Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH 3/3] powerpc/mm: call H_BLOCK_REMOVE when supported
To:     "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>,
        mpe@ellerman.id.au, benh@kernel.crashing.org, paulus@samba.org,
        npiggin@gmail.com
Cc:     linuxppc-dev@lists.ozlabs.org, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org
References: <20190830120712.22971-1-ldufour@linux.ibm.com>
 <20190830120712.22971-4-ldufour@linux.ibm.com>
 <5bcd3da7-1bc2-f3f9-3ed2-e3aa0bb540bd@linux.ibm.com>
From:   Laurent Dufour <ldufour@linux.ibm.com>
Date:   Fri, 13 Sep 2019 15:16:13 +0200
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.14; rv:60.0)
 Gecko/20100101 Thunderbird/60.9.0
MIME-Version: 1.0
In-Reply-To: <5bcd3da7-1bc2-f3f9-3ed2-e3aa0bb540bd@linux.ibm.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Message-Id: <cf06842a-ea37-9684-5b28-de4277e7bef3@linux.ibm.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Le 12/09/2019 à 16:20, Aneesh Kumar K.V a écrit :
> On 8/30/19 5:37 PM, Laurent Dufour wrote:
>> Instead of calling H_BLOCK_REMOVE all the time when the feature is
>> exhibited, call this hcall only when the couple base page size, page size
>> is supported as reported by the TLB Invalidate Characteristics.
>>
> 
> supported is not actually what we are checking here. We are checking 
> whether the base page size actual page size remove can be done in chunks of 
> 8 blocks. If we don't support 8 block you fallback to bulk invalidate. May 
> be update the commit message accordingly?

Yes that's correct.

I think I should also put the warning message displayed when reading the 
characteristic in that commit and explicitly mentioned that we only support 
8 entries size block for this hcall.
This way the limitation is limited to this commit.

> 
>> For regular pages and hugetlb, the assumption is made that the page size is
>> equal to the base page size. For THP the page size is assumed to be 16M.
>>
>> Signed-off-by: Laurent Dufour <ldufour@linux.ibm.com>
>> ---
>>   arch/powerpc/platforms/pseries/lpar.c | 11 +++++++++--
>>   1 file changed, 9 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/powerpc/platforms/pseries/lpar.c 
>> b/arch/powerpc/platforms/pseries/lpar.c
>> index 375e19b3cf53..ef3dbf108a65 100644
>> --- a/arch/powerpc/platforms/pseries/lpar.c
>> +++ b/arch/powerpc/platforms/pseries/lpar.c
>> @@ -1143,7 +1143,11 @@ static inline void 
>> __pSeries_lpar_hugepage_invalidate(unsigned long *slot,
>>       if (lock_tlbie)
>>           spin_lock_irqsave(&pSeries_lpar_tlbie_lock, flags);
>> -    if (firmware_has_feature(FW_FEATURE_BLOCK_REMOVE))
>> +    /*
>> +     * Assuming THP size is 16M, and we only support 8 bytes size buffer
>> +     * for the momment.
>> +     */
>> +    if (mmu_psize_defs[psize].hblk[MMU_PAGE_16M] == 8)
>>           hugepage_block_invalidate(slot, vpn, count, psize, ssize);
>>       else
>>           hugepage_bulk_invalidate(slot, vpn, count, psize, ssize);
> 
> 
> 
> So we don't use block invalidate if blocksize is != 8.

yes

> 
> 
>> @@ -1437,7 +1441,10 @@ static void pSeries_lpar_flush_hash_range(unsigned 
>> long number, int local)
>>       if (lock_tlbie)
>>           spin_lock_irqsave(&pSeries_lpar_tlbie_lock, flags);
>> -    if (firmware_has_feature(FW_FEATURE_BLOCK_REMOVE)) {
>> +    /*
>> +     * Currently, we only support 8 bytes size buffer in do_block_remove().
>> +     */
>> +    if (mmu_psize_defs[batch->psize].hblk[batch->psize] == 8) {
>>           do_block_remove(number, batch, param);
>>           goto out;
>>       }
>>
> 
> -aneesh