On 6/16/21 11:02 PM, shashi.mall...@linaro.org wrote:
> Hi Eric,
>
> Please find my responses inline (below):-
>
> On Sun, 2021-06-13 at 17:55 +0200, Eric Auger wrote:
>> Hi Shashi,
>>
>> On 6/2/21 8:00 PM, Shashi Mallela wrote:
>>> Added ITS command queue handling for MAPTI,MAPI commands,handled
>>> ITS
>>> translation which triggers an LPI via INT command as well as write
>>> to GITS_TRANSLATER register,defined enum to differentiate between
>>> ITS
>>> command interrupt trigger and GITS_TRANSLATER based interrupt
>>> trigger.
>>> Each of these commands make use of other functionalities
>>> implemented to
>>> get device table entry,collection table entry or interrupt
>>> translation
>>> table entry required for their processing.
>>>
>>> Signed-off-by: Shashi Mallela <shashi.mall...@linaro.org>
>>> ---
>>> hw/intc/arm_gicv3_its.c | 334
>>> +++++++++++++++++++++++++++++
>>> hw/intc/gicv3_internal.h | 12 ++
>>> include/hw/intc/arm_gicv3_common.h | 2 +
>>> 3 files changed, 348 insertions(+)
>>>
>>> diff --git a/hw/intc/arm_gicv3_its.c b/hw/intc/arm_gicv3_its.c
>>> index 6551c577b3..82bb5b84ef 100644
>>> --- a/hw/intc/arm_gicv3_its.c
>>> +++ b/hw/intc/arm_gicv3_its.c
>>> @@ -28,6 +28,13 @@ struct GICv3ITSClass {
>>> void (*parent_reset)(DeviceState *dev);
>>> };
>>>
>>> +typedef enum ItsCmdType {
>>> + NONE = 0, /* internal indication for GITS_TRANSLATER write */
>>> + CLEAR = 1,
>>> + DISCARD = 2,
>>> + INT = 3,
>>> +} ItsCmdType;
>> Add a comment to explain what this enum stand for. This sounds
>> misleading to me versus the command IDs. Why don't you use the cmd id
>> then and add NONE?
> This is an internal enum used to distinguish between interrupt
> triggered via command queue and interrupt triggered via gits_translater
> write.Will add the same comment in code.
> Since NONE is only 1 command applicable for GITS_TRANSLATER,started
> with it so that in the future if any further command queue commands
> have to be added we can just extend the numbering.
>>> +
>>> static uint64_t baser_base_addr(uint64_t value, uint32_t page_sz)
>>> {
>>> uint64_t result = 0;
>>> @@ -49,6 +56,315 @@ static uint64_t baser_base_addr(uint64_t value,
>>> uint32_t page_sz)
>>> return result;
>>> }
>>>
>>> +static bool get_cte(GICv3ITSState *s, uint16_t icid, uint64_t
>>> *cte,
>>> + MemTxResult *res)
>>> +{
>>> + AddressSpace *as = &s->gicv3->dma_as;
>>> + uint64_t l2t_addr;
>>> + uint64_t value;
>>> + bool valid_l2t;
>>> + uint32_t l2t_id;
>>> + uint32_t max_l2_entries;
>>> + bool status = false;
>>> +
>>> + if (s->ct.indirect) {
>>> + l2t_id = icid / (s->ct.page_sz / L1TABLE_ENTRY_SIZE);
>>> +
>>> + value = address_space_ldq_le(as,
>>> + s->ct.base_addr +
>>> + (l2t_id *
>>> L1TABLE_ENTRY_SIZE),
>>> + MEMTXATTRS_UNSPECIFIED, res);
>>> +
>>> + if (*res == MEMTX_OK) {
>>> + valid_l2t = (value >> VALID_SHIFT) & VALID_MASK;
>>> +
>>> + if (valid_l2t) {
>>> + max_l2_entries = s->ct.page_sz / s->ct.entry_sz;
>>> +
>>> + l2t_addr = value & ((1ULL << 51) - 1);
>>> +
>>> + *cte = address_space_ldq_le(as, l2t_addr +
>>> + ((icid % max_l2_entries) *
>>> GITS_CTE_SIZE),
>>> + MEMTXATTRS_UNSPECIFIED, res);
>>> + }
>>> + }
>>> + } else {
>>> + /* Flat level table */
>>> + *cte = address_space_ldq_le(as, s->ct.base_addr +
>>> + (icid * GITS_CTE_SIZE),
>>> + MEMTXATTRS_UNSPECIFIED,
>>> res);
>>> + }
>>> +
>>> + if (*cte & VALID_MASK) {
>>> + status = true;
>>> + }
>>> +
>>> + return status;
>>> +}
>>> +
>>> +static MemTxResult update_ite(GICv3ITSState *s, uint32_t eventid,
>>> uint64_t dte,
>>> + uint64_t itel, uint32_t iteh)
>> why not introducing an ite struct instead of the h/l args?based on
>> the same reason for packing individual fields within bit positions
>> instead of using more memory to store the same fields as struct
>> members.
> Will create an ite struct with existing itel & iteh as members and
> retain their bit fields processing to avoid extra params being passed.
>>> +{
>>> + AddressSpace *as = &s->gicv3->dma_as;
>>> + uint64_t itt_addr;
>>> + MemTxResult res = MEMTX_OK;
>>> +
>>> + itt_addr = (dte >> 6ULL) & ITTADDR_MASK;
>>> + itt_addr <<= ITTADDR_SHIFT; /* 256 byte aligned */
>>> +
>>> + address_space_stq_le(as, itt_addr + (eventid *
>>> sizeof(uint64_t)),
>>> + itel, MEMTXATTRS_UNSPECIFIED, &res);
>>> +
>>> + if (res == MEMTX_OK) {
>>> + address_space_stl_le(as, itt_addr + ((eventid +
>>> sizeof(uint64_t)) *
>>> + sizeof(uint32_t)), iteh,
>>> MEMTXATTRS_UNSPECIFIED,
>>> + &res);
>>> + }
>>> + return res;
>>> +}
>>> +
>>> +static bool get_ite(GICv3ITSState *s, uint32_t eventid, uint64_t
>>> dte,
>>> + uint16_t *icid, uint32_t *pIntid, MemTxResult
>>> *res)
>>> +{
>>> + AddressSpace *as = &s->gicv3->dma_as;
>>> + uint64_t itt_addr;
>>> + bool status = false;
>>> + uint64_t itel = 0;
>>> + uint32_t iteh = 0;
>>> +
>>> + itt_addr = (dte >> 6ULL) & ITTADDR_MASK;
>>> + itt_addr <<= ITTADDR_SHIFT; /* 256 byte aligned */
>>> +
>>> + itel = address_space_ldq_le(as, itt_addr + (eventid *
>>> sizeof(uint64_t)),
>>> + MEMTXATTRS_UNSPECIFIED, res);
>>> +
>>> + if (*res == MEMTX_OK) {
>>> + iteh = address_space_ldl_le(as, itt_addr + ((eventid +
>>> + sizeof(uint64_t)) *
>>> sizeof(uint32_t)),
>>> + MEMTXATTRS_UNSPECIFIED, res);
>>> +
>>> + if (*res == MEMTX_OK) {
>>> + if (itel & VALID_MASK) {
>>> + if ((itel >> ITE_ENTRY_INTTYPE_SHIFT) &
>>> GITS_TYPE_PHYSICAL) {
>>> + *pIntid = (itel >> ITE_ENTRY_INTID_SHIFT) &
>>> + ITE_ENTRY_INTID_MASK;
>>> + *icid = iteh & ITE_ENTRY_ICID_MASK;
>>> + status = true;
>>> + }
>>> + }
>>> + }
>>> + }
>>> + return status;
>>> +}
>>> +
>>> +static uint64_t get_dte(GICv3ITSState *s, uint32_t devid,
>>> MemTxResult *res)
>> maybe the code would be more readable if you were returning a strcut
>> for
>> dte/cte instead of uint64_t. The decoding of the fields would be done
>> here instead?
> Both dte and cte are 8 bytes and hence chose a single uint64_t to
> efficiently utilize the applicable bit fields in the variable instead
> of a struct(either with bitfields to match the current field layouts
> within dte/cte/ite entries or use more memory variables within the
> struct to hold each individual dte/cte/ite fields)
>>> +{
>>> + AddressSpace *as = &s->gicv3->dma_as;
>>> + uint64_t l2t_addr;
>>> + uint64_t value;
>>> + bool valid_l2t;
>>> + uint32_t l2t_id;
>>> + uint32_t max_l2_entries;
>>> +
>>> + if (s->dt.indirect) {
>>> + l2t_id = devid / (s->dt.page_sz / L1TABLE_ENTRY_SIZE);
>>> +
>>> + value = address_space_ldq_le(as,
>>> + s->dt.base_addr +
>>> + (l2t_id *
>>> L1TABLE_ENTRY_SIZE),
>>> + MEMTXATTRS_UNSPECIFIED, res);
>>> +
>>> + if (*res == MEMTX_OK) {
>>> + valid_l2t = (value >> VALID_SHIFT) & VALID_MASK;
>>> +
>>> + if (valid_l2t) {
>>> + max_l2_entries = s->dt.page_sz / s->dt.entry_sz;
>>> +
>>> + l2t_addr = value & ((1ULL << 51) - 1);
>>> +
>>> + value = address_space_ldq_le(as, l2t_addr +
>>> + ((devid % max_l2_entries) *
>>> GITS_DTE_SIZE),
>>> + MEMTXATTRS_UNSPECIFIED, res);
>>> + }
>>> + }
>>> + } else {
>>> + /* Flat level table */
>>> + value = address_space_ldq_le(as, s->dt.base_addr +
>>> + (devid * GITS_DTE_SIZE),
>>> + MEMTXATTRS_UNSPECIFIED, res);
>>> + }
>>> +
>>> + return value;
>>> +}
>> I think a common helper could be defined for get_cte and get_dte.
> was just trying to keep the code modular for getting dte and cte
> instead of various if else's within a common helper
>>> +
>>> +static MemTxResult process_int(GICv3ITSState *s, uint64_t value,
>>> + uint32_t offset, ItsCmdType cmd)
>> this is a bit misleanding as INT is a command. You should rename it I
>> think. Also it is not really homogeneous with other cmds, ie. you
>> have
>> process_mapti, process_mapd, process_mac and all the remaining cmds
>> are
>> handled though this one? At least add a doc comment to explain what
>> it does.
> the naming of this function is along the lines of other ITS commands
> i.e process_xxx format where xxx is the actual ITS command.
> There is no connection between handling of process_mapd,process_mapc
> with process_int.Each of them handle their respective command
> processing independently.
Well I am not big fan of keeping that name ;-) but well.
>>> +{
>>> + AddressSpace *as = &s->gicv3->dma_as;
>>> + uint32_t devid, eventid;
>>> + MemTxResult res = MEMTX_OK;
>>> + bool dte_valid;
>>> + uint64_t dte = 0;
>>> + uint32_t max_eventid;
>>> + uint16_t icid = 0;
>>> + uint32_t pIntid = 0;
>>> + bool ite_valid = false;
>>> + uint64_t cte = 0;
>>> + bool cte_valid = false;
>>> + uint64_t itel = 0;
>>> + uint32_t iteh = 0;
>>> +
>>> + if (cmd == NONE) {
>>> + devid = offset;
>>> + } else {
>>> + devid = (value >> DEVID_SHIFT) & DEVID_MASK;
>>> +
>>> + offset += NUM_BYTES_IN_DW;
>>> + value = address_space_ldq_le(as, s->cq.base_addr + offset,
>>> + MEMTXATTRS_UNSPECIFIED,
>>> &res);
>>> + }
>>> +
>>> + if (res != MEMTX_OK) {
>>> + return res;
>>> + }
>>> +
>>> + eventid = (value & EVENTID_MASK);
>>> +
>>> + dte = get_dte(s, devid, &res);
>>> +
>>> + if (res != MEMTX_OK) {
>>> + return res;
>>> + }
>>> + dte_valid = dte & VALID_MASK;
>>> +
>>> + if (dte_valid) {
>>> + max_eventid = (1UL << (((dte >> 1U) & SIZE_MASK) + 1));
>> THE DTE format is implementation defined, and the decoding is hard to
>> follow because it is documented somewhere else.
> The decoding of fields within each of dte,cte and ite has been
> explained as comments in gicv3_internal.h via
> GITS_DTE_SIZE,GITS_CTE_SIZE, ITS_ITT_ENTRY_SIZE defines
>>> +
>>> + ite_valid = get_ite(s, eventid, dte, &icid, &pIntid,
>>> &res);
>>> +
>>> + if (res != MEMTX_OK) {
>>> + return res;
>>> + }
>>> +
>>> + if (ite_valid) {
>>> + cte_valid = get_cte(s, icid, &cte, &res);
>>> + }
>>> +
>>> + if (res != MEMTX_OK) {
>>> + return res;
>>> + }
>> instead of having this process_int() helper, why not having a helper
>> doing just the decoding phase, ie.
>> get_dte -> get_ite -> get_cte and collect the relevant info and
>> collect
>> error and then keep the actual cmd processing in the switch?
> As explained previously chose to keep the code modular instead of a
> large monolothic structure with all the functionality in 1 place.
> process_cmdq handles the command queue (and their read/write offsets)
> while the individual ITS command handling is done by respective
> functions
OK That's not my opinion but it is not a blocker anyway ;-)
>>> + }
>>> +
>>> + if ((devid > s->dt.max_devids) || !dte_valid || !ite_valid ||
>>> + !cte_valid || (eventid > max_eventid)) {
>>> + qemu_log_mask(LOG_GUEST_ERROR,
>>> + "%s: invalid interrupt translation table
>>> attributes "
>>> + "devid %d or eventid %d\n",
>>> + __func__, devid, eventid);
>> the msg does not necessarily match the error case. You mention ITT
>> issue
>> whereas the problem may come from invalid DTE, CTE, or even devid.
> will change it to consider the missing cases
>>> + /*
>>> + * in this implementation,in case of error
>>> + * we ignore this command and move onto the next
>>> + * command in the queue
>>> + */
>> so you don't return an error?
> yes we dont,we just ignore this command and move onto next one in the
> queue
>>> + } else {
>>> + /*
>>> + * Current implementation only supports rdbase == procnum
>>> + * Hence rdbase physical address is ignored
>>> + */
>>> + if (cmd == DISCARD) {
>>> + /* remove mapping from interrupt translation table */
>>> + res = update_ite(s, eventid, dte, itel, iteh);
>> iteh and itel always are 0, why not use a struct ite with valid field
>> unset.
> based on the same reason for packing individual fields within bit
> positions instead of using more memory to store the same fields as
> struct members.
> I could however create an ite struct with existing itel & iteh as
> members and retain their bit fields processing.
>>> + }
>>> + }
>>> +
>>> + return res;
>>> +}
>>> +
>>> +static MemTxResult process_mapti(GICv3ITSState *s, uint64_t value,
>>> + uint32_t offset, bool
>>> ignore_pInt)
>>> +{
>>> + AddressSpace *as = &s->gicv3->dma_as;
>>> + uint32_t devid, eventid;
>>> + uint32_t pIntid = 0;
>>> + uint32_t max_eventid, max_Intid;
>>> + bool dte_valid;
>>> + MemTxResult res = MEMTX_OK;
>>> + uint16_t icid = 0;
>>> + uint64_t dte = 0;
>>> + uint64_t itel = 0;
>>> + uint32_t iteh = 0;
>>> + uint32_t int_spurious = INTID_SPURIOUS;
>>> +
>>> + devid = (value >> DEVID_SHIFT) & DEVID_MASK;
>>> + offset += NUM_BYTES_IN_DW;
>>> + value = address_space_ldq_le(as, s->cq.base_addr + offset,
>>> + MEMTXATTRS_UNSPECIFIED, &res);
>>> +
>>> + if (res != MEMTX_OK) {
>>> + return res;
>>> + }
>>> +
>>> + eventid = (value & EVENTID_MASK);
>>> +
>>> + if (!ignore_pInt) {
>>> + pIntid = (value >> pINTID_OFFSET) & pINTID_MASK;
>>> + }
>>> +
>>> + offset += NUM_BYTES_IN_DW;
>>> + value = address_space_ldq_le(as, s->cq.base_addr + offset,
>>> + MEMTXATTRS_UNSPECIFIED, &res);
>>> +
>>> + if (res != MEMTX_OK) {
>>> + return res;
>>> + }
>>> +
>>> + icid = value & ICID_MASK;
>>> +
>>> + dte = get_dte(s, devid, &res);
>>> +
>>> + if (res != MEMTX_OK) {
>>> + return res;
>>> + }
>>> + dte_valid = dte & VALID_MASK;
>>> +
>>> + max_eventid = (1UL << (((dte >> 1U) & SIZE_MASK) + 1));
>> Again I think you would gain in readibility if get_cte were to return
>> a
>> struct and you would avoid code duplication.
> Like mentioned earlier, preferred to use a single uint64_t for cte (and
> dte) instead of a struct (either with bitfields to match the current
> field layouts within dte/cte/ite entries or use more memory variables
> within the struct to hold each individual dte/cte/ite fields).The
> layout of all these functions follows the pseudocde format defined
> under each ITS command in the spec.
>>> +
>>> + if (!ignore_pInt) {
>>> + max_Intid = (1UL << (FIELD_EX64(s->typer, GITS_TYPER,
>>> IDBITS) + 1));
>>> + }
>>> +
>>> + if ((devid > s->dt.max_devids) || (icid > s->ct.max_collids)
>>> ||
>>> + !dte_valid || (eventid > max_eventid) ||
>>> + (!ignore_pInt && ((pIntid < GICV3_LPI_INTID_START) ||
>>> + (pIntid > max_Intid)))) {
>>> + qemu_log_mask(LOG_GUEST_ERROR,
>>> + "%s: invalid interrupt translation table
>>> attributes "
>>> + "devid %d or icid %d or eventid %d or pIntid
>>> %d\n",
>>> + __func__, devid, icid, eventid, pIntid);
>>> + /*
>>> + * in this implementation,in case of error
>>> + * we ignore this command and move onto the next
>>> + * command in the queue
>>> + */
>>> + } else {
>>> + /* add ite entry to interrupt translation table */
>>> + itel = (dte_valid & VALID_MASK) | (GITS_TYPE_PHYSICAL <<
>>> + ITE_ENTRY_INTTYPE_SHIFT
>>> );
>>> +
>>> + if (ignore_pInt) {
>>> + itel |= (eventid << ITE_ENTRY_INTID_SHIFT);
>>> + } else {
>>> + itel |= (pIntid << ITE_ENTRY_INTID_SHIFT);
>>> + }
>>> + itel |= (int_spurious << ITE_ENTRY_INTSP_SHIFT);
>>> + iteh |= icid;
>>> +
>>> + res = update_ite(s, eventid, dte, itel, iteh);
>>> + }
>>> +
>>> + return res;
>>> +}
>>> +
>>> static MemTxResult update_cte(GICv3ITSState *s, uint16_t icid,
>>> bool valid,
>>> uint64_t rdbase)
>>> {
>>> @@ -295,8 +611,10 @@ static void process_cmdq(GICv3ITSState *s)
>>>
>>> switch (cmd) {
>>> case GITS_CMD_INT:
>>> + res = process_int(s, data, cq_offset, INT);
>>> break;
>>> case GITS_CMD_CLEAR:
>>> + res = process_int(s, data, cq_offset, CLEAR);
>>> break;
>>> case GITS_CMD_SYNC:
>>> /*
>>> @@ -313,10 +631,13 @@ static void process_cmdq(GICv3ITSState *s)
>>> res = process_mapc(s, cq_offset);
>>> break;
>>> case GITS_CMD_MAPTI:
>>> + res = process_mapti(s, data, cq_offset, false);
>>> break;
>>> case GITS_CMD_MAPI:
>>> + res = process_mapti(s, data, cq_offset, true);
>>> break;
>>> case GITS_CMD_DISCARD:
>>> + res = process_int(s, data, cq_offset, DISCARD);
>>> break;
>>> default:
>>> break;
>>> @@ -472,7 +793,20 @@ static MemTxResult
>>> gicv3_its_translation_write(void *opaque, hwaddr offset,
>>> uint64_t data,
>>> unsigned size,
>>> MemTxAttrs attrs)
>>> {
>>> + GICv3ITSState *s = (GICv3ITSState *)opaque;
>>> MemTxResult result = MEMTX_OK;
>>> + uint32_t devid = 0;
>>> +
>>> + switch (offset) {
>>> + case GITS_TRANSLATER:
>>> + if (s->ctlr & ITS_CTLR_ENABLED) {
>>> + devid = attrs.requester_id;
>>> + result = process_int(s, data, devid, NONE);
>>> + }
>>> + break;
>>> + default:
>>> + break;
>>> + }
>>>
>>> return result;
>>> }
>>> diff --git a/hw/intc/gicv3_internal.h b/hw/intc/gicv3_internal.h
>>> index 0932a30560..ce45cd0ef6 100644
>>> --- a/hw/intc/gicv3_internal.h
>>> +++ b/hw/intc/gicv3_internal.h
>>> @@ -324,6 +324,13 @@ FIELD(MAPC, RDBASE, 16, 32)
>>> #define ITTADDR_MASK ((1ULL << ITTADDR_LENGTH) - 1)
>>> #define SIZE_MASK 0x1f
>>>
>>> +/* MAPI command fields */
>>> +#define EVENTID_MASK ((1ULL << 32) - 1)
>>> +
>>> +/* MAPTI command fields */
>>> +#define pINTID_OFFSET 32
>>> +#define pINTID_MASK ((1ULL << 32) - 1)
>>> +
>>> #define VALID_SHIFT 63
>>> #define VALID_MASK 1ULL
>>>
>>> @@ -344,6 +351,11 @@ FIELD(MAPC, RDBASE, 16, 32)
>>> * vPEID = 16 bits
>>> */
>>> #define ITS_ITT_ENTRY_SIZE 0xC
>>> +#define ITE_ENTRY_INTTYPE_SHIFT 1
>>> +#define ITE_ENTRY_INTID_SHIFT 2
>>> +#define ITE_ENTRY_INTID_MASK ((1ULL << 24) - 1)
>>> +#define ITE_ENTRY_INTSP_SHIFT 26
>>> +#define ITE_ENTRY_ICID_MASK ((1ULL << 16) - 1)
>>>
>>> /* 16 bits EventId */
>>> #define ITS_IDBITS GICD_TYPER_IDBITS
>>> diff --git a/include/hw/intc/arm_gicv3_common.h
>>> b/include/hw/intc/arm_gicv3_common.h
>>> index 1fd5cedbbd..0715b0bc2a 100644
>>> --- a/include/hw/intc/arm_gicv3_common.h
>>> +++ b/include/hw/intc/arm_gicv3_common.h
>>> @@ -36,6 +36,8 @@
>>> #define GICV3_MAXIRQ 1020
>>> #define GICV3_MAXSPI (GICV3_MAXIRQ - GIC_INTERNAL)
>>>
>>> +#define GICV3_LPI_INTID_START 8192
>>> +
>>> #define GICV3_REDIST_SIZE 0x20000
>>>
>>> /* Number of SGI target-list bits */
>>>
>> Thanks
>>
>> Eric
>>
>
Thanks
Eric
