| Message ID | 1497883044-1276-1-git-send-email-ard.biesheuvel@linaro.org |
|---|---|
| State | Superseded |
| Headers | show |
On Mon, Jun 19, 2017 at 04:37:24PM +0200, Ard Biesheuvel wrote: > On arm64, the /dev/kmem driver barely works, given that it assumes that > VMALLOC_START > PAGE_OFFSET, which is not the case on arm64. Due to the Probably worth being explicit about the failure mode here (the commit message you had in your RFC was much better ;). Will > dubious value of kmem on an architecture such as arm64, which is fairly > new and is not as forgiving as other architectures when it comes to > userland accesses to kernel memory using mismatched attributes*, let's > disable it altogether if CONFIG_ARM64=y. > > * Mappings in the vmalloc region will be accessed using vread/vwrite when > reading/writing to/from /dev/kmem, which will perform these accesses by > kmap()'ing each page using writeback cacheable attributes. This will be > correct in most cases, but this is not guaranteed. > > Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> > --- > drivers/char/Kconfig | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/drivers/char/Kconfig b/drivers/char/Kconfig > index 31adbebf812e..8102ee7b3247 100644 > --- a/drivers/char/Kconfig > +++ b/drivers/char/Kconfig > @@ -17,6 +17,8 @@ config DEVMEM > > config DEVKMEM > bool "/dev/kmem virtual device support" > + # On arm64, VMALLOC_START < PAGE_OFFSET, which confuses kmem read/write > + depends on !ARM64 > help > Say Y here if you want to support the /dev/kmem device. The > /dev/kmem device is rarely used, but can be used for certain > -- > 2.7.4 >
On 19 June 2017 at 17:03, Will Deacon <will.deacon@arm.com> wrote: > On Mon, Jun 19, 2017 at 04:37:24PM +0200, Ard Biesheuvel wrote: >> On arm64, the /dev/kmem driver barely works, given that it assumes that >> VMALLOC_START > PAGE_OFFSET, which is not the case on arm64. Due to the > > Probably worth being explicit about the failure mode here (the commit > message you had in your RFC was much better ;). > Yeah, that's a fair point, I was being lazy :-) I will repost with the v1 commit log integrated >> dubious value of kmem on an architecture such as arm64, which is fairly >> new and is not as forgiving as other architectures when it comes to >> userland accesses to kernel memory using mismatched attributes*, let's >> disable it altogether if CONFIG_ARM64=y. >> >> * Mappings in the vmalloc region will be accessed using vread/vwrite when >> reading/writing to/from /dev/kmem, which will perform these accesses by >> kmap()'ing each page using writeback cacheable attributes. This will be >> correct in most cases, but this is not guaranteed. >> >> Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> >> --- >> drivers/char/Kconfig | 2 ++ >> 1 file changed, 2 insertions(+) >> >> diff --git a/drivers/char/Kconfig b/drivers/char/Kconfig >> index 31adbebf812e..8102ee7b3247 100644 >> --- a/drivers/char/Kconfig >> +++ b/drivers/char/Kconfig >> @@ -17,6 +17,8 @@ config DEVMEM >> >> config DEVKMEM >> bool "/dev/kmem virtual device support" >> + # On arm64, VMALLOC_START < PAGE_OFFSET, which confuses kmem read/write >> + depends on !ARM64 >> help >> Say Y here if you want to support the /dev/kmem device. The >> /dev/kmem device is rarely used, but can be used for certain >> -- >> 2.7.4 >>
diff --git a/drivers/char/Kconfig b/drivers/char/Kconfig index 31adbebf812e..8102ee7b3247 100644 --- a/drivers/char/Kconfig +++ b/drivers/char/Kconfig @@ -17,6 +17,8 @@ config DEVMEM config DEVKMEM bool "/dev/kmem virtual device support" + # On arm64, VMALLOC_START < PAGE_OFFSET, which confuses kmem read/write + depends on !ARM64 help Say Y here if you want to support the /dev/kmem device. The /dev/kmem device is rarely used, but can be used for certain
On arm64, the /dev/kmem driver barely works, given that it assumes that VMALLOC_START > PAGE_OFFSET, which is not the case on arm64. Due to the dubious value of kmem on an architecture such as arm64, which is fairly new and is not as forgiving as other architectures when it comes to userland accesses to kernel memory using mismatched attributes*, let's disable it altogether if CONFIG_ARM64=y. * Mappings in the vmalloc region will be accessed using vread/vwrite when reading/writing to/from /dev/kmem, which will perform these accesses by kmap()'ing each page using writeback cacheable attributes. This will be correct in most cases, but this is not guaranteed. Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org> --- drivers/char/Kconfig | 2 ++ 1 file changed, 2 insertions(+) -- 2.7.4