Re: [PATCH v2 4/4] swiotlb: Free tbl memory in swiotlb_exit()

[Guenter Roeck <linux@roeck-us.net>]

On 7/31/21 7:29 PM, Konrad Rzeszutek Wilk wrote:
> On Sat, Jul 31, 2021 at 11:26:11AM -0700, Guenter Roeck wrote:
>> Hi,
>>
>> On Tue, Jul 20, 2021 at 02:38:26PM +0100, Will Deacon wrote:
>>> Although swiotlb_exit() frees the 'slots' metadata array referenced by
>>> 'io_tlb_default_mem', it leaves the underlying buffer pages allocated
>>> despite no longer being usable.
>>>
>>> Extend swiotlb_exit() to free the buffer pages as well as the slots
>>> array.
>>>
>>
>> This patch causes qemu pseries emulations to crash. Backtrace and bisect
>> log see below. Reverting it fixes the problem.
> 
> I am 99% sure it is fixed by this patch (which should be in linux-next
> in 5 minutes):
> 

Yes, it does.

FWIW:

Tested-by: Guenter Roeck <linux@roeck-us.net>

Thanks!
Guenter

> 
>>From a449ffaf9181b5a2dc705d8a06b13e0068207fd4 Mon Sep 17 00:00:00 2001
> From: Will Deacon <will@kernel.org>
> Date: Fri, 30 Jul 2021 12:42:31 +0100
> Subject: [PATCH] powerpc/svm: Don't issue ultracalls if !mem_encrypt_active()
> 
> Commit ad6c00283163 ("swiotlb: Free tbl memory in swiotlb_exit()")
> introduced a set_memory_encrypted() call to swiotlb_exit() so that the
> buffer pages are returned to an encrypted state prior to being freed.
> 
> Sachin reports that this leads to the following crash on a Power server:
> 
> [    0.010799] software IO TLB: tearing down default memory pool
> [    0.010805] ------------[ cut here ]------------
> [    0.010808] kernel BUG at arch/powerpc/kernel/interrupt.c:98!
> 
> Nick spotted that this is because set_memory_encrypted() is issuing an
> ultracall which doesn't exist for the processor, and should therefore
> be gated by mem_encrypt_active() to mirror the x86 implementation.
> 
> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> Cc: Claire Chang <tientzu@chromium.org>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: Robin Murphy <robin.murphy@arm.com>
> Fixes: ad6c00283163 ("swiotlb: Free tbl memory in swiotlb_exit()")
> Suggested-by: Nicholas Piggin <npiggin@gmail.com>
> Reported-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
> Tested-by: Sachin Sant <sachinp@linux.vnet.ibm.com>
> Tested-by: Nathan Chancellor <nathan@kernel.org>
> Link: https://lore.kernel.org/r/1905CD70-7656-42AE-99E2-A31FC3812EAC@linux.vnet.ibm.com/
> Signed-off-by: Will Deacon <will@kernel.org>
> Signed-off-by: Konrad Rzeszutek Wilk <konrad@kernel.org>
> ---
>   arch/powerpc/platforms/pseries/svm.c | 6 ++++++
>   1 file changed, 6 insertions(+)
> 
> diff --git a/arch/powerpc/platforms/pseries/svm.c b/arch/powerpc/platforms/pseries/svm.c
> index 1d829e257996..87f001b4c4e4 100644
> --- a/arch/powerpc/platforms/pseries/svm.c
> +++ b/arch/powerpc/platforms/pseries/svm.c
> @@ -63,6 +63,9 @@ void __init svm_swiotlb_init(void)
>   
>   int set_memory_encrypted(unsigned long addr, int numpages)
>   {
> +	if (!mem_encrypt_active())
> +		return 0;
> +
>   	if (!PAGE_ALIGNED(addr))
>   		return -EINVAL;
>   
> @@ -73,6 +76,9 @@ int set_memory_encrypted(unsigned long addr, int numpages)
>   
>   int set_memory_decrypted(unsigned long addr, int numpages)
>   {
> +	if (!mem_encrypt_active())
> +		return 0;
> +
>   	if (!PAGE_ALIGNED(addr))
>   		return -EINVAL;
>   
>