[PATCH 1/6] powerpc: Fix virt_addr_valid() for 64-bit Book3E & 32-bit

[PATCH 1/6] powerpc: Fix virt_addr_valid() for 64-bit Book3E & 32-bit

[Michael Ellerman]

From: Kefeng Wang <wangkefeng.wang@huawei.com>

mpe: On 64-bit Book3E vmalloc space starts at 0x8000000000000000.

Because of the way __pa() works we have:
  __pa(0x8000000000000000) == 0, and therefore
  virt_to_pfn(0x8000000000000000) == 0, and therefore
  virt_addr_valid(0x8000000000000000) == true

Which is wrong, virt_addr_valid() should be false for vmalloc space.
In fact all vmalloc addresses that alias with a valid PFN will return
true from virt_addr_valid(). That can cause bugs with hardened usercopy
as described below by Kefeng Wang:

  When running ethtool eth0 on 64-bit Book3E, a BUG occurred:

    usercopy: Kernel memory exposure attempt detected from SLUB object not in SLUB page?! (offset 0, size 1048)!
    kernel BUG at mm/usercopy.c:99
    ...
    usercopy_abort+0x64/0xa0 (unreliable)
    __check_heap_object+0x168/0x190
    __check_object_size+0x1a0/0x200
    dev_ethtool+0x2494/0x2b20
    dev_ioctl+0x5d0/0x770
    sock_do_ioctl+0xf0/0x1d0
    sock_ioctl+0x3ec/0x5a0
    __se_sys_ioctl+0xf0/0x160
    system_call_exception+0xfc/0x1f0
    system_call_common+0xf8/0x200

  The code shows below,

    data = vzalloc(array_size(gstrings.len, ETH_GSTRING_LEN));
    copy_to_user(useraddr, data, gstrings.len * ETH_GSTRING_LEN))

  The data is alloced by vmalloc(), virt_addr_valid(ptr) will return true
  on 64-bit Book3E, which leads to the panic.

  As commit 4dd7554a6456 ("powerpc/64: Add VIRTUAL_BUG_ON checks for __va
  and __pa addresses") does, make sure the virt addr above PAGE_OFFSET in
  the virt_addr_valid() for 64-bit, also add upper limit check to make
  sure the virt is below high_memory.

  Meanwhile, for 32-bit PAGE_OFFSET is the virtual address of the start
  of lowmem, high_memory is the upper low virtual address, the check is
  suitable for 32-bit, this will fix the issue mentioned in commit
  602946ec2f90 ("powerpc: Set max_mapnr correctly") too.

On 32-bit there is a similar problem with high memory, that was fixed in
commit 602946ec2f90 ("powerpc: Set max_mapnr correctly"), but that
commit breaks highmem and needs to be reverted.

We can't easily fix __pa(), we have code that relies on its current
behaviour. So for now add extra checks to virt_addr_valid().

For 64-bit Book3S the extra checks are not necessary, the combination of
virt_to_pfn() and pfn_valid() should yield the correct result, but they
are harmless.

Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu>
[mpe: Add additional change log detail]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
---
 arch/powerpc/include/asm/page.h | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/include/asm/page.h b/arch/powerpc/include/asm/page.h
index 254687258f42..f2c5c26869f1 100644
--- a/arch/powerpc/include/asm/page.h
+++ b/arch/powerpc/include/asm/page.h
@@ -132,7 +132,11 @@ static inline bool pfn_valid(unsigned long pfn)
 #define virt_to_page(kaddr)	pfn_to_page(virt_to_pfn(kaddr))
 #define pfn_to_kaddr(pfn)	__va((pfn) << PAGE_SHIFT)
 
-#define virt_addr_valid(kaddr)	pfn_valid(virt_to_pfn(kaddr))
+#define virt_addr_valid(vaddr)	({					\
+	unsigned long _addr = (unsigned long)vaddr;			\
+	_addr >= PAGE_OFFSET && _addr < (unsigned long)high_memory &&	\
+	pfn_valid(virt_to_pfn(_addr));					\
+})
 
 /*
  * On Book-E parts we need __va to parse the device tree and we can't
-- 
2.34.1

[PATCH 2/6] Revert "powerpc: Set max_mapnr correctly"

[Michael Ellerman]

From: Kefeng Wang <wangkefeng.wang@huawei.com>

This reverts commit 602946ec2f90d5bd965857753880db29d2d9a1e9.

If CONFIG_HIGHMEM is enabled, no highmem will be added with max_mapnr
set to max_low_pfn, see mem_init():

  for (pfn = highmem_mapnr; pfn < max_mapnr; ++pfn) {
        ...
        free_highmem_page();
  }

Now that virt_addr_valid() has been fixed in the previous commit, we can
revert the change to max_mapnr.

Fixes: 602946ec2f90 ("powerpc: Set max_mapnr correctly")
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Reported-by: Erhard F. <erhard_f@mailbox.org>
[mpe: Update change log to reflect series reordering]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
---
 arch/powerpc/mm/mem.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/powerpc/mm/mem.c b/arch/powerpc/mm/mem.c
index 8e301cd8925b..4d221d033804 100644
--- a/arch/powerpc/mm/mem.c
+++ b/arch/powerpc/mm/mem.c
@@ -255,7 +255,7 @@ void __init mem_init(void)
 #endif
 
 	high_memory = (void *) __va(max_low_pfn * PAGE_SIZE);
-	set_max_mapnr(max_low_pfn);
+	set_max_mapnr(max_pfn);
 
 	kasan_late_init();
 
-- 
2.34.1

[PATCH 3/6] powerpc/85xx: Fix virt_to_phys() off-by-one in smp_85xx_start_cpu()

[Michael Ellerman]

In smp_85xx_start_cpu() we are passed an address but we're unsure if
it's a real or virtual address, so there's a check to determine that.

The check has an off-by-one in that it tests if the address is greater
than high_memory, but high_memory is the first address of high memory,
so the check should be greater-or-equal.

It seems this has never been a problem in practice, but it also triggers
the DEBUG_VIRTUAL checks in __pa() which we would like to avoid. We can
fix both issues by converting high_memory - 1 to a physical address and
testing against that.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
---
 arch/powerpc/platforms/85xx/smp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/powerpc/platforms/85xx/smp.c b/arch/powerpc/platforms/85xx/smp.c
index a1c6a7827c8f..9c43cf32f4c9 100644
--- a/arch/powerpc/platforms/85xx/smp.c
+++ b/arch/powerpc/platforms/85xx/smp.c
@@ -208,7 +208,7 @@ static int smp_85xx_start_cpu(int cpu)
 	 * The bootpage and highmem can be accessed via ioremap(), but
 	 * we need to directly access the spinloop if its in lowmem.
 	 */
-	ioremappable = *cpu_rel_addr > virt_to_phys(high_memory);
+	ioremappable = *cpu_rel_addr > virt_to_phys(high_memory - 1);
 
 	/* Map the spin table */
 	if (ioremappable)
-- 
2.34.1

[PATCH 4/6] powerpc/vas: Fix __pa() handling in init_winctx_regs()

[Michael Ellerman]

In init_winctx_regs() we call __pa() on winctx->rx_fifo, but some
callers pass a real adress which causes errors with DEBUG_VIRTUAL
enabled.

So check first if we have a virtual address, and otherwise leave the
address unchanged.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
---
 arch/powerpc/platforms/powernv/vas-window.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/platforms/powernv/vas-window.c b/arch/powerpc/platforms/powernv/vas-window.c
index 0f8d39fbf2b2..25324390e292 100644
--- a/arch/powerpc/platforms/powernv/vas-window.c
+++ b/arch/powerpc/platforms/powernv/vas-window.c
@@ -404,7 +404,13 @@ static void init_winctx_regs(struct pnv_vas_window *window,
 	 *
 	 * See also: Design note in function header.
 	 */
-	val = __pa(winctx->rx_fifo);
+
+	// Some callers pass virtual addresses, others pass real
+	if (virt_addr_valid(winctx->rx_fifo))
+		val = virt_to_phys(winctx->rx_fifo);
+	else
+		val = (u64)winctx->rx_fifo;
+
 	val = SET_FIELD(VAS_PAGE_MIGRATION_SELECT, val, 0);
 	write_hvwc_reg(window, VREG(LFIFO_BAR), val);
 
-- 
2.34.1

[PATCH 5/6] powerpc/64: Only WARN if __pa()/__va() called with bad addresses

[Michael Ellerman]

We added checks to __pa() / __va() to ensure they're only called with
appropriate addresses. But using BUG_ON() is too strong, it means
virt_addr_valid() will BUG when DEBUG_VIRTUAL is enabled.

Instead switch them to warnings, arm64 does the same.

Fixes: 4dd7554a6456 ("powerpc/64: Add VIRTUAL_BUG_ON checks for __va and __pa addresses")
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
---
 arch/powerpc/include/asm/page.h | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/include/asm/page.h b/arch/powerpc/include/asm/page.h
index f2c5c26869f1..40a27a56ee40 100644
--- a/arch/powerpc/include/asm/page.h
+++ b/arch/powerpc/include/asm/page.h
@@ -216,6 +216,12 @@ static inline bool pfn_valid(unsigned long pfn)
 #define __pa(x) ((phys_addr_t)(unsigned long)(x) - VIRT_PHYS_OFFSET)
 #else
 #ifdef CONFIG_PPC64
+
+#ifdef CONFIG_DEBUG_VIRTUAL
+#define VIRTUAL_WARN_ON(x)	WARN_ON(x)
+#else
+#define VIRTUAL_WARN_ON(x)
+#endif
 /*
  * gcc miscompiles (unsigned long)(&static_var) - PAGE_OFFSET
  * with -mcmodel=medium, so we use & and | instead of - and + on 64-bit.
@@ -223,13 +229,13 @@ static inline bool pfn_valid(unsigned long pfn)
  */
 #define __va(x)								\
 ({									\
-	VIRTUAL_BUG_ON((unsigned long)(x) >= PAGE_OFFSET);		\
+	VIRTUAL_WARN_ON((unsigned long)(x) >= PAGE_OFFSET);		\
 	(void *)(unsigned long)((phys_addr_t)(x) | PAGE_OFFSET);	\
 })
 
 #define __pa(x)								\
 ({									\
-	VIRTUAL_BUG_ON((unsigned long)(x) < PAGE_OFFSET);		\
+	VIRTUAL_WARN_ON((unsigned long)(x) < PAGE_OFFSET);		\
 	(unsigned long)(x) & 0x0fffffffffffffffUL;			\
 })
 
-- 
2.34.1

[RFC PATCH 6/6] powerpc/mm: Add virt_addr_valid() checks

[Michael Ellerman]

We've had several bugs now with virt_addr_valid() being wrong, so lets
add some always-enabled boot time checks that it behaves as expected.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
---
 arch/powerpc/mm/mem.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/arch/powerpc/mm/mem.c b/arch/powerpc/mm/mem.c
index 4d221d033804..81e9d948a8e8 100644
--- a/arch/powerpc/mm/mem.c
+++ b/arch/powerpc/mm/mem.c
@@ -305,6 +305,13 @@ void __init mem_init(void)
 		MODULES_VADDR, MODULES_END);
 #endif
 #endif /* CONFIG_PPC32 */
+
+	// Check virt_addr_valid() works as expected
+	WARN_ON(!virt_addr_valid(PAGE_OFFSET));
+	WARN_ON(virt_addr_valid(PAGE_OFFSET - 1));
+	WARN_ON(virt_addr_valid(high_memory));
+	WARN_ON(virt_addr_valid(VMALLOC_START));
+	WARN_ON(virt_addr_valid(VMALLOC_END - 1));
 }
 
 void free_initmem(void)
-- 
2.34.1

Re: [PATCH 5/6] powerpc/64: Only WARN if __pa()/__va() called with bad addresses

[Christophe Leroy]

Le 06/04/2022 à 16:58, Michael Ellerman a écrit :
> We added checks to __pa() / __va() to ensure they're only called with
> appropriate addresses. But using BUG_ON() is too strong, it means
> virt_addr_valid() will BUG when DEBUG_VIRTUAL is enabled.
> 
> Instead switch them to warnings, arm64 does the same.
> 
> Fixes: 4dd7554a6456 ("powerpc/64: Add VIRTUAL_BUG_ON checks for __va and __pa addresses")
> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
> ---
>   arch/powerpc/include/asm/page.h | 10 ++++++++--
>   1 file changed, 8 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/powerpc/include/asm/page.h b/arch/powerpc/include/asm/page.h
> index f2c5c26869f1..40a27a56ee40 100644
> --- a/arch/powerpc/include/asm/page.h
> +++ b/arch/powerpc/include/asm/page.h
> @@ -216,6 +216,12 @@ static inline bool pfn_valid(unsigned long pfn)
>   #define __pa(x) ((phys_addr_t)(unsigned long)(x) - VIRT_PHYS_OFFSET)
>   #else
>   #ifdef CONFIG_PPC64
> +
> +#ifdef CONFIG_DEBUG_VIRTUAL
> +#define VIRTUAL_WARN_ON(x)	WARN_ON(x)
> +#else
> +#define VIRTUAL_WARN_ON(x)
> +#endif

Could be:

#define VIRTUAL_WARN_ON(x)	WARN_ON(IS_ENABLED(CONFIG_DEBUG_VIRTUAL) && (x))

>   /*
>    * gcc miscompiles (unsigned long)(&static_var) - PAGE_OFFSET
>    * with -mcmodel=medium, so we use & and | instead of - and + on 64-bit.
> @@ -223,13 +229,13 @@ static inline bool pfn_valid(unsigned long pfn)
>    */
>   #define __va(x)								\
>   ({									\
> -	VIRTUAL_BUG_ON((unsigned long)(x) >= PAGE_OFFSET);		\
> +	VIRTUAL_WARN_ON((unsigned long)(x) >= PAGE_OFFSET);		\
>   	(void *)(unsigned long)((phys_addr_t)(x) | PAGE_OFFSET);	\
>   })
>   
>   #define __pa(x)								\
>   ({									\
> -	VIRTUAL_BUG_ON((unsigned long)(x) < PAGE_OFFSET);		\
> +	VIRTUAL_WARN_ON((unsigned long)(x) < PAGE_OFFSET);		\
>   	(unsigned long)(x) & 0x0fffffffffffffffUL;			\
>   })
>   

Isn't it dangerous to WARN (or BUG) here ? __pa() can be used very early 
during boot, like in prom_init.c

Some other architectures have a __pa_nodebug(). The __pa() does the 
WARN() then calls __pa_nodebug(). Early users call __pa_nodebug() directly.

Christophe

Re: [PATCH 5/6] powerpc/64: Only WARN if __pa()/__va() called with bad addresses

[Michael Ellerman]

Christophe Leroy <christophe.leroy@csgroup.eu> writes:
> Le 06/04/2022 à 16:58, Michael Ellerman a écrit :
>> We added checks to __pa() / __va() to ensure they're only called with
>> appropriate addresses. But using BUG_ON() is too strong, it means
>> virt_addr_valid() will BUG when DEBUG_VIRTUAL is enabled.
>> 
>> Instead switch them to warnings, arm64 does the same.
>> 
>> Fixes: 4dd7554a6456 ("powerpc/64: Add VIRTUAL_BUG_ON checks for __va and __pa addresses")
>> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
>> ---
>>   arch/powerpc/include/asm/page.h | 10 ++++++++--
>>   1 file changed, 8 insertions(+), 2 deletions(-)
>> 
>> diff --git a/arch/powerpc/include/asm/page.h b/arch/powerpc/include/asm/page.h
>> index f2c5c26869f1..40a27a56ee40 100644
>> --- a/arch/powerpc/include/asm/page.h
>> +++ b/arch/powerpc/include/asm/page.h
>> @@ -216,6 +216,12 @@ static inline bool pfn_valid(unsigned long pfn)
>>   #define __pa(x) ((phys_addr_t)(unsigned long)(x) - VIRT_PHYS_OFFSET)
>>   #else
>>   #ifdef CONFIG_PPC64
>> +
>> +#ifdef CONFIG_DEBUG_VIRTUAL
>> +#define VIRTUAL_WARN_ON(x)	WARN_ON(x)
>> +#else
>> +#define VIRTUAL_WARN_ON(x)
>> +#endif
>
> Could be:
>
> #define VIRTUAL_WARN_ON(x)	WARN_ON(IS_ENABLED(CONFIG_DEBUG_VIRTUAL) && (x))
>
>>   /*
>>    * gcc miscompiles (unsigned long)(&static_var) - PAGE_OFFSET
>>    * with -mcmodel=medium, so we use & and | instead of - and + on 64-bit.
>> @@ -223,13 +229,13 @@ static inline bool pfn_valid(unsigned long pfn)
>>    */
>>   #define __va(x)								\
>>   ({									\
>> -	VIRTUAL_BUG_ON((unsigned long)(x) >= PAGE_OFFSET);		\
>> +	VIRTUAL_WARN_ON((unsigned long)(x) >= PAGE_OFFSET);		\
>>   	(void *)(unsigned long)((phys_addr_t)(x) | PAGE_OFFSET);	\
>>   })
>>   
>>   #define __pa(x)								\
>>   ({									\
>> -	VIRTUAL_BUG_ON((unsigned long)(x) < PAGE_OFFSET);		\
>> +	VIRTUAL_WARN_ON((unsigned long)(x) < PAGE_OFFSET);		\
>>   	(unsigned long)(x) & 0x0fffffffffffffffUL;			\
>>   })
>>   
>
> Isn't it dangerous to WARN (or BUG) here ? __pa() can be used very early 
> during boot, like in prom_init.c

Yes. WARN is a bit less dangerous though :)

> Some other architectures have a __pa_nodebug(). The __pa() does the 
> WARN() then calls __pa_nodebug(). Early users call __pa_nodebug() directly.

Yeah I saw that, we could go that way.

I think possibly the better option is for __pa() to have no checks,
instead the checks go in the higher level routines like virt_to_phys()
and phys_to_virt().

And then we can check uses of __pa() and any that are *not* early boot
or low level stuff can be converted to virt_to_phys().

cheers

Re: [PATCH 1/6] powerpc: Fix virt_addr_valid() for 64-bit Book3E & 32-bit

[Michael Ellerman]

On Thu, 7 Apr 2022 00:57:57 +1000, Michael Ellerman wrote:
> From: Kefeng Wang <wangkefeng.wang@huawei.com>
> 
> mpe: On 64-bit Book3E vmalloc space starts at 0x8000000000000000.
> 
> Because of the way __pa() works we have:
>   __pa(0x8000000000000000) == 0, and therefore
>   virt_to_pfn(0x8000000000000000) == 0, and therefore
>   virt_addr_valid(0x8000000000000000) == true
> 
> [...]

Patches 1 & 2 applied to powerpc/fixes.

[1/6] powerpc: Fix virt_addr_valid() for 64-bit Book3E & 32-bit
      https://git.kernel.org/powerpc/c/ffa0b64e3be58519ae472ea29a1a1ad681e32f48
[2/6] Revert "powerpc: Set max_mapnr correctly"
      https://git.kernel.org/powerpc/c/1ff5c8e8c835e8a81c0868e3050c76563dd56a2c

cheers

Re: [PATCH 3/6] powerpc/85xx: Fix virt_to_phys() off-by-one in smp_85xx_start_cpu()

[Michael Ellerman]

On Thu, 7 Apr 2022 00:57:59 +1000, Michael Ellerman wrote:
> In smp_85xx_start_cpu() we are passed an address but we're unsure if
> it's a real or virtual address, so there's a check to determine that.
> 
> [...]

Applied to powerpc/next.

[3/6] powerpc/85xx: Fix virt_to_phys() off-by-one in smp_85xx_start_cpu()
      https://git.kernel.org/powerpc/c/0d897255e79e26f471d10bbf72db9eee6f9cb723

cheers