Message-ID: <10be86c3-96ac-399f-850d-2ceaace22e24@arm.com>
Date:   Fri, 3 Mar 2023 14:04:56 +0000
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.7.1
Subject: Re: [RFC PATCH 09/28] arm64: RME: RTT handling
To:     Zhi Wang <zhi.wang.linux@gmail.com>
Cc:     kvm@vger.kernel.org, kvmarm@lists.linux.dev,
        Catalin Marinas <catalin.marinas@arm.com>,
        Marc Zyngier <maz@kernel.org>, Will Deacon <will@kernel.org>,
        James Morse <james.morse@arm.com>,
        Oliver Upton <oliver.upton@linux.dev>,
        Suzuki K Poulose <suzuki.poulose@arm.com>,
        Zenghui Yu <yuzenghui@huawei.com>,
        linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org,
        Joey Gouly <joey.gouly@arm.com>,
        Alexandru Elisei <alexandru.elisei@arm.com>,
        Christoffer Dall <christoffer.dall@arm.com>,
        Fuad Tabba <tabba@google.com>, linux-coco@lists.linux.dev
References: <20230127112248.136810-1-suzuki.poulose@arm.com>
 <20230127112932.38045-1-steven.price@arm.com>
 <20230127112932.38045-10-steven.price@arm.com>
 <20230213194421.00005f7b@gmail.com>
Content-Language: en-GB
From:   Steven Price <steven.price@arm.com>
In-Reply-To: <20230213194421.00005f7b@gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 13/02/2023 17:44, Zhi Wang wrote:
> On Fri, 27 Jan 2023 11:29:13 +0000
> Steven Price <steven.price@arm.com> wrote:
> 
>> The RMM owns the stage 2 page tables for a realm, and KVM must request
>> that the RMM creates/destroys entries as necessary. The physical pages
>> to store the page tables are delegated to the realm as required, and can
>> be undelegated when no longer used.
>>
> 
> This is only an introduction to RTT handling. While this patch is mostly like
> RTT teardown, better add more introduction to this patch. Also maybe refine
> the tittle to reflect what this patch is actually doing.

You've a definite point that this patch is mostly about RTT teardown.
Technically it also adds the RTT creation path (realm_rtt_create) -
hence the generic patch title.

But I'll definitely expand the commit message to mention the complexity
of tear down which is the bulk of the patch.

>> Signed-off-by: Steven Price <steven.price@arm.com>
>> ---
>>  arch/arm64/include/asm/kvm_rme.h |  19 +++++
>>  arch/arm64/kvm/mmu.c             |   7 +-
>>  arch/arm64/kvm/rme.c             | 139 +++++++++++++++++++++++++++++++
>>  3 files changed, 162 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/arm64/include/asm/kvm_rme.h b/arch/arm64/include/asm/kvm_rme.h
>> index a6318af3ed11..eea5118dfa8a 100644
>> --- a/arch/arm64/include/asm/kvm_rme.h
>> +++ b/arch/arm64/include/asm/kvm_rme.h
>> @@ -35,5 +35,24 @@ u32 kvm_realm_ipa_limit(void);
>>  int kvm_realm_enable_cap(struct kvm *kvm, struct kvm_enable_cap *cap);
>>  int kvm_init_realm_vm(struct kvm *kvm);
>>  void kvm_destroy_realm(struct kvm *kvm);
>> +void kvm_realm_destroy_rtts(struct realm *realm, u32 ia_bits, u32 start_level);
>> +
>> +#define RME_RTT_BLOCK_LEVEL	2
>> +#define RME_RTT_MAX_LEVEL	3
>> +
>> +#define RME_PAGE_SHIFT		12
>> +#define RME_PAGE_SIZE		BIT(RME_PAGE_SHIFT)
>> +/* See ARM64_HW_PGTABLE_LEVEL_SHIFT() */
>> +#define RME_RTT_LEVEL_SHIFT(l)	\
>> +	((RME_PAGE_SHIFT - 3) * (4 - (l)) + 3)
>> +#define RME_L2_BLOCK_SIZE	BIT(RME_RTT_LEVEL_SHIFT(2))
>> +
>> +static inline unsigned long rme_rtt_level_mapsize(int level)
>> +{
>> +	if (WARN_ON(level > RME_RTT_MAX_LEVEL))
>> +		return RME_PAGE_SIZE;
>> +
>> +	return (1UL << RME_RTT_LEVEL_SHIFT(level));
>> +}
>>  
>>  #endif
>> diff --git a/arch/arm64/kvm/mmu.c b/arch/arm64/kvm/mmu.c
>> index 22c00274884a..f29558c5dcbc 100644
>> --- a/arch/arm64/kvm/mmu.c
>> +++ b/arch/arm64/kvm/mmu.c
>> @@ -834,16 +834,17 @@ void stage2_unmap_vm(struct kvm *kvm)
>>  void kvm_free_stage2_pgd(struct kvm_s2_mmu *mmu)
>>  {
>>  	struct kvm *kvm = kvm_s2_mmu_to_kvm(mmu);
>> -	struct kvm_pgtable *pgt = NULL;
>> +	struct kvm_pgtable *pgt;
>>  
>>  	write_lock(&kvm->mmu_lock);
>> +	pgt = mmu->pgt;
>>  	if (kvm_is_realm(kvm) &&
>>  	    kvm_realm_state(kvm) != REALM_STATE_DYING) {
>> -		/* TODO: teardown rtts */
>>  		write_unlock(&kvm->mmu_lock);
>> +		kvm_realm_destroy_rtts(&kvm->arch.realm, pgt->ia_bits,
>> +				       pgt->start_level);
>>  		return;
>>  	}
>> -	pgt = mmu->pgt;
>>  	if (pgt) {
>>  		mmu->pgd_phys = 0;
>>  		mmu->pgt = NULL;
>> diff --git a/arch/arm64/kvm/rme.c b/arch/arm64/kvm/rme.c
>> index 0c9d70e4d9e6..f7b0e5a779f8 100644
>> --- a/arch/arm64/kvm/rme.c
>> +++ b/arch/arm64/kvm/rme.c
>> @@ -73,6 +73,28 @@ static int rmi_check_version(void)
>>  	return 0;
>>  }
>>  
>> +static void realm_destroy_undelegate_range(struct realm *realm,
>> +					   unsigned long ipa,
>> +					   unsigned long addr,
>> +					   ssize_t size)
>> +{
>> +	unsigned long rd = virt_to_phys(realm->rd);
>> +	int ret;
>> +
>> +	while (size > 0) {
>> +		ret = rmi_data_destroy(rd, ipa);
>> +		WARN_ON(ret);
>> +		ret = rmi_granule_undelegate(addr);
>> +
> As the return value is not documented, what will happen if a page undelegate
> failed? Leaked? Some explanation is required here.

Yes - it's leaked. I'll add a comment to explain the get_page() call.

Thanks,

Steve

>> +		if (ret)
>> +			get_page(phys_to_page(addr));
>> +
>> +		addr += PAGE_SIZE;
>> +		ipa += PAGE_SIZE;
>> +		size -= PAGE_SIZE;
>> +	}
>> +}
>> +
>>  static unsigned long create_realm_feat_reg0(struct kvm *kvm)
>>  {
>>  	unsigned long ia_bits = VTCR_EL2_IPA(kvm->arch.vtcr);
>> @@ -170,6 +192,123 @@ static int realm_create_rd(struct kvm *kvm)
>>  	return r;
>>  }
>>  
>> +static int realm_rtt_destroy(struct realm *realm, unsigned long addr,
>> +			     int level, phys_addr_t rtt_granule)
>> +{
>> +	addr = ALIGN_DOWN(addr, rme_rtt_level_mapsize(level - 1));
>> +	return rmi_rtt_destroy(rtt_granule, virt_to_phys(realm->rd), addr,
>> +			level);
>> +}
>> +
>> +static int realm_destroy_free_rtt(struct realm *realm, unsigned long addr,
>> +				  int level, phys_addr_t rtt_granule)
>> +{
>> +	if (realm_rtt_destroy(realm, addr, level, rtt_granule))
>> +		return -ENXIO;
>> +	if (!WARN_ON(rmi_granule_undelegate(rtt_granule)))
>> +		put_page(phys_to_page(rtt_granule));
>> +
>> +	return 0;
>> +}
>> +
>> +static int realm_rtt_create(struct realm *realm,
>> +			    unsigned long addr,
>> +			    int level,
>> +			    phys_addr_t phys)
>> +{
>> +	addr = ALIGN_DOWN(addr, rme_rtt_level_mapsize(level - 1));
>> +	return rmi_rtt_create(phys, virt_to_phys(realm->rd), addr, level);
>> +}
>> +
>> +static int realm_tear_down_rtt_range(struct realm *realm, int level,
>> +				     unsigned long start, unsigned long end)
>> +{
>> +	phys_addr_t rd = virt_to_phys(realm->rd);
>> +	ssize_t map_size = rme_rtt_level_mapsize(level);
>> +	unsigned long addr, next_addr;
>> +	bool failed = false;
>> +
>> +	for (addr = start; addr < end; addr = next_addr) {
>> +		phys_addr_t rtt_addr, tmp_rtt;
>> +		struct rtt_entry rtt;
>> +		unsigned long end_addr;
>> +
>> +		next_addr = ALIGN(addr + 1, map_size);
>> +
>> +		end_addr = min(next_addr, end);
>> +
>> +		if (rmi_rtt_read_entry(rd, ALIGN_DOWN(addr, map_size),
>> +				       level, &rtt)) {
>> +			failed = true;
>> +			continue;
>> +		}
>> +
>> +		rtt_addr = rmi_rtt_get_phys(&rtt);
>> +		WARN_ON(level != rtt.walk_level);
>> +
>> +		switch (rtt.state) {
>> +		case RMI_UNASSIGNED:
>> +		case RMI_DESTROYED:
>> +			break;
>> +		case RMI_TABLE:
>> +			if (realm_tear_down_rtt_range(realm, level + 1,
>> +						      addr, end_addr)) {
>> +				failed = true;
>> +				break;
>> +			}
>> +			if (IS_ALIGNED(addr, map_size) &&
>> +			    next_addr <= end &&
>> +			    realm_destroy_free_rtt(realm, addr, level + 1,
>> +						   rtt_addr))
>> +				failed = true;
>> +			break;
>> +		case RMI_ASSIGNED:
>> +			WARN_ON(!rtt_addr);
>> +			/*
>> +			 * If there is a block mapping, break it now, using the
>> +			 * spare_page. We are sure to have a valid delegated
>> +			 * page at spare_page before we enter here, otherwise
>> +			 * WARN once, which will be followed by further
>> +			 * warnings.
>> +			 */
>> +			tmp_rtt = realm->spare_page;
>> +			if (level == 2 &&
>> +			    !WARN_ON_ONCE(tmp_rtt == PHYS_ADDR_MAX) &&
>> +			    realm_rtt_create(realm, addr,
>> +					     RME_RTT_MAX_LEVEL, tmp_rtt)) {
>> +				WARN_ON(1);
>> +				failed = true;
>> +				break;
>> +			}
>> +			realm_destroy_undelegate_range(realm, addr,
>> +						       rtt_addr, map_size);
>> +			/*
>> +			 * Collapse the last level table and make the spare page
>> +			 * reusable again.
>> +			 */
>> +			if (level == 2 &&
>> +			    realm_rtt_destroy(realm, addr, RME_RTT_MAX_LEVEL,
>> +					      tmp_rtt))
>> +				failed = true;
>> +			break;
>> +		case RMI_VALID_NS:
>> +			WARN_ON(rmi_rtt_unmap_unprotected(rd, addr, level));
>> +			break;
>> +		default:
>> +			WARN_ON(1);
>> +			failed = true;
>> +			break;
>> +		}
>> +	}
>> +
>> +	return failed ? -EINVAL : 0;
>> +}
>> +
>> +void kvm_realm_destroy_rtts(struct realm *realm, u32 ia_bits, u32 start_level)
>> +{
>> +	realm_tear_down_rtt_range(realm, start_level, 0, (1UL << ia_bits));
>> +}
>> +
>>  /* Protects access to rme_vmid_bitmap */
>>  static DEFINE_SPINLOCK(rme_vmid_lock);
>>  static unsigned long *rme_vmid_bitmap;
>