963,6 +963,15 @@ Controls for the dom0 IOMMU setup.
>
> Specify a list of IO ports to be excluded from dom0 access.
>
> +### dom0-llc-colors (arm64)
> +> `= List of [ <integer> | <integer>-<integer> ]`
> +
> +> Default: `All available LLC colors`
> +
> +Specify dom0 LLC color configuration. This option is available only when
> +`CONFIG_LLC_COLORING` is enabled. If the parameter is not set, all available
> +colors are used.
> +
> ### dom0_max_vcpus
>
> Either:
> diff --git a/xen/arch/arm/domain_build.c b/xen/arch/arm/domain_build.c
> index a95376dcdc..01aa9016f5 100644
> --- a/xen/arch/arm/domain_build.c
> +++ b/xen/arch/arm/domain_build.c
> @@ -2,6 +2,7 @@
> #include <xen/init.h>
> #include <xen/compile.h>
> #include <xen/lib.h>
> +#include <xen/llc-coloring.h>
> #include <xen/mm.h>
> #include <xen/param.h>
> #include <xen/domain_page.h>
> @@ -2284,6 +2285,7 @@ void __init create_dom0(void)
> .max_maptrack_frames = -1,
> .grant_opts = XEN_DOMCTL_GRANT_version(opt_gnttab_max_version),
> };
> + unsigned int flags = CDF_privileged;
> int rc;
>
> /* The vGIC for DOM0 is exactly emulating the hardware GIC */
> @@ -2311,10 +2313,16 @@ void __init create_dom0(void)
> panic("SVE vector length error\n");
> }
>
> - dom0 = domain_create(0, &dom0_cfg, CDF_privileged | CDF_directmap);
> + if ( !llc_coloring_enabled )
> + flags |= CDF_directmap;
> +
> + dom0 = domain_create(0, &dom0_cfg, flags);
> if ( IS_ERR(dom0) )
> panic("Error creating domain 0 (rc = %ld)\n", PTR_ERR(dom0));
>
> + if ( llc_coloring_enabled && (rc = dom0_set_llc_colors(dom0)) )
> + panic("Error initializing LLC coloring for domain 0 (rc = %d)", rc);
Missing newline \n
> +
> if ( alloc_dom0_vcpu0(dom0) == NULL )
> panic("Error creating domain 0 vcpu0\n");
>
> diff --git a/xen/common/llc-coloring.c b/xen/common/llc-coloring.c
> index 45b001b105..740b5b9e4f 100644
> --- a/xen/common/llc-coloring.c
> +++ b/xen/common/llc-coloring.c
> @@ -20,6 +20,66 @@ static unsigned int __initdata llc_nr_ways;
> integer_param("llc-nr-ways", llc_nr_ways);
> /* Number of colors available in the LLC */
> static unsigned int __ro_after_init max_nr_colors;
> +/* Default coloring configuration */
> +static unsigned int __ro_after_init default_colors[NR_LLC_COLORS];
> +
> +static unsigned int __initdata dom0_colors[NR_LLC_COLORS];
> +static unsigned int __initdata dom0_num_colors;
> +
> +/*
> + * Parse the coloring configuration given in the buf string, following the
> + * syntax below.
> + *
> + * COLOR_CONFIGURATION ::= COLOR | RANGE,...,COLOR | RANGE
> + * RANGE ::= COLOR-COLOR
> + *
> + * Example: "0,2-6,15-16" represents the set of colors: 0,2,3,4,5,6,15,16.
> + */
> +static int __init parse_color_config(const char *buf, unsigned int colors[],
> + unsigned int max_num_colors,
> + unsigned int *num_colors)
> +{
> + const char *s = buf;
> +
> + *num_colors = 0;
> +
> + while ( *s != '\0' )
> + {
> + unsigned int color, start, end;
> +
> + start = simple_strtoul(s, &s, 0);
> +
> + if ( *s == '-' ) /* Range */
> + {
> + s++;
> + end = simple_strtoul(s, &s, 0);
> + }
> + else /* Single value */
> + end = start;
> +
> + if ( start > end || (end - start) > (UINT_MAX - *num_colors) ||
> + (*num_colors + (end - start)) >= max_num_colors )
> + return -EINVAL;
> +
> + /* Colors are range checked in check_colors() */
> + for ( color = start; color <= end; color++ )
> + colors[(*num_colors)++] = color;
> +
> + if ( *s == ',' )
> + s++;
> + else if ( *s != '\0' )
> + break;
> + }
> +
> + return *s ? -EINVAL : 0;
> +}
> +
> +static int __init parse_dom0_colors(const char *s)
> +{
> + return parse_color_config(s, dom0_colors, ARRAY_SIZE(dom0_colors),
> + &dom0_num_colors);
> +}
> +custom_param("dom0-llc-colors", parse_dom0_colors);
>
> static void print_colors(const unsigned int colors[], unsigned int num_colors)
> {
> @@ -44,9 +104,26 @@ static void print_colors(const unsigned int colors[], unsigned int num_colors)
> printk(" }\n");
> }
>
> +static bool __init check_colors(const unsigned int colors[],
> + unsigned int num_colors)
> +{
> + unsigned int i;
> +
> + for ( i = 0; i < num_colors; i++ )
> + {
> + if ( colors[i] >= max_nr_colors )
> + {
> + printk(XENLOG_ERR "LLC color %u >= %u\n", colors[i], max_nr_colors);
The information about max colors is not printed at the start of day, so a message like:
LLC color 20 >= 10
is not that meaningful. Could we perhaps append (max allowable) or similar to the message?
> + return false;
> + }
> + }
> +
> + return true;
> +}
> +
> void __init llc_coloring_init(void)
> {
> - unsigned int way_size;
> + unsigned int way_size, i;
>
> if ( llc_size && llc_nr_ways )
> {
> @@ -83,6 +160,9 @@ void __init llc_coloring_init(void)
> } else if ( max_nr_colors < 2 )
> panic("Number of LLC colors %u < 2\n", max_nr_colors);
>
> + for ( i = 0; i < max_nr_colors; i++ )
> + default_colors[i] = i;
> +
> arch_llc_coloring_init();
> }
>
> @@ -104,6 +184,43 @@ void domain_dump_llc_colors(const struct domain *d)
> print_colors(d->llc_colors, d->num_llc_colors);
> }
>
> +static void __init domain_set_default_colors(struct domain *d)
> +{
> + printk(XENLOG_WARNING
> + "LLC color config not found for %pd, using all colors\n", d);
> +
> + d->llc_colors = default_colors;
> + d->num_llc_colors = max_nr_colors;
> +}
> +
> +int __init dom0_set_llc_colors(struct domain *d)
> +{
> + typeof(*dom0_colors) *colors;
> +
> + if ( !dom0_num_colors )
> + {
> + domain_set_default_colors(d);
> + return 0;
> + }
> +
> + if ( dom0_num_colors > max_nr_colors ||
NIT: Can we surround dom0_num_colors > max_nr_colors with brackets?
Other than that:
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
~Michal
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 09/12] xen: add cache coloring allocator for domains
2024-11-19 14:13 ` [PATCH v10 09/12] xen: add cache coloring allocator for domains Carlo Nonato
@ 2024-11-28 11:43 ` Jan Beulich
2024-11-29 9:37 ` Carlo Nonato
0 siblings, 1 reply; 36+ messages in thread
From: Jan Beulich @ 2024-11-28 11:43 UTC (permalink / raw)
To: Carlo Nonato, Stefano Stabellini
Cc: andrea.bastoni, marco.solieri, Andrew Cooper, Julien Grall,
Bertrand Marquis, Michal Orzel, Volodymyr Babchuk, xen-devel
On 19.11.2024 15:13, Carlo Nonato wrote:
> --- a/docs/misc/xen-command-line.pandoc
> +++ b/docs/misc/xen-command-line.pandoc
> @@ -270,6 +270,20 @@ and not running softirqs. Reduce this if softirqs are not being run frequently
> enough. Setting this to a high value may cause boot failure, particularly if
> the NMI watchdog is also enabled.
>
> +### buddy-alloc-size (arm64)
> +> `= <size>`
> +
> +> Default: `64M`
> +
> +Amount of memory reserved for the buddy allocator when colored allocator is
> +active. This option is available only when `CONFIG_LLC_COLORING` is enabled.
> +The colored allocator is meant as an alternative to the buddy allocator,
> +because its allocation policy is by definition incompatible with the generic
> +one. Since the Xen heap systems is not colored yet, we need to support the
> +coexistence of the two allocators for now. This parameter, which is optional
> +and for expert only, it's used to set the amount of memory reserved to the
> +buddy allocator.
Every time I see this, and I further see the title of patch 12, I'm confused,
expecting that what's being said here needs adjusting (or even undoing) there.
The issue is that patch 12's title says just "Xen" when, if I'm not mistaken,
it really only means the Xen image.
> --- a/xen/common/Kconfig
> +++ b/xen/common/Kconfig
> @@ -537,4 +537,12 @@ config LLC_COLORS_ORDER
> The default value corresponds to an 8 MiB 16-ways LLC, which should be
> more than what's needed in the general case.
>
> +config BUDDY_ALLOCATOR_SIZE
> + int "Buddy allocator reserved memory size (MiB)" if LLC_COLORING
> + default "0" if !LLC_COLORING
> + default "64"
> + help
> + Amount of memory reserved for the buddy allocator to serve Xen heap,
> + working alongside the colored one.
As the description has nothing in this regard: Why again is it that this
can't simply have "depends on LLC_COLORING"? Even if it ends up with a
value of 0, in an LLC_COLORING=n (or LLC_COLORING entirely absent) .config
I'd find it at least irritating to see this setting to be there.
> @@ -1945,6 +1960,155 @@ static unsigned long avail_heap_pages(
> return free_pages;
> }
>
> +/*************************
> + * COLORED SIDE-ALLOCATOR
> + *
> + * Pages are grouped by LLC color in lists which are globally referred to as the
> + * color heap. Lists are populated in end_boot_allocator().
> + * After initialization there will be N lists where N is the number of
> + * available colors on the platform.
> + */
> +static struct page_list_head *__ro_after_init _color_heap;
> +#define color_heap(color) (&_color_heap[color])
> +
> +static unsigned long *__ro_after_init free_colored_pages;
> +
> +/* Memory required for buddy allocator to work with colored one */
> +static unsigned long __initdata buddy_alloc_size =
> + MB(CONFIG_BUDDY_ALLOCATOR_SIZE);
> +size_param("buddy-alloc-size", buddy_alloc_size);
> +
> +#ifdef CONFIG_LLC_COLORING
According to the (updated) command line doc, this #ifdef needs to move up
so the command line option indeed is unrecognized when !LLC_COLORING.
Question then is whether the variable is actually needed. With the variable
also moved into the #ifdef, the need for BUDDY_ALLOCATOR_SIZE also goes
away when !LLC_COLORING (see the comment on the common/Kconfig change). Of
course you'll then need "#ifndef buddy_alloc_size" in
init_color_heap_pages(), around the respective if() there.
> +#define domain_num_llc_colors(d) ((d)->num_llc_colors)
> +#define domain_llc_color(d, i) ((d)->llc_colors[i])
> +#else
> +#define domain_num_llc_colors(d) 0
> +#define domain_llc_color(d, i) 0
> +#endif
> +
> +static void free_color_heap_page(struct page_info *pg, bool need_scrub)
> +{
> + unsigned int color;
> +
> + color = page_to_llc_color(pg);
> + free_colored_pages[color]++;
> + /*
> + * Head insertion allows re-using cache-hot pages in configurations without
> + * sharing of colors.
> + */
> + page_list_add(pg, color_heap(color));
> +}
Iirc it was me who had asked to keep this and further functions outside of
#ifdef-s, for the compiler's DCE to take care of. However, with all that
Misra work that has been going on since then, I now wonder what Misra
thinks of this: When PGC_colored is 0, functions like the above are
unreachable, and any code or data the compiler doesn't manage to eliminate
would be dead. Imo if the code is to remain as is, correctness Misra-wise
may want mentioning in the description (this isn't the only place we have
such, so an overall position towards this is going to be needed).
> +static void __init init_color_heap_pages(struct page_info *pg,
> + unsigned long nr_pages)
> +{
> + unsigned int i;
> + bool need_scrub = opt_bootscrub == BOOTSCRUB_IDLE;
> +
> + if ( buddy_alloc_size >= PAGE_SIZE )
> + {
> + unsigned long buddy_pages = min(PFN_DOWN(buddy_alloc_size), nr_pages);
> +
> + init_heap_pages(pg, buddy_pages);
> + nr_pages -= buddy_pages;
> + buddy_alloc_size -= buddy_pages << PAGE_SHIFT;
> + pg += buddy_pages;
> + }
> +
> + if ( !_color_heap )
> + {
> + unsigned int max_nr_colors = get_max_nr_llc_colors();
> +
> + _color_heap = xmalloc_array(struct page_list_head, max_nr_colors);
> + free_colored_pages = xzalloc_array(unsigned long, max_nr_colors);
At this point, xvmalloc() and friends want using by all new code, unless
explicitly justified otherwise.
> + if ( !_color_heap || !free_colored_pages )
> + panic("Can't allocate colored heap. Buddy reserved size is too low");
> +
> + for ( i = 0; i < max_nr_colors; i++ )
> + INIT_PAGE_LIST_HEAD(color_heap(i));
While for this loop i being unsigned int is okay, I fear that ...
> + }
> +
> + for ( i = 0; i < nr_pages; i++ )
> + {
> + pg[i].count_info = PGC_colored;
> + free_color_heap_page(&pg[i], need_scrub);
> + }
... for this loop it isn't.
Jan
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 12/12] xen/arm: add cache coloring support for Xen
2024-11-19 14:13 ` [PATCH v10 12/12] xen/arm: add cache coloring support for Xen Carlo Nonato
@ 2024-11-28 11:53 ` Jan Beulich
0 siblings, 0 replies; 36+ messages in thread
From: Jan Beulich @ 2024-11-28 11:53 UTC (permalink / raw)
To: Carlo Nonato
Cc: andrea.bastoni, marco.solieri, Stefano Stabellini, Julien Grall,
Bertrand Marquis, Michal Orzel, Volodymyr Babchuk, Andrew Cooper,
xen-devel
On 19.11.2024 15:13, Carlo Nonato wrote:
> Add the cache coloring support for Xen physical space.
>
> Since Xen must be relocated to a new physical space, some relocation
> functionalities must be brought back:
> - the virtual address of the new space is taken from 0c18fb76323b
> ("xen/arm: Remove unused BOOT_RELOC_VIRT_START").
> - relocate_xen() and get_xen_paddr() are taken from f60658c6ae47
> ("xen/arm: Stop relocating Xen").
>
> setup_pagetables() must be adapted for coloring and for relocation. Runtime
> page tables are used to map the colored space, but they are also linked in
> boot tables so that the new space is temporarily available for relocation.
> This implies that Xen protection must happen after the copy.
>
> Finally, since the alternative framework needs to remap the Xen text and
> inittext sections, this operation must be done in a coloring-aware way.
> The function xen_remap_colored() is introduced for that.
>
> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
Reviewed-by: Jan Beulich <jbeulich@suse.com> # common
preferably with ...
> @@ -353,6 +355,22 @@ unsigned int get_max_nr_llc_colors(void)
> return max_nr_colors;
> }
>
> +mfn_t __init xen_colored_mfn(mfn_t mfn)
> +{
> + unsigned int i, color = mfn_to_color(mfn);
> +
> + for ( i = 0; i < xen_num_colors; i++ )
> + {
> + if ( color == xen_colors[i] )
> + return mfn;
> + else if ( color < xen_colors[i] )
... "else" dropped from here.
Jan
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 07/12] xen/arm: add support for cache coloring configuration via device-tree
2024-11-19 14:13 ` [PATCH v10 07/12] xen/arm: add support for cache coloring configuration via device-tree Carlo Nonato
@ 2024-11-28 12:00 ` Michal Orzel
0 siblings, 0 replies; 36+ messages in thread
From: Michal Orzel @ 2024-11-28 12:00 UTC (permalink / raw)
To: Carlo Nonato, xen-devel
Cc: andrea.bastoni, marco.solieri, Stefano Stabellini, Julien Grall,
Bertrand Marquis, Volodymyr Babchuk, Andrew Cooper, Jan Beulich
On 19/11/2024 15:13, Carlo Nonato wrote:
>
>
> Add the "llc-colors" Device Tree attribute to express DomUs and Dom0less
NIT: s/attribute/property/
> color configurations.
>
> Based on original work from: Luca Miccio <lucmiccio@gmail.com>
>
> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> Signed-off-by: Marco Solieri <marco.solieri@minervasys.tech>
> Reviewed-by: Jan Beulich <jbeulich@suse.com> # non-Arm
> ---
> v10:
> - no changes
> v9:
> - use best-effort allocation in domain_set_llc_colors_from_str()
> v8:
> - fixed memory leak on error path of domain_set_llc_colors_from_str()
> - realloc colors array after parsing from string to reduce memory usage
> v7:
> - removed alloc_colors() helper usage from domain_set_llc_colors_from_str()
> v6:
> - rewrote domain_set_llc_colors_from_str() to be more explicit
> v5:
> - static-mem check has been moved in a previous patch
> - added domain_set_llc_colors_from_str() to set colors after domain creation
> ---
> docs/misc/arm/device-tree/booting.txt | 4 +++
> docs/misc/cache-coloring.rst | 48 +++++++++++++++++++++++++++
> xen/arch/arm/dom0less-build.c | 10 ++++++
> xen/common/llc-coloring.c | 41 +++++++++++++++++++++++
> xen/include/xen/llc-coloring.h | 1 +
> xen/include/xen/xmalloc.h | 12 +++++++
> 6 files changed, 116 insertions(+)
>
> diff --git a/docs/misc/arm/device-tree/booting.txt b/docs/misc/arm/device-tree/booting.txt
> index 3a04f5c57f..9085645433 100644
> --- a/docs/misc/arm/device-tree/booting.txt
> +++ b/docs/misc/arm/device-tree/booting.txt
> @@ -162,6 +162,10 @@ with the following properties:
>
> An integer specifying the number of vcpus to allocate to the guest.
>
> +- llc-colors
> + A string specifying the LLC color configuration for the guest.
> + Refer to docs/misc/cache_coloring.rst for syntax.
Shouldn't we make it clear that this property is Arm64 only?
> +
> - vpl011
>
> An empty property to enable/disable a virtual pl011 for the guest to
> diff --git a/docs/misc/cache-coloring.rst b/docs/misc/cache-coloring.rst
> index c5fb33996c..c14b6ee2f4 100644
> --- a/docs/misc/cache-coloring.rst
> +++ b/docs/misc/cache-coloring.rst
> @@ -12,6 +12,7 @@ If needed, change the maximum number of colors with
> ``CONFIG_LLC_COLORS_ORDER=<n>``.
>
> Runtime configuration is done via `Command line parameters`_.
> +For DomUs follow `DomUs configuration`_.
>
> Background
> **********
> @@ -147,6 +148,53 @@ LLC specs can be manually set via the above command line parameters. This
> bypasses any auto-probing and it's used to overcome failing situations, such as
> flawed probing logic, or for debugging/testing purposes.
>
> +DomUs configuration
> +*******************
> +
> +DomUs colors can be set either in the ``xl`` configuration file (documentation
> +at `docs/man/xl.cfg.pod.5.in`) or via Device Tree, also for Dom0less
> +configurations (documentation at `docs/misc/arm/device-tree/booting.txt`) using
NIT: I think you can remove the part "also for Dom0less configurations" if you already
mentioned "or via Device Tree".
Other than that:
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
~Michal
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 03/12] xen/arm: permit non direct-mapped Dom0 construction
2024-11-28 10:34 ` Michal Orzel
@ 2024-11-29 9:06 ` Carlo Nonato
0 siblings, 0 replies; 36+ messages in thread
From: Carlo Nonato @ 2024-11-29 9:06 UTC (permalink / raw)
To: Michal Orzel
Cc: xen-devel, andrea.bastoni, marco.solieri, Stefano Stabellini,
Julien Grall, Bertrand Marquis, Volodymyr Babchuk
Hi Michal,
On Thu, Nov 28, 2024 at 11:34 AM Michal Orzel <michal.orzel@amd.com> wrote:
> On 19/11/2024 15:13, Carlo Nonato wrote:
> >
> >
> > Cache coloring requires Dom0 not to be direct-mapped because of its non
> > contiguous mapping nature, so allocate_memory() is needed in this case.
> > 8d2c3ab18cc1 ("arm/dom0less: put dom0less feature code in a separate module")
> > moved allocate_memory() in dom0less_build.c. In order to use it
> > in Dom0 construction bring it back to domain_build.c and declare it in
> > domain_build.h.
> >
> > Take the opportunity to adapt the implementation of allocate_memory() so
> > that it uses the host layout when called on the hwdom, via
> > find_unallocated_memory().
> >
> > Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> > ---
> > v10:
> > - fixed a compilation bug that happened when dom0less support was disabled
> > v9:
> > - no changes
> > v8:
> > - patch adapted to new changes to allocate_memory()
> > v7:
> > - allocate_memory() now uses the host layout when called on the hwdom
> > v6:
> > - new patch
> > ---
> > xen/arch/arm/dom0less-build.c | 44 ------------
> > xen/arch/arm/domain_build.c | 96 +++++++++++++++++++++++--
> > xen/arch/arm/include/asm/domain_build.h | 1 +
> > 3 files changed, 93 insertions(+), 48 deletions(-)
> >
> > diff --git a/xen/arch/arm/dom0less-build.c b/xen/arch/arm/dom0less-build.c
> > index d93a85434e..67b1503647 100644
> > --- a/xen/arch/arm/dom0less-build.c
> > +++ b/xen/arch/arm/dom0less-build.c
> > @@ -49,50 +49,6 @@ bool __init is_dom0less_mode(void)
> > return ( !dom0found && domUfound );
> > }
> >
> > -static void __init allocate_memory(struct domain *d, struct kernel_info *kinfo)
> > -{
> > - struct membanks *mem = kernel_info_get_mem(kinfo);
> > - unsigned int i;
> > - paddr_t bank_size;
> > -
> > - printk(XENLOG_INFO "Allocating mappings totalling %ldMB for %pd:\n",
> > - /* Don't want format this as PRIpaddr (16 digit hex) */
> > - (unsigned long)(kinfo->unassigned_mem >> 20), d);
> > -
> > - mem->nr_banks = 0;
> > - bank_size = MIN(GUEST_RAM0_SIZE, kinfo->unassigned_mem);
> > - if ( !allocate_bank_memory(kinfo, gaddr_to_gfn(GUEST_RAM0_BASE),
> > - bank_size) )
> > - goto fail;
> > -
> > - bank_size = MIN(GUEST_RAM1_SIZE, kinfo->unassigned_mem);
> > - if ( !allocate_bank_memory(kinfo, gaddr_to_gfn(GUEST_RAM1_BASE),
> > - bank_size) )
> > - goto fail;
> > -
> > - if ( kinfo->unassigned_mem )
> > - goto fail;
> > -
> > - for( i = 0; i < mem->nr_banks; i++ )
> > - {
> > - printk(XENLOG_INFO "%pd BANK[%d] %#"PRIpaddr"-%#"PRIpaddr" (%ldMB)\n",
> > - d,
> > - i,
> > - mem->bank[i].start,
> > - mem->bank[i].start + mem->bank[i].size,
> > - /* Don't want format this as PRIpaddr (16 digit hex) */
> > - (unsigned long)(mem->bank[i].size >> 20));
> > - }
> > -
> > - return;
> > -
> > -fail:
> > - panic("Failed to allocate requested domain memory."
> > - /* Don't want format this as PRIpaddr (16 digit hex) */
> > - " %ldKB unallocated. Fix the VMs configurations.\n",
> > - (unsigned long)kinfo->unassigned_mem >> 10);
> > -}
> > -
> > #ifdef CONFIG_VGICV2
> > static int __init make_gicv2_domU_node(struct kernel_info *kinfo)
> > {
> > diff --git a/xen/arch/arm/domain_build.c b/xen/arch/arm/domain_build.c
> > index 2c30792de8..a95376dcdc 100644
> > --- a/xen/arch/arm/domain_build.c
> > +++ b/xen/arch/arm/domain_build.c
> > @@ -416,7 +416,6 @@ static void __init allocate_memory_11(struct domain *d,
> > }
> > }
> >
> > -#ifdef CONFIG_DOM0LESS_BOOT
> > bool __init allocate_domheap_memory(struct domain *d, paddr_t tot_size,
> > alloc_domheap_mem_cb cb, void *extra)
> > {
> > @@ -508,7 +507,6 @@ bool __init allocate_bank_memory(struct kernel_info *kinfo, gfn_t sgfn,
> >
> > return true;
> > }
> > -#endif
> >
> > /*
> > * When PCI passthrough is available we want to keep the
> > @@ -1003,6 +1001,93 @@ out:
> > return res;
> > }
> >
> > +void __init allocate_memory(struct domain *d, struct kernel_info *kinfo)
> > +{
> > + struct membanks *mem = kernel_info_get_mem(kinfo);
> > + unsigned int i, nr_banks = 2;
> Instead of hardcoding, shouldn't it be GUEST_RAM_BANKS?
Right.
> Also, the second bank won't work with CONFIG_ARM_PA_BITS_32 which limits PA to 32bit.
How is this being addressed in the current allocate_memory?
Also, LLC_COLORING depends on ARM_64. ARM_PA_BITS_32 requires ARM_32, so the
two configurations should be incompatible as of now.
> > + paddr_t bank_start, bank_size;
> > + struct membanks *hwdom_ext_regions = NULL;
> AFAICT you search for free memory. Naming it as extended regions is not a good choice.
> Instead, hwdom_free_mem?
Yes.
> > +
> > + printk(XENLOG_INFO "Allocating mappings totalling %ldMB for %pd:\n",
> > + /* Don't want format this as PRIpaddr (16 digit hex) */
> > + (unsigned long)(kinfo->unassigned_mem >> 20), d);
> > +
> > + mem->nr_banks = 0;
> > + /*
> > + * Use host memory layout for hwdom. Only case for this is when LLC coloring
> > + * is enabled.
> > + */
> > + if ( is_hardware_domain(d) )
> > + {
> > + ASSERT(llc_coloring_enabled);
> > +
> > + hwdom_ext_regions = xzalloc_flex_struct(struct membanks, bank,
> > + NR_MEM_BANKS);
> > + if ( !hwdom_ext_regions )
> > + goto fail;
> empty line here please
>
> > + hwdom_ext_regions->max_banks = NR_MEM_BANKS;
> > +
> > + if ( find_unallocated_memory(kinfo, hwdom_ext_regions) )
> If you reuse find_unallocated_memory for a purpose other than extended regions, I think
> the description of this function should change as well as comments inside.
I can definetely change that.
> Today, the function gets all RAM and exclude dom0 RAM (in your case is 0 at this point, reserved memory,
> static shmem and gnttab (in your case is 0 at this point). I think we cannot get away without
> making this function more generic. Maybe it should take as a parameter struct membanks * array?
> Also, the callback add_ext_regions() may not be suited for all purposes (i.e. it takes only banks
> > 64MB into account). I know that there will be more use cases for a function that will return the
> free memory for domains. As an example, today, for direct mapped domains we hardcode the gnttab region
> (only dom0 is special cased). This should not be like that. We would need to find a free memory region
> to expose as gnttab.
For the current cases (including llc coloring) it works. If there are no
objections, a TODO plus comments and description changes you talked above is
probably sufficient to cover the (current) use-cases and "ensure" this is not
forgotten for the future extension you mention.
> > + goto fail;
> > +
> > + nr_banks = hwdom_ext_regions->nr_banks;
> > + }
> > +
> > + for ( i = 0; kinfo->unassigned_mem > 0 && nr_banks > 0; i++, nr_banks-- )
> > + {
> > + if ( is_hardware_domain(d) )
> > + {
> > + bank_start = hwdom_ext_regions->bank[i].start;
> > + bank_size = hwdom_ext_regions->bank[i].size;
> > +
> > + if ( bank_size < min_t(paddr_t, kinfo->unassigned_mem, MB(128)) )
> I would expect some explanation.
>
> > + continue;
> > + }
> > + else
> > + {
> > + if ( i == 0 )
> > + {
> > + bank_start = GUEST_RAM0_BASE;
> > + bank_size = GUEST_RAM0_SIZE;
> > + }
> > + else if ( i == 1 )
> > + {
> > + bank_start = GUEST_RAM1_BASE;
> > + bank_size = GUEST_RAM1_SIZE;
> > + }
> > + else
> > + goto fail;
> This could be simplified:
> const uint64_t bankbase[] = GUEST_RAM_BANK_BASES;
> const uint64_t banksize[] = GUEST_RAM_BANK_SIZES;
>
> if ( i >= GUEST_RAM_BANKS )
> goto fail;
>
> bank_start = bankbase[i];
> bank_size = banksize[i];
Ok.
> This patch requires also opinion of other maintainers.
>
> ~Michal
Thanks.
- Carlo
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 08/12] xen/page_alloc: introduce preserved page flags macro
2024-11-28 11:04 ` Jan Beulich
@ 2024-11-29 9:32 ` Carlo Nonato
2024-11-29 11:09 ` Jan Beulich
0 siblings, 1 reply; 36+ messages in thread
From: Carlo Nonato @ 2024-11-29 9:32 UTC (permalink / raw)
To: Jan Beulich
Cc: andrea.bastoni, marco.solieri, Andrew Cooper, Julien Grall,
Stefano Stabellini, xen-devel
Hi Jan,
On Thu, Nov 28, 2024 at 12:05 PM Jan Beulich <jbeulich@suse.com> wrote:
>
> On 19.11.2024 15:13, Carlo Nonato wrote:
> > PGC_static, PGC_extra and PGC_broken need to be preserved when assigning a
> > page. Define a new macro that groups those flags and use it instead of or'ing
> > every time.
> >
> > To make preserved flags even more meaningful, they are kept also when
> > switching state in mark_page_free().
> > Enforce the removal of PGC_extra before freeing domain pages as this is
> > considered an error and can cause ASSERT violations.
> >
> > Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> > ---
> > v10:
> > - fixed commit message
> > v9:
> > - add PGC_broken to PGC_preserved
> > - clear PGC_extra in alloc_domheap_pages() only if MEMF_no_refcount is set
> > v8:
> > - fixed PGC_extra ASSERT fail in alloc_domheap_pages() by removing PGC_extra
> > before freeing
> > v7:
> > - PGC_preserved used also in mark_page_free()
> > v6:
> > - preserved_flags renamed to PGC_preserved
> > - PGC_preserved is used only in assign_pages()
> > v5:
> > - new patch
> > ---
> > xen/common/page_alloc.c | 19 ++++++++++++++-----
> > 1 file changed, 14 insertions(+), 5 deletions(-)
> >
> > diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
> > index 7b911b5ed9..34cd473150 100644
> > --- a/xen/common/page_alloc.c
> > +++ b/xen/common/page_alloc.c
> > @@ -160,6 +160,7 @@
> > #endif
> >
> > #define PGC_no_buddy_merge PGC_static
> > +#define PGC_preserved (PGC_extra | PGC_static | PGC_broken)
> >
> > #ifndef PGT_TYPE_INFO_INITIALIZER
> > #define PGT_TYPE_INFO_INITIALIZER 0
> > @@ -1427,12 +1428,11 @@ static bool mark_page_free(struct page_info *pg, mfn_t mfn)
> > {
> > case PGC_state_inuse:
> > BUG_ON(pg->count_info & PGC_broken);
> > - pg->count_info = PGC_state_free;
> > + pg->count_info = PGC_state_free | (pg->count_info & PGC_preserved);
> > break;
>
> PGC_extra doesn't want preserving here. Since it's mark_page_free(), and
> since PGC_extra is removed before freeing, the state change is apparently
> fine. But an assertion may want adding, for documentation purposes if
> nothing else.
>
> Alternatively it may make sense to indeed exclude PGC_extra here. In fact
> PGC_static doesn't need using here either, as unprepare_staticmem_pages()
> will explicitly set it again anyway.
>
> Hence I wonder whether the change here really is necessary (one will then
> be needed in the next patch aiui, when PGC_colored is introduced). Which
> would then eliminate the need for the final two hunks of the patch, I
> think.
>
> > case PGC_state_offlining:
> > - pg->count_info = (pg->count_info & PGC_broken) |
> > - PGC_state_offlined;
> > + pg->count_info = (pg->count_info & PGC_preserved) | PGC_state_offlined;
> > pg_offlined = true;
> > break;
>
> I'm similarly unconvinced that anything other than PGC_broken (and
> subsequently perhaps PGC_colored) would need preserving here.
Yes, we (re)checked the code and also believe that the introduction of
PGC_preserved is generating more confusion (and code) then the actual logical
help it provides.
We'll remove this patch and integrate PGC_colored explicitly in the flags to
be preserved. This avoid the clumsy logic of preserving something (extra)
when it's not needed and then handling the special case to remove it
explicitly.
Basically my goal is to go back to v4 where this patch didn't exist.
> Jan
Thanks.
- Carlo
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 09/12] xen: add cache coloring allocator for domains
2024-11-28 11:43 ` Jan Beulich
@ 2024-11-29 9:37 ` Carlo Nonato
2024-11-29 11:17 ` Jan Beulich
0 siblings, 1 reply; 36+ messages in thread
From: Carlo Nonato @ 2024-11-29 9:37 UTC (permalink / raw)
To: Jan Beulich
Cc: Stefano Stabellini, andrea.bastoni, marco.solieri, Andrew Cooper,
Julien Grall, Bertrand Marquis, Michal Orzel, Volodymyr Babchuk,
xen-devel
Hi Jan,
On Thu, Nov 28, 2024 at 12:43 PM Jan Beulich <jbeulich@suse.com> wrote:
>
> On 19.11.2024 15:13, Carlo Nonato wrote:
> > --- a/docs/misc/xen-command-line.pandoc
> > +++ b/docs/misc/xen-command-line.pandoc
> > @@ -270,6 +270,20 @@ and not running softirqs. Reduce this if softirqs are not being run frequently
> > enough. Setting this to a high value may cause boot failure, particularly if
> > the NMI watchdog is also enabled.
> >
> > +### buddy-alloc-size (arm64)
> > +> `= <size>`
> > +
> > +> Default: `64M`
> > +
> > +Amount of memory reserved for the buddy allocator when colored allocator is
> > +active. This option is available only when `CONFIG_LLC_COLORING` is enabled.
> > +The colored allocator is meant as an alternative to the buddy allocator,
> > +because its allocation policy is by definition incompatible with the generic
> > +one. Since the Xen heap systems is not colored yet, we need to support the
> > +coexistence of the two allocators for now. This parameter, which is optional
> > +and for expert only, it's used to set the amount of memory reserved to the
> > +buddy allocator.
>
> Every time I see this, and I further see the title of patch 12, I'm confused,
> expecting that what's being said here needs adjusting (or even undoing) there.
> The issue is that patch 12's title says just "Xen" when, if I'm not mistaken,
> it really only means the Xen image.
I'll change the patch 12 title to reflect what you said.
> > --- a/xen/common/Kconfig
> > +++ b/xen/common/Kconfig
> > @@ -537,4 +537,12 @@ config LLC_COLORS_ORDER
> > The default value corresponds to an 8 MiB 16-ways LLC, which should be
> > more than what's needed in the general case.
> >
> > +config BUDDY_ALLOCATOR_SIZE
> > + int "Buddy allocator reserved memory size (MiB)" if LLC_COLORING
> > + default "0" if !LLC_COLORING
> > + default "64"
> > + help
> > + Amount of memory reserved for the buddy allocator to serve Xen heap,
> > + working alongside the colored one.
>
> As the description has nothing in this regard: Why again is it that this
> can't simply have "depends on LLC_COLORING"? Even if it ends up with a
> value of 0, in an LLC_COLORING=n (or LLC_COLORING entirely absent) .config
> I'd find it at least irritating to see this setting to be there.
Quoting you from v8 (6 months ago, a lot of time ago I know, link here
https://patchew.org/Xen/20240502165533.319988-1-carlo.nonato@minervasys.tech/20240502165533.319988-10-carlo.nonato@minervasys.tech/#5c16f53f-3bb0-49d6-b174-b0e8c6ceff4c@suse.com):
> > +/* Memory required for buddy allocator to work with colored one */
> > +#ifdef CONFIG_LLC_COLORING
> > +static unsigned long __initdata buddy_alloc_size =
> > + MB(CONFIG_BUDDY_ALLOCATOR_SIZE);
>
> I think it would be quite nice if this and ...
>
> > +size_param("buddy-alloc-size", buddy_alloc_size);
> > +
> > +#define domain_num_llc_colors(d) (d)->num_llc_colors
> > +#define domain_llc_color(d, i) (d)->llc_colors[i]
> > +#else
> > +static unsigned long __initdata buddy_alloc_size;
>
> ... this were folded. Which I think would be possible if the Kconfig
> addition went like this:
>
> config BUDDY_ALLOCATOR_SIZE
> int "Buddy allocator reserved memory size (MiB)" if LLC_COLORING
> default "0" if !LLC_COLORING
> default "64"
But I know that...
> > @@ -1945,6 +1960,155 @@ static unsigned long avail_heap_pages(
> > return free_pages;
> > }
> >
> > +/*************************
> > + * COLORED SIDE-ALLOCATOR
> > + *
> > + * Pages are grouped by LLC color in lists which are globally referred to as the
> > + * color heap. Lists are populated in end_boot_allocator().
> > + * After initialization there will be N lists where N is the number of
> > + * available colors on the platform.
> > + */
> > +static struct page_list_head *__ro_after_init _color_heap;
> > +#define color_heap(color) (&_color_heap[color])
> > +
> > +static unsigned long *__ro_after_init free_colored_pages;
> > +
> > +/* Memory required for buddy allocator to work with colored one */
> > +static unsigned long __initdata buddy_alloc_size =
> > + MB(CONFIG_BUDDY_ALLOCATOR_SIZE);
> > +size_param("buddy-alloc-size", buddy_alloc_size);
> > +
> > +#ifdef CONFIG_LLC_COLORING
>
> According to the (updated) command line doc, this #ifdef needs to move up
> so the command line option indeed is unrecognized when !LLC_COLORING.
> Question then is whether the variable is actually needed. With the variable
> also moved into the #ifdef, the need for BUDDY_ALLOCATOR_SIZE also goes
> away when !LLC_COLORING (see the comment on the common/Kconfig change). Of
> course you'll then need "#ifndef buddy_alloc_size" in
> init_color_heap_pages(), around the respective if() there.
buddy-alloc-size must not be parsed when !LLC_COLORING. I'll change the code
accordingly to what you said.
> > +#define domain_num_llc_colors(d) ((d)->num_llc_colors)
> > +#define domain_llc_color(d, i) ((d)->llc_colors[i])
> > +#else
> > +#define domain_num_llc_colors(d) 0
> > +#define domain_llc_color(d, i) 0
> > +#endif
> > +
> > +static void free_color_heap_page(struct page_info *pg, bool need_scrub)
> > +{
> > + unsigned int color;
> > +
> > + color = page_to_llc_color(pg);
> > + free_colored_pages[color]++;
> > + /*
> > + * Head insertion allows re-using cache-hot pages in configurations without
> > + * sharing of colors.
> > + */
> > + page_list_add(pg, color_heap(color));
> > +}
>
> Iirc it was me who had asked to keep this and further functions outside of
> #ifdef-s, for the compiler's DCE to take care of. However, with all that
> Misra work that has been going on since then, I now wonder what Misra
> thinks of this: When PGC_colored is 0, functions like the above are
> unreachable, and any code or data the compiler doesn't manage to eliminate
> would be dead. Imo if the code is to remain as is, correctness Misra-wise
> may want mentioning in the description (this isn't the only place we have
> such, so an overall position towards this is going to be needed).
For the time being, I'll add a note that can be updated when the overall
position is found.
> > +static void __init init_color_heap_pages(struct page_info *pg,
> > + unsigned long nr_pages)
> > +{
> > + unsigned int i;
> > + bool need_scrub = opt_bootscrub == BOOTSCRUB_IDLE;
> > +
> > + if ( buddy_alloc_size >= PAGE_SIZE )
> > + {
> > + unsigned long buddy_pages = min(PFN_DOWN(buddy_alloc_size), nr_pages);
> > +
> > + init_heap_pages(pg, buddy_pages);
> > + nr_pages -= buddy_pages;
> > + buddy_alloc_size -= buddy_pages << PAGE_SHIFT;
> > + pg += buddy_pages;
> > + }
> > +
> > + if ( !_color_heap )
> > + {
> > + unsigned int max_nr_colors = get_max_nr_llc_colors();
> > +
> > + _color_heap = xmalloc_array(struct page_list_head, max_nr_colors);
> > + free_colored_pages = xzalloc_array(unsigned long, max_nr_colors);
>
> At this point, xvmalloc() and friends want using by all new code, unless
> explicitly justified otherwise.
Yes.
> > + if ( !_color_heap || !free_colored_pages )
> > + panic("Can't allocate colored heap. Buddy reserved size is too low");
> > +
> > + for ( i = 0; i < max_nr_colors; i++ )
> > + INIT_PAGE_LIST_HEAD(color_heap(i));
>
> While for this loop i being unsigned int is okay, I fear that ...
>
> > + }
> > +
> > + for ( i = 0; i < nr_pages; i++ )
> > + {
> > + pg[i].count_info = PGC_colored;
> > + free_color_heap_page(&pg[i], need_scrub);
> > + }
>
> ... for this loop it isn't.
Ok.
> Jan
Thanks.
- Carlo
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 08/12] xen/page_alloc: introduce preserved page flags macro
2024-11-29 9:32 ` Carlo Nonato
@ 2024-11-29 11:09 ` Jan Beulich
2024-12-02 12:51 ` Andrea Bastoni
0 siblings, 1 reply; 36+ messages in thread
From: Jan Beulich @ 2024-11-29 11:09 UTC (permalink / raw)
To: Carlo Nonato
Cc: andrea.bastoni, marco.solieri, Andrew Cooper, Julien Grall,
Stefano Stabellini, xen-devel
On 29.11.2024 10:32, Carlo Nonato wrote:
> Hi Jan,
>
> On Thu, Nov 28, 2024 at 12:05 PM Jan Beulich <jbeulich@suse.com> wrote:
>>
>> On 19.11.2024 15:13, Carlo Nonato wrote:
>>> PGC_static, PGC_extra and PGC_broken need to be preserved when assigning a
>>> page. Define a new macro that groups those flags and use it instead of or'ing
>>> every time.
>>>
>>> To make preserved flags even more meaningful, they are kept also when
>>> switching state in mark_page_free().
>>> Enforce the removal of PGC_extra before freeing domain pages as this is
>>> considered an error and can cause ASSERT violations.
>>>
>>> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
>>> ---
>>> v10:
>>> - fixed commit message
>>> v9:
>>> - add PGC_broken to PGC_preserved
>>> - clear PGC_extra in alloc_domheap_pages() only if MEMF_no_refcount is set
>>> v8:
>>> - fixed PGC_extra ASSERT fail in alloc_domheap_pages() by removing PGC_extra
>>> before freeing
>>> v7:
>>> - PGC_preserved used also in mark_page_free()
>>> v6:
>>> - preserved_flags renamed to PGC_preserved
>>> - PGC_preserved is used only in assign_pages()
>>> v5:
>>> - new patch
>>> ---
>>> xen/common/page_alloc.c | 19 ++++++++++++++-----
>>> 1 file changed, 14 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
>>> index 7b911b5ed9..34cd473150 100644
>>> --- a/xen/common/page_alloc.c
>>> +++ b/xen/common/page_alloc.c
>>> @@ -160,6 +160,7 @@
>>> #endif
>>>
>>> #define PGC_no_buddy_merge PGC_static
>>> +#define PGC_preserved (PGC_extra | PGC_static | PGC_broken)
>>>
>>> #ifndef PGT_TYPE_INFO_INITIALIZER
>>> #define PGT_TYPE_INFO_INITIALIZER 0
>>> @@ -1427,12 +1428,11 @@ static bool mark_page_free(struct page_info *pg, mfn_t mfn)
>>> {
>>> case PGC_state_inuse:
>>> BUG_ON(pg->count_info & PGC_broken);
>>> - pg->count_info = PGC_state_free;
>>> + pg->count_info = PGC_state_free | (pg->count_info & PGC_preserved);
>>> break;
>>
>> PGC_extra doesn't want preserving here. Since it's mark_page_free(), and
>> since PGC_extra is removed before freeing, the state change is apparently
>> fine. But an assertion may want adding, for documentation purposes if
>> nothing else.
>>
>> Alternatively it may make sense to indeed exclude PGC_extra here. In fact
>> PGC_static doesn't need using here either, as unprepare_staticmem_pages()
>> will explicitly set it again anyway.
>>
>> Hence I wonder whether the change here really is necessary (one will then
>> be needed in the next patch aiui, when PGC_colored is introduced). Which
>> would then eliminate the need for the final two hunks of the patch, I
>> think.
>>
>>> case PGC_state_offlining:
>>> - pg->count_info = (pg->count_info & PGC_broken) |
>>> - PGC_state_offlined;
>>> + pg->count_info = (pg->count_info & PGC_preserved) | PGC_state_offlined;
>>> pg_offlined = true;
>>> break;
>>
>> I'm similarly unconvinced that anything other than PGC_broken (and
>> subsequently perhaps PGC_colored) would need preserving here.
>
> Yes, we (re)checked the code and also believe that the introduction of
> PGC_preserved is generating more confusion (and code) then the actual logical
> help it provides.
>
> We'll remove this patch and integrate PGC_colored explicitly in the flags to
> be preserved. This avoid the clumsy logic of preserving something (extra)
> when it's not needed and then handling the special case to remove it
> explicitly.
> Basically my goal is to go back to v4 where this patch didn't exist.
Hmm, no, I don't think I said anything in the direction of removing PGC_preserved
again. I merely think you went too far in where it actually wants using.
Jan
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 09/12] xen: add cache coloring allocator for domains
2024-11-29 9:37 ` Carlo Nonato
@ 2024-11-29 11:17 ` Jan Beulich
0 siblings, 0 replies; 36+ messages in thread
From: Jan Beulich @ 2024-11-29 11:17 UTC (permalink / raw)
To: Carlo Nonato
Cc: Stefano Stabellini, andrea.bastoni, marco.solieri, Andrew Cooper,
Julien Grall, Bertrand Marquis, Michal Orzel, Volodymyr Babchuk,
xen-devel
On 29.11.2024 10:37, Carlo Nonato wrote:
> On Thu, Nov 28, 2024 at 12:43 PM Jan Beulich <jbeulich@suse.com> wrote:
>> On 19.11.2024 15:13, Carlo Nonato wrote:
>>> --- a/xen/common/Kconfig
>>> +++ b/xen/common/Kconfig
>>> @@ -537,4 +537,12 @@ config LLC_COLORS_ORDER
>>> The default value corresponds to an 8 MiB 16-ways LLC, which should be
>>> more than what's needed in the general case.
>>>
>>> +config BUDDY_ALLOCATOR_SIZE
>>> + int "Buddy allocator reserved memory size (MiB)" if LLC_COLORING
>>> + default "0" if !LLC_COLORING
>>> + default "64"
>>> + help
>>> + Amount of memory reserved for the buddy allocator to serve Xen heap,
>>> + working alongside the colored one.
>>
>> As the description has nothing in this regard: Why again is it that this
>> can't simply have "depends on LLC_COLORING"? Even if it ends up with a
>> value of 0, in an LLC_COLORING=n (or LLC_COLORING entirely absent) .config
>> I'd find it at least irritating to see this setting to be there.
>
> Quoting you from v8 (6 months ago, a lot of time ago I know, link here
> https://patchew.org/Xen/20240502165533.319988-1-carlo.nonato@minervasys.tech/20240502165533.319988-10-carlo.nonato@minervasys.tech/#5c16f53f-3bb0-49d6-b174-b0e8c6ceff4c@suse.com):
>>> +/* Memory required for buddy allocator to work with colored one */
>>> +#ifdef CONFIG_LLC_COLORING
>>> +static unsigned long __initdata buddy_alloc_size =
>>> + MB(CONFIG_BUDDY_ALLOCATOR_SIZE);
>>
>> I think it would be quite nice if this and ...
>>
>>> +size_param("buddy-alloc-size", buddy_alloc_size);
>>> +
>>> +#define domain_num_llc_colors(d) (d)->num_llc_colors
>>> +#define domain_llc_color(d, i) (d)->llc_colors[i]
>>> +#else
>>> +static unsigned long __initdata buddy_alloc_size;
>>
>> ... this were folded. Which I think would be possible if the Kconfig
>> addition went like this:
>>
>> config BUDDY_ALLOCATOR_SIZE
>> int "Buddy allocator reserved memory size (MiB)" if LLC_COLORING
>> default "0" if !LLC_COLORING
>> default "64"
Hmm, I see. I didn't consider the negative effect back at the time, it seems.
The main goal was anyway to get rid of the multiple instances of
buddy_alloc_size. I'm sure that can be easily achieved also when going back
to the original form of the Kconfig option (at least partly by following what
I said in the initial reply to this v10 patch, I think).
Jan
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 11/12] xen/arm: make consider_modules() available for xen relocation
2024-11-19 14:13 ` [PATCH v10 11/12] xen/arm: make consider_modules() available for xen relocation Carlo Nonato
@ 2024-12-02 8:59 ` Michal Orzel
0 siblings, 0 replies; 36+ messages in thread
From: Michal Orzel @ 2024-12-02 8:59 UTC (permalink / raw)
To: Carlo Nonato, xen-devel
Cc: andrea.bastoni, marco.solieri, Stefano Stabellini, Julien Grall,
Bertrand Marquis, Volodymyr Babchuk
On 19/11/2024 15:13, Carlo Nonato wrote:
>
>
> Cache coloring must physically relocate Xen in order to color the hypervisor
> and consider_modules() is a key function that is needed to find a new
> available physical address.
>
> 672d67f339c0 ("xen/arm: Split MMU-specific setup_mm() and related code out")
> moved consider_modules() under arm32. Move it to mmu/setup.c and make it
> non-static so that it can be used outside.
>
> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
> ---
> v10:
> - no changes
> v9:
> - no changes
> v8:
> - patch adapted to new changes to consider_modules()
> v7:
> - moved consider_modules() to arm/mmu/setup.c
> v6:
> - new patch
> ---
> xen/arch/arm/arm32/mmu/mm.c | 95 +------------------------------
> xen/arch/arm/include/asm/setup.h | 3 +
> xen/arch/arm/mmu/setup.c | 97 ++++++++++++++++++++++++++++++++
> 3 files changed, 101 insertions(+), 94 deletions(-)
>
> diff --git a/xen/arch/arm/arm32/mmu/mm.c b/xen/arch/arm/arm32/mmu/mm.c
> index 063611412b..c5fcd19291 100644
> --- a/xen/arch/arm/arm32/mmu/mm.c
> +++ b/xen/arch/arm/arm32/mmu/mm.c
> @@ -9,6 +9,7 @@
> #include <asm/fixmap.h>
> #include <asm/static-memory.h>
> #include <asm/static-shmem.h>
> +#include <asm/setup.h>
Sort alphabetically.
>
> static unsigned long opt_xenheap_megabytes __initdata;
> integer_param("xenheap_megabytes", opt_xenheap_megabytes);
> @@ -31,100 +32,6 @@ static void __init setup_directmap_mappings(unsigned long base_mfn,
> directmap_virt_end = XENHEAP_VIRT_START + nr_mfns * PAGE_SIZE;
> }
>
> -/*
> - * Returns the end address of the highest region in the range s..e
> - * with required size and alignment that does not conflict with the
> - * modules from first_mod to nr_modules.
> - *
> - * For non-recursive callers first_mod should normally be 0 (all
> - * modules and Xen itself) or 1 (all modules but not Xen).
> - */
> -static paddr_t __init consider_modules(paddr_t s, paddr_t e,
> - uint32_t size, paddr_t align,
> - int first_mod)
> -{
> - const struct membanks *reserved_mem = bootinfo_get_reserved_mem();
> -#ifdef CONFIG_STATIC_SHM
> - const struct membanks *shmem = bootinfo_get_shmem();
> -#endif
> - const struct bootmodules *mi = &bootinfo.modules;
> - int i;
> - int nr;
> -
> - s = (s+align-1) & ~(align-1);
> - e = e & ~(align-1);
> -
> - if ( s > e || e - s < size )
> - return 0;
> -
> - /* First check the boot modules */
> - for ( i = first_mod; i < mi->nr_mods; i++ )
> - {
> - paddr_t mod_s = mi->module[i].start;
> - paddr_t mod_e = mod_s + mi->module[i].size;
> -
> - if ( s < mod_e && mod_s < e )
> - {
> - mod_e = consider_modules(mod_e, e, size, align, i+1);
> - if ( mod_e )
> - return mod_e;
> -
> - return consider_modules(s, mod_s, size, align, i+1);
> - }
> - }
> -
> - /*
> - * i is the current bootmodule we are evaluating, across all
> - * possible kinds of bootmodules.
> - *
> - * When retrieving the corresponding reserved-memory addresses, we
> - * need to index the reserved_mem bank starting from 0, and only counting
> - * the reserved-memory modules. Hence, we need to use i - nr.
> - */
> - nr = mi->nr_mods;
> - for ( ; i - nr < reserved_mem->nr_banks; i++ )
> - {
> - paddr_t r_s = reserved_mem->bank[i - nr].start;
> - paddr_t r_e = r_s + reserved_mem->bank[i - nr].size;
> -
> - if ( s < r_e && r_s < e )
> - {
> - r_e = consider_modules(r_e, e, size, align, i + 1);
> - if ( r_e )
> - return r_e;
> -
> - return consider_modules(s, r_s, size, align, i + 1);
> - }
> - }
> -
> -#ifdef CONFIG_STATIC_SHM
> - nr += reserved_mem->nr_banks;
> - for ( ; i - nr < shmem->nr_banks; i++ )
> - {
> - paddr_t r_s, r_e;
> -
> - r_s = shmem->bank[i - nr].start;
> -
> - /* Shared memory banks can contain INVALID_PADDR as start */
> - if ( INVALID_PADDR == r_s )
> - continue;
> -
> - r_e = r_s + shmem->bank[i - nr].size;
> -
> - if ( s < r_e && r_s < e )
> - {
> - r_e = consider_modules(r_e, e, size, align, i + 1);
> - if ( r_e )
> - return r_e;
> -
> - return consider_modules(s, r_s, size, align, i + 1);
> - }
> - }
> -#endif
> -
> - return e;
> -}
> -
> /*
> * Find a contiguous region that fits in the static heap region with
> * required size and alignment, and return the end address of the region
> diff --git a/xen/arch/arm/include/asm/setup.h b/xen/arch/arm/include/asm/setup.h
> index 64c227d171..0c560d141f 100644
> --- a/xen/arch/arm/include/asm/setup.h
> +++ b/xen/arch/arm/include/asm/setup.h
> @@ -89,6 +89,9 @@ struct init_info
> unsigned int cpuid;
> };
>
> +paddr_t consider_modules(paddr_t s, paddr_t e, uint32_t size, paddr_t align,
> + int first_mod);
> +
> #endif
> /*
> * Local variables:
> diff --git a/xen/arch/arm/mmu/setup.c b/xen/arch/arm/mmu/setup.c
> index 9664e85ee6..1cf62390e3 100644
> --- a/xen/arch/arm/mmu/setup.c
> +++ b/xen/arch/arm/mmu/setup.c
> @@ -6,7 +6,10 @@
> */
>
> #include <xen/init.h>
> +#include <xen/lib.h>
What's in lib.h that consider_modules() requires?
> #include <xen/libfdt/libfdt.h>
> +#include <xen/libfdt/libfdt-xen.h>
Why? AFAICS, this header defines fdt_get_mem_rsv_paddr() which is not used in this file.
> +#include <xen/llc-coloring.h>
Why? There's nothing LLC related here at this point.
With the header inclusions sorted out:
Reviewed-by: Michal Orzel <michal.orzel@amd.com>
~Michal
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 08/12] xen/page_alloc: introduce preserved page flags macro
2024-11-29 11:09 ` Jan Beulich
@ 2024-12-02 12:51 ` Andrea Bastoni
2024-12-02 14:28 ` Jan Beulich
0 siblings, 1 reply; 36+ messages in thread
From: Andrea Bastoni @ 2024-12-02 12:51 UTC (permalink / raw)
To: Jan Beulich, Carlo Nonato
Cc: marco.solieri, Andrew Cooper, Julien Grall, Stefano Stabellini,
xen-devel
Hi,
On 29/11/2024 12:09, Jan Beulich wrote:
> On 29.11.2024 10:32, Carlo Nonato wrote:
>> Hi Jan,
>>
>> On Thu, Nov 28, 2024 at 12:05 PM Jan Beulich <jbeulich@suse.com> wrote:
>>>
>>> On 19.11.2024 15:13, Carlo Nonato wrote:
>>>> PGC_static, PGC_extra and PGC_broken need to be preserved when assigning a
>>>> page. Define a new macro that groups those flags and use it instead of or'ing
>>>> every time.
>>>>
>>>> To make preserved flags even more meaningful, they are kept also when
>>>> switching state in mark_page_free().
>>>> Enforce the removal of PGC_extra before freeing domain pages as this is
>>>> considered an error and can cause ASSERT violations.
>>>>
>>>> Signed-off-by: Carlo Nonato <carlo.nonato@minervasys.tech>
>>>> ---
>>>> v10:
>>>> - fixed commit message
>>>> v9:
>>>> - add PGC_broken to PGC_preserved
>>>> - clear PGC_extra in alloc_domheap_pages() only if MEMF_no_refcount is set
>>>> v8:
>>>> - fixed PGC_extra ASSERT fail in alloc_domheap_pages() by removing PGC_extra
>>>> before freeing
>>>> v7:
>>>> - PGC_preserved used also in mark_page_free()
>>>> v6:
>>>> - preserved_flags renamed to PGC_preserved
>>>> - PGC_preserved is used only in assign_pages()
>>>> v5:
>>>> - new patch
>>>> ---
>>>> xen/common/page_alloc.c | 19 ++++++++++++++-----
>>>> 1 file changed, 14 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
>>>> index 7b911b5ed9..34cd473150 100644
>>>> --- a/xen/common/page_alloc.c
>>>> +++ b/xen/common/page_alloc.c
>>>> @@ -160,6 +160,7 @@
>>>> #endif
>>>>
>>>> #define PGC_no_buddy_merge PGC_static
>>>> +#define PGC_preserved (PGC_extra | PGC_static | PGC_broken)
>>>>
>>>> #ifndef PGT_TYPE_INFO_INITIALIZER
>>>> #define PGT_TYPE_INFO_INITIALIZER 0
>>>> @@ -1427,12 +1428,11 @@ static bool mark_page_free(struct page_info *pg, mfn_t mfn)
>>>> {
>>>> case PGC_state_inuse:
>>>> BUG_ON(pg->count_info & PGC_broken);
>>>> - pg->count_info = PGC_state_free;
>>>> + pg->count_info = PGC_state_free | (pg->count_info & PGC_preserved);
>>>> break;
>>>
>>> PGC_extra doesn't want preserving here. Since it's mark_page_free(), and
>>> since PGC_extra is removed before freeing, the state change is apparently
>>> fine. But an assertion may want adding, for documentation purposes if
>>> nothing else.
>>>
>>> Alternatively it may make sense to indeed exclude PGC_extra here. In fact
>>> PGC_static doesn't need using here either, as unprepare_staticmem_pages()
>>> will explicitly set it again anyway.
>>>
>>> Hence I wonder whether the change here really is necessary (one will then
>>> be needed in the next patch aiui, when PGC_colored is introduced). Which
>>> would then eliminate the need for the final two hunks of the patch, I
>>> think.
>>>
>>>> case PGC_state_offlining:
>>>> - pg->count_info = (pg->count_info & PGC_broken) |
>>>> - PGC_state_offlined;
>>>> + pg->count_info = (pg->count_info & PGC_preserved) | PGC_state_offlined;
>>>> pg_offlined = true;
>>>> break;
>>>
>>> I'm similarly unconvinced that anything other than PGC_broken (and
>>> subsequently perhaps PGC_colored) would need preserving here.
>>
>> Yes, we (re)checked the code and also believe that the introduction of
>> PGC_preserved is generating more confusion (and code) then the actual logical
>> help it provides.
>>
>> We'll remove this patch and integrate PGC_colored explicitly in the flags to
>> be preserved. This avoid the clumsy logic of preserving something (extra)
>> when it's not needed and then handling the special case to remove it
>> explicitly.
>> Basically my goal is to go back to v4 where this patch didn't exist.
>
> Hmm, no, I don't think I said anything in the direction of removing PGC_preserved
> again. I merely think you went too far in where it actually wants using.
Let me try to better detail the intention of what we have in mind.
Here again parts of this patch:
> diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
> index 7b911b5ed9..34cd473150 100644
> --- a/xen/common/page_alloc.c
> +++ b/xen/common/page_alloc.c
> @@ -160,6 +160,7 @@
> #endif
>
> #define PGC_no_buddy_merge PGC_static
> +#define PGC_preserved (PGC_extra | PGC_static | PGC_broken)
The granularity of preserved as implemented by v10 is different than what you require...
>
> #ifndef PGT_TYPE_INFO_INITIALIZER
> #define PGT_TYPE_INFO_INITIALIZER 0
> @@ -1427,12 +1428,11 @@ static bool mark_page_free(struct page_info *pg, mfn_t mfn)
> {
> case PGC_state_inuse:
> BUG_ON(pg->count_info & PGC_broken);
> - pg->count_info = PGC_state_free;
> + pg->count_info = PGC_state_free | (pg->count_info & PGC_preserved);
here (no need),
> break;
>
> case PGC_state_offlining:
> - pg->count_info = (pg->count_info & PGC_broken) |
> - PGC_state_offlined;
> + pg->count_info = (pg->count_info & PGC_preserved) | PGC_state_offlined;
here (only broken),
> pg_offlined = true;
> break;
>
> @@ -2366,7 +2366,7 @@ int assign_pages(
>
> for ( i = 0; i < nr; i++ )
> {
> - ASSERT(!(pg[i].count_info & ~(PGC_extra | PGC_static)));
> + ASSERT(!(pg[i].count_info & ~PGC_preserved));
here (no broken),
> if ( pg[i].count_info & PGC_extra )
> extra_pages++;
> }
> @@ -2426,7 +2426,7 @@ int assign_pages(
> page_set_owner(&pg[i], d);
> smp_wmb(); /* Domain pointer must be visible before updating refcnt. */
> pg[i].count_info =
> - (pg[i].count_info & (PGC_extra | PGC_static)) | PGC_allocated | 1;
> + (pg[i].count_info & PGC_preserved) | PGC_allocated | 1;
here (no broken).
And it also leads (as noted) to the special management of extra...
>
> page_list_add_tail(&pg[i], page_to_list(d, &pg[i]));
> }
> @@ -2485,6 +2485,14 @@ struct page_info *alloc_domheap_pages(
> }
> if ( assign_page(pg, order, d, memflags) )
> {
> + if ( memflags & MEMF_no_refcount )
> + {
> + unsigned long i;
> +
> + for ( i = 0; i < (1UL << order); i++ )
> + pg[i].count_info &= ~PGC_extra;
> + }
here (likely not need, should be impossible to trigger),
> +
> free_heap_pages(pg, order, memflags & MEMF_no_scrub);
> return NULL;
> }
> @@ -2539,6 +2547,7 @@ void free_domheap_pages(struct page_info *pg, unsigned int order)
> {
> ASSERT(d->extra_pages);
> d->extra_pages--;
> + pg[i].count_info &= ~PGC_extra;
and here.
The proposal would be to go back to v9, which reduces (for the PGC)
the management to colored only:
> diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
...
> @@ -2382,7 +2556,7 @@ int assign_pages(
>
> for ( i = 0; i < nr; i++ )
> {
> - ASSERT(!(pg[i].count_info & ~(PGC_extra | PGC_static)));
> + ASSERT(!(pg[i].count_info & ~(PGC_extra | PGC_static | PGC_colored)));
> if ( pg[i].count_info & PGC_extra )
> extra_pages++;
> }
> @@ -2442,7 +2616,8 @@ int assign_pages(
> page_set_owner(&pg[i], d);
> smp_wmb(); /* Domain pointer must be visible before updating refcnt. */
> pg[i].count_info =
> - (pg[i].count_info & (PGC_extra | PGC_static)) | PGC_allocated | 1;
> + (pg[i].count_info & (PGC_extra | PGC_static | PGC_colored)) |
> + PGC_allocated | 1;
>
> page_list_add_tail(&pg[i], page_to_list(d, &pg[i]));
Thanks,
Andrea
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH v10 08/12] xen/page_alloc: introduce preserved page flags macro
2024-12-02 12:51 ` Andrea Bastoni
@ 2024-12-02 14:28 ` Jan Beulich
0 siblings, 0 replies; 36+ messages in thread
From: Jan Beulich @ 2024-12-02 14:28 UTC (permalink / raw)
To: Andrea Bastoni, Carlo Nonato
Cc: marco.solieri, Andrew Cooper, Julien Grall, Stefano Stabellini,
xen-devel
On 02.12.2024 13:51, Andrea Bastoni wrote:
> The proposal would be to go back to v9, which reduces (for the PGC)
> the management to colored only:
>
>> diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
> ...
>> @@ -2382,7 +2556,7 @@ int assign_pages(
>>
>> for ( i = 0; i < nr; i++ )
>> {
>> - ASSERT(!(pg[i].count_info & ~(PGC_extra | PGC_static)));
>> + ASSERT(!(pg[i].count_info & ~(PGC_extra | PGC_static | PGC_colored)));
>> if ( pg[i].count_info & PGC_extra )
>> extra_pages++;
>> }
>> @@ -2442,7 +2616,8 @@ int assign_pages(
>> page_set_owner(&pg[i], d);
>> smp_wmb(); /* Domain pointer must be visible before updating refcnt. */
>> pg[i].count_info =
>> - (pg[i].count_info & (PGC_extra | PGC_static)) | PGC_allocated | 1;
>> + (pg[i].count_info & (PGC_extra | PGC_static | PGC_colored)) |
>> + PGC_allocated | 1;
>>
>> page_list_add_tail(&pg[i], page_to_list(d, &pg[i]));
So what's wrong with having PGC_preserved _just_ for these two instances?
Jan
^ permalink raw reply [flat|nested] 36+ messages in thread
end of thread, other threads:[~2024-12-02 14:29 UTC | newest]
Thread overview: 36+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2024-11-19 14:13 [PATCH v10 00/12] Arm cache coloring Carlo Nonato
2024-11-19 14:13 ` [PATCH v10 01/12] xen/common: add cache coloring common code Carlo Nonato
2024-11-27 10:48 ` Michal Orzel
2024-11-27 13:24 ` Carlo Nonato
2024-11-27 13:57 ` Michal Orzel
2024-11-28 10:39 ` Jan Beulich
2024-11-19 14:13 ` [PATCH v10 02/12] xen/arm: add initial support for LLC coloring on arm64 Carlo Nonato
2024-11-27 14:32 ` Michal Orzel
2024-11-28 10:42 ` Jan Beulich
2024-11-19 14:13 ` [PATCH v10 03/12] xen/arm: permit non direct-mapped Dom0 construction Carlo Nonato
2024-11-28 10:34 ` Michal Orzel
2024-11-29 9:06 ` Carlo Nonato
2024-11-19 14:13 ` [PATCH v10 04/12] xen/arm: add Dom0 cache coloring support Carlo Nonato
2024-11-28 11:27 ` Michal Orzel
2024-11-19 14:13 ` [PATCH v10 05/12] xen: extend domctl interface for cache coloring Carlo Nonato
2024-11-19 14:13 ` [PATCH v10 06/12] tools: add support for cache coloring configuration Carlo Nonato
2024-11-20 16:25 ` Anthony PERARD
2024-11-21 11:48 ` Carlo Nonato
2024-11-20 16:48 ` Anthony PERARD
2024-11-19 14:13 ` [PATCH v10 07/12] xen/arm: add support for cache coloring configuration via device-tree Carlo Nonato
2024-11-28 12:00 ` Michal Orzel
2024-11-19 14:13 ` [PATCH v10 08/12] xen/page_alloc: introduce preserved page flags macro Carlo Nonato
2024-11-28 11:04 ` Jan Beulich
2024-11-29 9:32 ` Carlo Nonato
2024-11-29 11:09 ` Jan Beulich
2024-12-02 12:51 ` Andrea Bastoni
2024-12-02 14:28 ` Jan Beulich
2024-11-19 14:13 ` [PATCH v10 09/12] xen: add cache coloring allocator for domains Carlo Nonato
2024-11-28 11:43 ` Jan Beulich
2024-11-29 9:37 ` Carlo Nonato
2024-11-29 11:17 ` Jan Beulich
2024-11-19 14:13 ` [PATCH v10 10/12] xen/arm: add Xen cache colors command line parameter Carlo Nonato
2024-11-19 14:13 ` [PATCH v10 11/12] xen/arm: make consider_modules() available for xen relocation Carlo Nonato
2024-12-02 8:59 ` Michal Orzel
2024-11-19 14:13 ` [PATCH v10 12/12] xen/arm: add cache coloring support for Xen Carlo Nonato
2024-11-28 11:53 ` Jan Beulich
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.