hw/arm/smmuv3: Correct stream ID and event address handling
cover hw/arm/smmuv3: Correct stream ID and event address handling
While working on the Linux SMMUv3 driver, I noticed a few cases where the QEMU SMMUv3 behavior relating to stream tables was inconsistent with our hardware.
Also, when debugging those differences, I found that the errors reported through the QEMU SMMUv3 event queue contained the address fields in an incorrect position.
These patches correct the QEMU SMMUv3 behavior to match the specification (and the behavior that I observed in our hardware). Linux guests normally will not notice these issues, but other SMMUv3 driver implementations might.
Simon Veith (5):
hw/arm/smmuv3: Apply address mask to linear strtab base address hw/arm/smmuv3: Check stream IDs against actual table LOG2SIZE hw/arm/smmuv3: Align stream table base address to table size hw/arm/smmuv3: Use correct bit positions in EVT_SET_ADDR2 macro hw/arm/smmuv3: Report F_STE_FETCH fault address in correct word
position
hw/arm/smmuv3-internal.h | 4 ++-- hw/arm/smmuv3.c | 39 ++++++++++++++++++++++++++++++++------- 2 files changed, 34 insertions(+), 9 deletions(-)
Signature · 1 line
2.7.4
PATCH 1/5 hw/arm/smmuv3: Apply address mask to linear strtab base address
In the SMMU_STRTAB_BASE register, the stream table base address only occupies bits [51:6]. Other bits, such as RA (bit [62]), must be masked out to obtain the base address.
The branch for 2-level stream tables correctly applies this mask by way of SMMU_BASE_ADDR_MASK, but the one for linear stream tables does not.
Apply the missing mask in that case as well so that the correct stream base address is used by guests which configure a linear stream table.
Linux guests are unaffected by this change because they choose a 2-level stream table layout for the QEMU SMMUv3, based on the size of its stream ID space.
ref. ARM IHI 0070C, section 6.3.23.
--- hw/arm/smmuv3.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
hw/arm/smmuv3.c +1 -1
@@ -429,7 +429,7 @@ static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, } addr = l2ptr + l2_ste_offset *sizeof (*ste); }else { - addr = s->strtab_base + sid *sizeof (*ste); + addr = (s->strtab_base &SMMU_BASE_ADDR_MASK ) + sid *sizeof (*ste); }if (smmu_get_ste (s, addr, ste, event)) {
Signature · 1 line
2.7.4
Re: [PATCH 1/5] hw/arm/smmuv3: Apply address mask to linear strtab base address
Hi Simon,
Quoted 29 lines · Simon Veith
In the
SMMU_STRTAB_BASEregister, the stream table base address only occupies bits [51:6]. Other bits, such as RA (bit [62]), must be masked out to obtain the base address.The branch for 2-level stream tables correctly applies this mask by way of
SMMU_BASE_ADDR_MASK, but the one for linear stream tables does not.Apply the missing mask in that case as well so that the correct stream base address is used by guests which configure a linear stream table.
Linux guests are unaffected by this change because they choose a 2-level stream table layout for the QEMU SMMUv3, based on the size of its stream ID space.
ref. ARM IHI 0070C, section 6.3.23.
Signed-off-by Simon Veith <[email protected]>Cc Eric Auger <[email protected]>--- hw/arm/smmuv3.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)hw/arm/smmuv3.c +0
-0
hw/arm/smmuv3.c static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, · line 432
Quoted 5 lines
hw/arm/smmuv3.c +1
-1@@ -429,7 +429,7 @@ static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, } addr = l2ptr + l2_ste_offset *sizeof (*ste); }else { - addr = s->strtab_base + sid *sizeof (*ste); + addr = (s->strtab_base &SMMU_BASE_ADDR_MASK ) + sid *sizeof (*ste);
Not related to this patch but I noticed SMMU_BASE_ADDR_MASK should be 0xffffffffffc0 and not 0xffffffffffe0. I can fix it separately or if you respin, you may fix it as well?
hw/arm/smmuv3.c static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, · line 435
Quoted 4 lines
hw/arm/smmuv3.c +0
-0}if (smmu_get_ste (s, addr, ste, event)) {
Besides
Thanks
Eric
Re: [PATCH 1/5] hw/arm/smmuv3: Apply address mask to linear strtab base address
Hello Eric,
Quoted 3 lines · Auger Eric
Not related to this patch but I noticed
SMMU_BASE_ADDR_MASKshould be0xffffffffffc0and not0xffffffffffe0. I can fix it separately or if you respin, you may fix it as well?
Good catch, thank you. I'll fix it in the next version.
Looking at the upper end of that mask, it seems that we are currently masking out bits 48 through 63, rather than just 51 through 63. The reference manual says that this should be done to match the system physical address size as given by SMMU_IDR5.OAS.
We define SMMU_IDR5_OAS to be 4, which selects a physical address size of 44 bits (ref. section 6.3.6). I think the mask should therefore be 0xfffffffffc0 to clear bits 44 and above. Do you agree?
Ideally, we would derive this mask from our definition of SMMU_IDR5_OAS, but I'm not sure it's okay to stuff a call to oas2bits() into the SMMU_BASE_ADDR_MASK macro.
Regards
Simon
Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Ralf Herbrich Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879
Re: [PATCH 1/5] hw/arm/smmuv3: Apply address mask to linear strtab base address
Hi Simon,
Quoted 11 lines · Simon Veith
Hello Eric,
Quoted 3 lines · Auger Eric
Not related to this patch but I noticed
SMMU_BASE_ADDR_MASKshould be0xffffffffffc0and not0xffffffffffe0. I can fix it separately or if you respin, you may fix it as well?Good catch, thank you. I'll fix it in the next version.
Looking at the upper end of that mask, it seems that we are currently masking out bits 48 through 63, rather than just 51 through 63. The reference manual says that this should be done to match the system physical address size as given by SMMU_IDR5.OAS.
Yes you're right. This should go up to 51 as per the field range definition. Spec says address bits and above this field range are treated as zero.
Quoted 2 lines
We define
SMMU_IDR5_OASto be 4, which selects a physical address size of 44 bits (ref. section 6.3.6). I think the mask should therefore be0xfffffffffc0to clear bits 44 and above. Do you agree?
bits beyond the OAS are RES0. The spec does not says those fields are treated as zero, as explicitly mentioned for bits > 51. Normally the guest should not set them to something != 0, this would be a programming error right? Guest is supposed to read the IDR5 and program accordingly?
Quoted 2 lines
Ideally, we would derive this mask from our definition of
SMMU_IDR5_OAS, but I'm not sure it's okay to stuff a call tooas2bits()into theSMMU_BASE_ADDR_MASKmacro.
Well I am not sure this is worth the candle. I am not sure we are obliged to enforce this.
Thanks
Eric
Quoted 15 lines
Regards
SimonAmazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Ralf Herbrich Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879
PATCH 2/5 hw/arm/smmuv3: Check stream IDs against actual table LOG2SIZE
In reply to Simon VeithWhen checking whether a stream ID is in range of the stream table, we have so far been only checking it against our implementation limit (SMMU_IDR1_SIDSIZE). However, the guest can program the STRTAB_BASE_CFG.LOG2SIZE field to a size that is smaller than this limit.
Check the stream ID against this limit as well to match the hardware behavior of raising C_BAD_STREAMID events in case the limit is exceeded.
ref. ARM IHI 0070C, section 6.3.24.
--- hw/arm/smmuv3.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-)
hw/arm/smmuv3.c +6 -2
@@ -377,11 +377,15 @@ static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, SMMUEventInfo *event) {dma_addr_t addr; +uint32_t log2size;int ret;trace_smmuv3_find_ste (sid, s->features, s->sid_split); -/* Check SID range */ -if (sid > (1 <<SMMU_IDR1_SIDSIZE )) { + log2size =FIELD_EX32 (s->strtab_base_cfg,STRTAB_BASE_CFG ,LOG2SIZE ); +/* +* Check SID range against both guest-configured and implementation limits +*/ +if (sid > (1 <<MIN (log2size,SMMU_IDR1_SIDSIZE ))) { event->type =SMMU_EVT_C_BAD_STREAMID ;return -EINVAL ; }
Signature · 1 line
2.7.4
Re: [PATCH 2/5] hw/arm/smmuv3: Check stream IDs against actual table LOG2SIZE
Hi Simon,
Quoted 28 lines · Simon Veith
When checking whether a stream ID is in range of the stream table, we have so far been only checking it against our implementation limit (
SMMU_IDR1_SIDSIZE). However, the guest can program the STRTAB_BASE_CFG.LOG2SIZE field to a size that is smaller than this limit.Check the stream ID against this limit as well to match the hardware behavior of raising
C_BAD_STREAMIDevents in case the limit is exceeded.ref. ARM IHI 0070C, section 6.3.24.
Signed-off-by Simon Veith <[email protected]>Cc Eric Auger <[email protected]>--- hw/arm/smmuv3.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-)hw/arm/smmuv3.c +1
-0@@ -377,11 +377,15 @@ static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, SMMUEventInfo *event) {dma_addr_t addr; +uint32_t log2size;
hw/arm/smmuv3.c static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, · line 388
Quoted 10 lines
hw/arm/smmuv3.c +5
-2int ret;trace_smmuv3_find_ste (sid, s->features, s->sid_split); -/* Check SID range */ -if (sid > (1 <<SMMU_IDR1_SIDSIZE )) { + log2size =FIELD_EX32 (s->strtab_base_cfg,STRTAB_BASE_CFG ,LOG2SIZE ); +/* +* Check SID range against both guest-configured and implementation limits +*/ + if (sid > (1 << MIN(log2size, SMMU_IDR1_SIDSIZE))) {I think this should be sid >= (1 << MIN(log2size, SMMU_IDR1_SIDSIZE))
If you agree can you fix it at the same time?
hw/arm/smmuv3.c static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, · line 391
Quoted 4 lines
hw/arm/smmuv3.c +0
-0event->type =SMMU_EVT_C_BAD_STREAMID ;return -EINVAL ; }
Thanks
Eric
PATCH 3/5 hw/arm/smmuv3: Align stream table base address to table size
In reply to Simon VeithPer the specification, and as observed in hardware, the SMMUv3 aligns the SMMU_STRTAB_BASE address to the size of the table by masking out the respective least significant bits in the ADDR field.
Apply this masking logic to our smmu_find_ste() lookup function per the specification.
ref. ARM IHI 0070C, section 6.3.23.
--- hw/arm/smmuv3.c | 29 +++++++++++++++++++++++++---- 1 file changed, 25 insertions(+), 4 deletions(-)
hw/arm/smmuv3.c +25 -4
@@ -376,8 +376,9 @@ bad_ste:static int smmu_find_ste (SMMUv3State *s,uint32_t sid,STE *ste,SMMUEventInfo *event) { -dma_addr_t addr; +dma_addr_t addr, strtab_base;uint32_t log2size; +int strtab_size_shift;int ret;trace_smmuv3_find_ste (sid, s->features, s->sid_split);
@@ -391,10 +392,23 @@ static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, }if (s->features &SMMU_FEATURE_2LVL_STE ) {int l1_ste_offset, l2_ste_offset, max_l2_ste, span; -dma_addr_t strtab_base, l1ptr, l2ptr; +dma_addr_t l1ptr, l2ptr;STEDesc l1std; - strtab_base = s->strtab_base &SMMU_BASE_ADDR_MASK ; +/* +* Align strtab base address to table size. For this purpose, assume it +* is not bounded by SMMU_IDR1_SIDSIZE. +*/ + strtab_size_shift = log2size - s->sid_split -1 +3 ; +if (strtab_size_shift <DMA_ADDR_BITS ) { +if (strtab_size_shift <5 ) { + strtab_size_shift =5 ; + } + strtab_base = s->strtab_base &SMMU_BASE_ADDR_MASK & + ~((1ULL << strtab_size_shift) -1 ); + }else { + strtab_base =0 ; + } l1_ste_offset = sid >> s->sid_split; l2_ste_offset = sid & ((1 << s->sid_split) -1 ); l1ptr = (dma_addr_t)(strtab_base + l1_ste_offset *sizeof (l1std));
@@ -433,7 +447,14 @@ static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, } addr = l2ptr + l2_ste_offset *sizeof (*ste); }else { - addr = (s->strtab_base &SMMU_BASE_ADDR_MASK ) + sid *sizeof (*ste); + strtab_size_shift = log2size +5 ; +if (strtab_size_shift <DMA_ADDR_BITS ) { + strtab_base = s->strtab_base &SMMU_BASE_ADDR_MASK & + ~((1ULL << strtab_size_shift) -1 ); + }else { + strtab_base =0 ; + } + addr = strtab_base + sid *sizeof (*ste); }if (smmu_get_ste (s, addr, ste, event)) {
Signature · 1 line
2.7.4
Re: [PATCH 3/5] hw/arm/smmuv3: Align stream table base address to table size
Hi Simon,
Quoted 36 lines · Simon Veith
Per the specification, and as observed in hardware, the SMMUv3 aligns the
SMMU_STRTAB_BASEaddress to the size of the table by masking out the respective least significant bits in the ADDR field.Apply this masking logic to our
smmu_find_ste()lookup function per the specification.ref. ARM IHI 0070C, section 6.3.23.
Signed-off-by Simon Veith <[email protected]>Cc Eric Auger <[email protected]>--- hw/arm/smmuv3.c | 29 +++++++++++++++++++++++++---- 1 file changed, 25 insertions(+), 4 deletions(-)hw/arm/smmuv3.c +2
-1@@ -376,8 +376,9 @@ bad_ste:static int smmu_find_ste (SMMUv3State *s,uint32_t sid,STE *ste,SMMUEventInfo *event) { -dma_addr_t addr; +dma_addr_t addr, strtab_base;uint32_t log2size; +int strtab_size_shift;int ret;trace_smmuv3_find_ste (sid, s->features, s->sid_split);@@ -391,10 +392,23 @@ static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, }if (s->features &SMMU_FEATURE_2LVL_STE ) {int l1_ste_offset, l2_ste_offset, max_l2_ste, span;
hw/arm/smmuv3.c static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, · line 402
Quoted 10 lines
hw/arm/smmuv3.c +6
-2-dma_addr_t strtab_base, l1ptr, l2ptr; +dma_addr_t l1ptr, l2ptr;STEDesc l1std; - strtab_base = s->strtab_base &SMMU_BASE_ADDR_MASK ; +/* +* Align strtab base address to table size. For this purpose, assume it +* is not bounded by SMMU_IDR1_SIDSIZE. +*/ + strtab_size_shift = log2size - s->sid_split -1 +3 ;
Maybe just as the spec use MAX(5, (LOG2SIZE-SPLIT-1+3))?
hw/arm/smmuv3.c static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, · line 408
Quoted 6 lines
hw/arm/smmuv3.c +6
-0+if (strtab_size_shift <DMA_ADDR_BITS ) { +if (strtab_size_shift <5 ) { + strtab_size_shift =5 ; + } + strtab_base = s->strtab_base &SMMU_BASE_ADDR_MASK & + ~((1ULL << strtab_size_shift) -1 );
nit: you may use ~MAKE_64BIT_MASK(0, strtab_size_shift)
hw/arm/smmuv3.c static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, · line 409
Quoted 1 line
hw/arm/smmuv3.c +1
-0+ }else {
see below
Quoted 6 lines
hw/arm/smmuv3.c +2
-0+ strtab_base =0 ; + } l1_ste_offset = sid >> s->sid_split; l2_ste_offset = sid & ((1 << s->sid_split) -1 ); l1ptr = (dma_addr_t)(strtab_base + l1_ste_offset *sizeof (l1std));
hw/arm/smmuv3.c static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, · line 454
Quoted 9 lines
hw/arm/smmuv3.c +5
-1@@ -433,7 +447,14 @@ static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, } addr = l2ptr + l2_ste_offset *sizeof (*ste); }else { - addr = (s->strtab_base &SMMU_BASE_ADDR_MASK ) + sid *sizeof (*ste); + strtab_size_shift = log2size +5 ; +if (strtab_size_shift <DMA_ADDR_BITS ) { + strtab_base = s->strtab_base &SMMU_BASE_ADDR_MASK & + ~((1ULL << strtab_size_shift) -1 ); + }else {
Can it happen? I understand LOG2SIZE <= SMMU_S_IDR_5.SIDSIZE and SIDSIZE is max 32. Same above?
hw/arm/smmuv3.c static int smmu_find_ste(SMMUv3State *s, uint32_t sid, STE *ste, · line 460
Quoted 7 lines
hw/arm/smmuv3.c +3
-0+ strtab_base =0 ; + } + addr = strtab_base + sid *sizeof (*ste); }if (smmu_get_ste (s, addr, ste, event)) {
Thank you for those series fixes.
Best Regards
Eric
PATCH 4/5 hw/arm/smmuv3: Use correct bit positions in EVT_SET_ADDR2 macro
In reply to Simon VeithThe bit offsets in the EVT_SET_ADDR2 macro do not match those specified in the ARM SMMUv3 Architecture Specification. In all events that use this macro, e.g. F_WALK_EABT, the faulting fetch address or IPA actually occupies the 32-bit words 6 and 7 in the event record contiguously, with the upper and lower unused bits clear due to alignment or maximum supported address bits. How many bits are clear depends on the individual event type.
Update the macro to write to the correct words in the event record so that guest drivers can obtain accurate address information on events.
ref. ARM IHI 0070C, sections 7.3.12 through 7.3.16.
--- hw/arm/smmuv3-internal.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
hw/arm/smmuv3-internal.h +2 -2
@@ -461,8 +461,8 @@ typedef struct SMMUEventInfo { } while (0)#define EVT_SET_ADDR2 (x, addr) \ do { \ - (x)->word[7] = deposit32((x)->word[7], 3, 29, addr >> 16); \ - (x)->word[7] = deposit32((x)->word[7], 0, 16, addr & 0xffff);\ + (x)->word[7] = (uint32_t)(addr >> 32); \ + (x)->word[6] = (uint32_t)(addr & 0xffffffff); \ } while (0)void smmuv3_record_event (SMMUv3State *s,SMMUEventInfo *event);
Signature · 1 line
2.7.4
Re: [PATCH 4/5] hw/arm/smmuv3: Use correct bit positions in EVT_SET_ADDR2 macro
Hi Simon,
Quoted 8 lines · Simon Veith
The bit offsets in the
EVT_SET_ADDR2macro do not match those specified in the ARM SMMUv3 Architecture Specification. In all events that use this macro, e.g.F_WALK_EABT, the faulting fetch address or IPA actually occupies the 32-bit words 6 and 7 in the event record contiguously, with the upper and lower unused bits clear due to alignment or maximum supported address bits. How many bits are clear depends on the individual event type.
Quoted 6 lines
Update the macro to write to the correct words in the event record so that guest drivers can obtain accurate address information on events.
ref. ARM IHI 0070C, sections 7.3.12 through 7.3.16.
Signed-off-by Simon Veith <[email protected]>
Thanks
Eric
Quoted 23 lines
--- hw/arm/smmuv3-internal.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)hw/arm/smmuv3-internal.h +2
-2@@ -461,8 +461,8 @@ typedef struct SMMUEventInfo { } while (0)#define EVT_SET_ADDR2 (x, addr) \ do { \ - (x)->word[7] = deposit32((x)->word[7], 3, 29, addr >> 16); \ - (x)->word[7] = deposit32((x)->word[7], 0, 16, addr & 0xffff);\ + (x)->word[7] = (uint32_t)(addr >> 32); \ + (x)->word[6] = (uint32_t)(addr & 0xffffffff); \ } while (0)void smmuv3_record_event (SMMUv3State *s,SMMUEventInfo *event);
PATCH 5/5 hw/arm/smmuv3: Report F_STE_FETCH fault address in correct word position
In reply to Simon VeithThe smmuv3_record_event() function that generates the F_STE_FETCH error uses the EVT_SET_ADDR macro to record the fetch address, placing it in 32-bit words 4 and 5.
The correct position for this address is in words 6 and 7, per the SMMUv3 Architecture Specification.
Update the function to use the EVT_SET_ADDR2 macro instead, which is the macro intended for writing to these words.
ref. ARM IHI 0070C, section 7.3.4.
--- hw/arm/smmuv3.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
hw/arm/smmuv3.c +1 -1
@@ -172,7 +172,7 @@ void smmuv3_record_event(SMMUv3State *s, SMMUEventInfo *info)case SMMU_EVT_F_STE_FETCH :EVT_SET_SSID (&evt, info->u.f_ste_fetch.ssid);EVT_SET_SSV (&evt, info->u.f_ste_fetch.ssv); -EVT_SET_ADDR (&evt, info->u.f_ste_fetch.addr); +EVT_SET_ADDR2 (&evt, info->u.f_ste_fetch.addr);break ;case SMMU_EVT_C_BAD_STE :EVT_SET_SSID (&evt, info->u.c_bad_ste.ssid);
Signature · 1 line
2.7.4
Re: [PATCH 5/5] hw/arm/smmuv3: Report F_STE_FETCH fault address in correct word position
Hi Simon,
Quoted 26 lines · Simon Veith
The
smmuv3_record_event()function that generates theF_STE_FETCHerror uses theEVT_SET_ADDRmacro to record the fetch address, placing it in 32-bit words 4 and 5.The correct position for this address is in words 6 and 7, per the SMMUv3 Architecture Specification.
Update the function to use the
EVT_SET_ADDR2macro instead, which is the macro intended for writing to these words.ref. ARM IHI 0070C, section 7.3.4.
Signed-off-by Simon Veith <[email protected]>Cc Eric Auger <[email protected]>--- hw/arm/smmuv3.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)hw/arm/smmuv3.c +0
-0
hw/arm/smmuv3.c void smmuv3_record_event(SMMUv3State *s, SMMUEventInfo *info) · line 178
Quoted 8 lines
hw/arm/smmuv3.c +1
-1@@ -172,7 +172,7 @@ void smmuv3_record_event(SMMUv3State *s, SMMUEventInfo *info)case SMMU_EVT_F_STE_FETCH :EVT_SET_SSID (&evt, info->u.f_ste_fetch.ssid);EVT_SET_SSV (&evt, info->u.f_ste_fetch.ssv); -EVT_SET_ADDR (&evt, info->u.f_ste_fetch.addr); +EVT_SET_ADDR2 (&evt, info->u.f_ste_fetch.addr);break ;case SMMU_EVT_C_BAD_STE :EVT_SET_SSID (&evt, info->u.c_bad_ste.ssid);
Thanks
Eric