On Thu, 09 Jul 2020 13:49:21 PDT (-0700), rppt@xxxxxxxxxxxxx wrote:
Hi Palmer,
On Thu, Jul 09, 2020 at 01:05:48PM -0700, Palmer Dabbelt wrote:
From: Palmer Dabbelt <palmerdabbelt@xxxxxxxxxx>
As part of adding support for STRICT_DEVMEM to the RISC-V port, Zong
provided a devmem_is_allowed() implementation that's exactly the same as
all the others I checked. Instead I'm adding a generic version, which
will soon be used.
Signed-off-by: Palmer Dabbelt <palmerdabbelt@xxxxxxxxxx>
---
include/asm-generic/io.h | 4 ++++
lib/Kconfig | 4 ++++
lib/Makefile | 2 ++
lib/devmem_is_allowed.c | 27 +++++++++++++++++++++++++++
4 files changed, 37 insertions(+)
create mode 100644 lib/devmem_is_allowed.c
diff --git a/include/asm-generic/io.h b/include/asm-generic/io.h
index 8b1e020e9a03..69e3db65fba0 100644
--- a/include/asm-generic/io.h
+++ b/include/asm-generic/io.h
@@ -1122,6 +1122,10 @@ static inline void memcpy_toio(volatile void __iomem *addr, const void *buffer,
}
#endif
+#ifndef CONFIG_GENERIC_DEVMEM_IS_ALLOWED
+extern int devmem_is_allowed(unsigned long pfn);
+#endif
+
#endif /* __KERNEL__ */
#endif /* __ASM_GENERIC_IO_H */
diff --git a/lib/Kconfig b/lib/Kconfig
index df3f3da95990..3b1b6481e073 100644
--- a/lib/Kconfig
+++ b/lib/Kconfig
@@ -676,3 +676,7 @@ config GENERIC_LIB_CMPDI2
config GENERIC_LIB_UCMPDI2
bool
+
+config GENERIC_LIB_DEVMEM_IS_ALLOWED
+ bool
+ select ARCH_HAS_DEVMEM_IS_ALLOWED
This seems to work the other way around from the usual Kconfig chains.
In the most cases ARCH_HAS_SOMETHING selects GENERIC_SOMETHING.
Ya, it seemed kind of odd.
I believe nicer way would be to make
config STRICT_DEVMEM
bool "Filter access to /dev/mem"
depends on MMU && DEVMEM
depends on ARCH_HAS_DEVMEM_IS_ALLOWED || GENERIC_LIB_DEVMEM_IS_ALLOWED
config GENERIC_LIB_DEVMEM_IS_ALLOWED
bool
and then s/select ARCH_HAS_DEVMEM_IS_ALLOWED/select GENERIC_LIB_DEVMEM_IS_ALLOWED/
in the arch Kconfigs and drop ARCH_HAS_DEVMEM_IS_ALLOWED in the end.
There's some arches that can't be converted to the generic version, at least
not trivially, so we wouldn't drop it. I think it's still cleaner, though.
I'll send a v2.
diff --git a/lib/Makefile b/lib/Makefile
index b1c42c10073b..554ef14f9be5 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -318,3 +318,5 @@ obj-$(CONFIG_OBJAGG) += objagg.o
# KUnit tests
obj-$(CONFIG_LIST_KUNIT_TEST) += list-test.o
obj-$(CONFIG_LINEAR_RANGES_TEST) += test_linear_ranges.o
+
+obj-$(CONFIG_GENERIC_LIB_DEVMEM_IS_ALLOWED) += devmem_is_allowed.o
diff --git a/lib/devmem_is_allowed.c b/lib/devmem_is_allowed.c
new file mode 100644
index 000000000000..c0d67c541849
--- /dev/null
+++ b/lib/devmem_is_allowed.c
@@ -0,0 +1,27 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * A generic version of devmem_is_allowed.
+ *
+ * Based on arch/arm64/mm/mmap.c
+ *
+ * Copyright (C) 2020 Google, Inc.
+ * Copyright (C) 2012 ARM Ltd.
+ */
+
+#include <linux/mm.h>
+#include <linux/ioport.h>
+
+/*
+ * devmem_is_allowed() checks to see if /dev/mem access to a certain address
+ * is valid. The argument is a physical page number. We mimic x86 here by
+ * disallowing access to system RAM as well as device-exclusive MMIO regions.
+ * This effectively disable read()/write() on /dev/mem.
+ */
+int devmem_is_allowed(unsigned long pfn)
+{
+ if (iomem_is_exclusive(pfn << PAGE_SHIFT))
+ return 0;
+ if (!page_is_ram(pfn))
+ return 1;
+ return 0;
+}
--
2.27.0.383.g050319c2ae-goog
