diff options
Diffstat (limited to 'src/import/generic/memory/lib/utils/mc/gen_mss_port.H')
-rw-r--r-- | src/import/generic/memory/lib/utils/mc/gen_mss_port.H | 698 |
1 files changed, 692 insertions, 6 deletions
diff --git a/src/import/generic/memory/lib/utils/mc/gen_mss_port.H b/src/import/generic/memory/lib/utils/mc/gen_mss_port.H index 01c704e89..435d93a50 100644 --- a/src/import/generic/memory/lib/utils/mc/gen_mss_port.H +++ b/src/import/generic/memory/lib/utils/mc/gen_mss_port.H @@ -5,7 +5,7 @@ /* */ /* OpenPOWER HostBoot Project */ /* */ -/* Contributors Listed Below - COPYRIGHT 2018,2019 */ +/* Contributors Listed Below - COPYRIGHT 2018,2020 */ /* [+] International Business Machines Corp. */ /* */ /* */ @@ -43,6 +43,9 @@ #include <generic/memory/lib/utils/mc/gen_mss_port_traits.H> #include <generic/memory/lib/utils/scom.H> #include <generic/memory/lib/utils/c_str.H> +#include <generic/memory/lib/ecc/ecc.H> +#include <generic/memory/lib/utils/mss_bad_bits.H> +#include <generic/memory/lib/utils/mss_rank.H> namespace mss { @@ -163,11 +166,10 @@ void get_bw_snapshot( const fapi2::buffer<uint64_t>& i_data, uint64_t& o_bw_snap /// @param[in] const ref to the fapi2::Target<fapi2::TargetType> /// @param[out] uint32_t* memory to store the value /// @return fapi2::ReturnCode - FAPI2_RC_SUCCESS iff get is OK -/// @note Mark store records from MPVD Lx -/// keyword +/// @note Mark store records from MPVD Lx keyword /// template< mss::mc_type MC = DEFAULT_MC_TYPE, fapi2::TargetType T> -inline fapi2::ReturnCode mvpd_fwms(const fapi2::Target< T>& i_target, uint32_t (&o_array)[MARK_STORE_COUNT]); +fapi2::ReturnCode mvpd_fwms(const fapi2::Target< T>& i_target, uint32_t (&o_array)[MARK_STORE_COUNT]); /// /// @brief Enable power management @@ -604,17 +606,571 @@ fapi_try_exit: return fapi2::current_err; } +/// +/// @brief Place a symbol mark in a Firmware Mark Store register +/// @tparam MC the memory controller type +/// @tparam T the fapi2 target type of the target +/// @tparam TT the class traits for the port +/// @param[in] i_target the DIMM target +/// @param[in] i_rank the rank +/// @param[in] i_dq the bad DQ bit +/// @return FAPI2_RC_SUCCESS if and only if ok +/// +template< mss::mc_type MC = DEFAULT_MC_TYPE, fapi2::TargetType T, typename TT = portTraits<MC> > +inline fapi2::ReturnCode place_symbol_mark(const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq) +{ + const auto& l_port = mss::find_target<TT::PORT_TYPE>(i_target); + const auto l_dimm_idx = mss::index(i_target); + const auto l_rank_idx = mss::index(i_rank); + + uint8_t l_galois = 0; + mss::mcbist::address l_addr; + + // For symbol marks, we set the appropriate Firmware Mark Store reg, with the symbol's + // Galois code, mark_type=SYMBOL, mark_region=MRANK, and the address of the DIMM+MRANK + // TODO RTC:165133 Remove static_cast once Galois API is updated to accept uint64_t input + FAPI_TRY( mss::ecc::dq_to_galois(static_cast<uint8_t>(i_dq), l_galois) ); + + l_addr.set_dimm(l_dimm_idx).set_master_rank(l_rank_idx); + + FAPI_INF("%s Setting firmware symbol mark on rank:%d dq:%d galois:0x%02x", + mss::c_str(i_target), i_rank, i_dq, l_galois); + FAPI_TRY( mss::ecc::set_fwms(l_port, i_rank, l_galois, + mss::ecc::fwms::mark_type::SYMBOL, + mss::ecc::fwms::mark_region::MRANK, + l_addr) ); + +fapi_try_exit: + return fapi2::current_err; +} + +/// +/// @brief Place a chip mark in a Hardware Mark Store register +/// @tparam MC the memory controller type +/// @tparam T the fapi2 target type of the target +/// @tparam TT the class traits for the port +/// @param[in] i_target the DIMM target +/// @param[in] i_rank the rank +/// @param[in] i_dq one of the bad DQ bits in the bad nibble +/// @return FAPI2_RC_SUCCESS if and only if ok +/// +template< mss::mc_type MC = DEFAULT_MC_TYPE, fapi2::TargetType T, typename TT = portTraits<MC> > +inline fapi2::ReturnCode place_chip_mark(const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq) +{ + const auto& l_port = mss::find_target<TT::PORT_TYPE>(i_target); + + uint8_t l_galois = 0; + uint8_t l_symbol = 0; + + // For chip marks, we set the appropriate Hardware Mark Store reg, with the Galois code + // of the first (smallest) symbol in the bad nibble, and both confirmed and exit1 bits set + FAPI_TRY( mss::ecc::dq_to_symbol(static_cast<uint8_t>(i_dq), l_symbol) ); + + // Round down to the nearest "nibble" to get the correct symbol, then get the Galois code for it + l_symbol = (l_symbol / BITS_PER_NIBBLE) * BITS_PER_NIBBLE; + FAPI_TRY( mss::ecc::symbol_to_galois(l_symbol, l_galois) ); + + FAPI_INF("%s Setting hardware (chip) mark on rank:%d galois:0x%02x", mss::c_str(i_target), i_rank, l_galois); + FAPI_TRY( mss::ecc::set_hwms(l_port, i_rank, l_galois) ); + +fapi_try_exit: + return fapi2::current_err; +} + +// Forward declaration for use in repair_state classes +template< fapi2::TargetType T > +class repair_state_machine; + +/// +/// @class mss::repair_state +/// @brief A class for keeping track of bad bit repair states in a repair_state_machine +/// @tparam T, the fapi2 target type of the DIMM +/// @note this is a base class +/// +template< fapi2::TargetType T > +class repair_state +{ + public: + /// @brief default contructor + repair_state() = default; + /// @brief default destructor + virtual ~repair_state() = default; + + /// + /// @brief Perform a repair for a single bad DQ bit in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq the DQ bit index + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + virtual fapi2::ReturnCode one_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) = 0; + + /// + /// @brief Perform a repair for multiple bad DQ bits in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq one of the bad DQ bit indexes + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + virtual fapi2::ReturnCode multiple_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) = 0; + + protected: + /// + /// @brief Set a new state in the repair state machine + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_state pointer to the new state to set + /// + inline void set_state(repair_state_machine<T>& io_machine, std::shared_ptr<repair_state<T>> i_state) + { + io_machine.update_state(i_state); + } +}; + +/// +/// @class mss::chip_and_symbol_mark +/// @brief repair_state class for when both a chip mark and a symbol mark have been used +/// @tparam T, the fapi2 target type of the DIMM +/// +template< fapi2::TargetType T > +class chip_and_symbol_mark : public repair_state<T> +{ + public: + /// @brief default contructor + chip_and_symbol_mark() = default; + /// @brief default destructor + ~chip_and_symbol_mark() = default; + + /// + /// @brief Perform a repair for a single bad DQ bit in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq the DQ bit index + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode one_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + // repairs exceeded + FAPI_TRY( io_repairs_exceeded.setBit(mss::index(i_target)) ); + FAPI_INF("%s Repairs exceeded (chip mark and symbol mark exist, plus one bad DQ) on rank:%d DQ:%d", + mss::c_str(i_target), i_rank, i_dq); + return fapi2::FAPI2_RC_SUCCESS; + fapi_try_exit: + return fapi2::current_err; + } + + /// + /// @brief Perform a repair for multiple bad DQ bits in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq one of the bad DQ bit indexes + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode multiple_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + // repairs exceeded + FAPI_TRY( io_repairs_exceeded.setBit(mss::index(i_target)) ); + FAPI_INF("%s Repairs exceeded (chip mark and symbol mark exist, plus one bad nibble) on rank:%d DQ:%d", + mss::c_str(i_target), i_rank, i_dq); + return fapi2::FAPI2_RC_SUCCESS; + fapi_try_exit: + return fapi2::current_err; + } +}; + +/// +/// @class mss::symbol_mark_plus_unrepaired_dq +/// @brief repair_state class for when only a symbol mark has been used, and one DQ bit remains unrepaired +/// @tparam T, the fapi2 target type of the DIMM +/// +template< fapi2::TargetType T > +class symbol_mark_plus_unrepaired_dq : public repair_state<T> +{ + public: + /// @brief default contructor + symbol_mark_plus_unrepaired_dq() = default; + /// @brief default destructor + ~symbol_mark_plus_unrepaired_dq() = default; + + /// + /// @brief Perform a repair for a single bad DQ bit in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq the DQ bit index + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode one_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + // repairs exceeded + FAPI_INF("%s Repairs exceeded (symbol mark and unrepaired DQ exist, plus bad DQ) on rank:%d DQ:%d", + mss::c_str(i_target), i_rank, i_dq); + FAPI_TRY( io_repairs_exceeded.setBit(mss::index(i_target)) ); + return fapi2::FAPI2_RC_SUCCESS; + fapi_try_exit: + return fapi2::current_err; + } + + /// + /// @brief Perform a repair for multiple bad DQ bits in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq one of the bad DQ bit indexes + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode multiple_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + // place a chip mark, but also repairs exceeded + FAPI_TRY( place_chip_mark(i_target, i_rank, i_dq) ); + FAPI_TRY( io_repairs_applied.setBit(i_rank) ); + FAPI_TRY( io_repairs_exceeded.setBit(mss::index(i_target)) ); + FAPI_INF("%s Repairs exceeded (symbol mark and unrepaired DQ exist, plus bad nibble) on rank:%d DQ:%d", + mss::c_str(i_target), i_rank, i_dq); + { + const auto new_state = std::make_shared<chip_and_symbol_mark<fapi2::TARGET_TYPE_DIMM>>(); + mss::repair_state<T>::set_state(io_machine, new_state); + } + fapi_try_exit: + return fapi2::current_err; + } + +}; + +/// +/// @class mss::symbol_mark_only +/// @brief repair_state class for when only a symbol mark has been used +/// @tparam T, the fapi2 target type of the DIMM +/// +template< fapi2::TargetType T > +class symbol_mark_only : public repair_state<T> +{ + public: + /// @brief default contructor + symbol_mark_only() = default; + /// @brief default destructor + ~symbol_mark_only() = default; + + /// + /// @brief Perform a repair for a single bad DQ bit in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq the DQ bit index + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode one_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + // leave an unrepaired DQ + const auto new_state = std::make_shared< symbol_mark_plus_unrepaired_dq<T> >(); + mss::repair_state<T>::set_state(io_machine, new_state); + return fapi2::FAPI2_RC_SUCCESS; + } + + /// + /// @brief Perform a repair for multiple bad DQ bits in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq one of the bad DQ bit indexes + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode multiple_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + // place a chip mark + FAPI_TRY( place_chip_mark(i_target, i_rank, i_dq) ); + FAPI_TRY( io_repairs_applied.setBit(i_rank) ); + { + const auto new_state = std::make_shared< chip_and_symbol_mark<T> >(); + mss::repair_state<T>::set_state(io_machine, new_state); + } + fapi_try_exit: + return fapi2::current_err; + } +}; + +/// +/// @class mss::chip_mark_only +/// @brief repair_state class for when only a chip mark has been used +/// @tparam T, the fapi2 target type of the DIMM +/// +template< fapi2::TargetType T > +class chip_mark_only : public repair_state<T> +{ + public: + /// @brief default contructor + chip_mark_only() = default; + /// @brief default destructor + ~chip_mark_only() = default; + + /// + /// @brief Perform a repair for a single bad DQ bit in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq the DQ bit index + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode one_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + // place a symbol mark + FAPI_TRY( place_symbol_mark(i_target, i_rank, i_dq) ); + FAPI_TRY( io_repairs_applied.setBit(i_rank) ); + { + const auto new_state = std::make_shared<chip_and_symbol_mark<fapi2::TARGET_TYPE_DIMM>>(); + mss::repair_state<T>::set_state(io_machine, new_state); + } + fapi_try_exit: + return fapi2::current_err; + } + + /// + /// @brief Perform a repair for multiple bad DQ bits in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq one of the bad DQ bit indexes + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode multiple_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + // repairs exceeded + FAPI_TRY( io_repairs_exceeded.setBit(mss::index(i_target)) ); + FAPI_INF("%s Repairs exceeded (chip mark exists, plus bad nibble) on rank:%d DQ:%d", + mss::c_str(i_target), i_rank, i_dq); + return fapi2::FAPI2_RC_SUCCESS; + fapi_try_exit: + return fapi2::current_err; + } +}; + +/// +/// @class mss::no_fails +/// @brief repair_state class for no fails (no marks applied) +/// @tparam T, the fapi2 target type of the DIMM +/// +template< fapi2::TargetType T > +class no_fails : public repair_state<T> +{ + public: + /// @brief default contructor + no_fails() = default; + /// @brief default destructor + ~no_fails() = default; + + /// + /// @brief Perform a repair for a single bad DQ bit in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq the DQ bit index + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode one_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + // place a symbol mark + FAPI_TRY( place_symbol_mark(i_target, i_rank, i_dq) ); + FAPI_TRY( io_repairs_applied.setBit(i_rank) ); + { + const auto new_state = std::make_shared<symbol_mark_only<fapi2::TARGET_TYPE_DIMM>>(); + mss::repair_state<T>::set_state(io_machine, new_state); + } + fapi_try_exit: + return fapi2::current_err; + } + + /// + /// @brief Perform a repair for multiple bad DQ bits in a nibble + /// @param[in,out] io_machine the repair state machine + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq one of the bad DQ bit indexes + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode multiple_bad_dq(repair_state_machine<T>& io_machine, + const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + // place a chip mark + FAPI_TRY( place_chip_mark(i_target, i_rank, i_dq) ); + FAPI_TRY( io_repairs_applied.setBit(i_rank) ); + { + const auto new_state = std::make_shared<chip_mark_only<fapi2::TARGET_TYPE_DIMM>>(); + mss::repair_state<T>::set_state(io_machine, new_state); + } + fapi_try_exit: + return fapi2::current_err; + } +}; + +/// +/// @class mss::repair_state_machine +/// @brief state machine class used in restore_repairs_helper +/// @tparam T, the fapi2 target type of the DIMM +/// +template< fapi2::TargetType T > +class repair_state_machine +{ + public: + /// @brief constructor + repair_state_machine() + : iv_repair_state(std::make_shared<no_fails<T>>()) {} + + /// @brief default destructor + ~repair_state_machine() = default; + + /// + /// @brief Perform a repair for a single bad DQ bit in a nibble + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq the DQ bit index + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repai:rs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode one_bad_dq(const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + FAPI_TRY( iv_repair_state->one_bad_dq(*this, i_target, i_rank, i_dq, io_repairs_applied, io_repairs_exceeded) ); + fapi_try_exit: + return fapi2::current_err; + } + + /// + /// @brief Perform a repair for multiple bad DQ bits in a nibble + /// @param[in] i_target the DIMM target + /// @param[in] i_rank the rank + /// @param[in] i_dq one of the bad DQ bit indexes + /// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied + /// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired + /// @return FAPI2_RC_SUCCESS if and only if ok + /// + fapi2::ReturnCode multiple_bad_dq(const fapi2::Target<T>& i_target, + const uint64_t i_rank, + const uint64_t i_dq, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) + { + FAPI_TRY( iv_repair_state->multiple_bad_dq(*this, i_target, i_rank, i_dq, io_repairs_applied, io_repairs_exceeded) ); + fapi_try_exit: + return fapi2::current_err; + } + + /// + /// @brief Update the state of the state machine + /// @param[in] i_state shared pointer to the new state + /// + void update_state(std::shared_ptr<repair_state<T>> i_state) + { + iv_repair_state = i_state; + } + + private: + std::shared_ptr<repair_state<T>> iv_repair_state; +}; /// @brief Get the attributes for the reorder queue setting /// @tparam MC the memory controller type -/// @tparam T, the mc +/// @tparam T, the fapi2 target type of the target /// @param[in] const ref to the mc target /// @param[out] uint8_t& reference to store the value /// @return fapi2::ReturnCode - FAPI2_RC_SUCCESS iff get is OK /// @note Contains the settings for write/read reorder queue /// template< mss::mc_type MC = DEFAULT_MC_TYPE, fapi2::TargetType T > -inline fapi2::ReturnCode reorder_queue_setting(const fapi2::Target<T>& i_target, uint8_t& o_value); +fapi2::ReturnCode reorder_queue_setting(const fapi2::Target<T>& i_target, uint8_t& o_value); /// /// @brief Resets the write/read reorder queue values - needs to be called after MCBIST execution @@ -644,6 +1200,136 @@ fapi_try_exit: return fapi2::current_err; } +/// +/// @brief Convert a bitmap from the BAD_DQ_BITMAP attribute to a vector of bad DQ indexes +/// @param[in] i_bad_bits an 8-bit bitmap of bad bits +/// @param[in] i_nibble which nibble of the bitmap to convert +/// @return std::vector of DQ bits marked as bad in the bitmap +/// +inline std::vector<uint64_t> bad_bit_helper(const uint8_t i_bad_bits, const size_t i_nibble) +{ + std::vector<uint64_t> l_output; + fapi2::buffer<uint8_t> l_bit_buffer(i_bad_bits); + + const size_t l_start = (i_nibble == 0) ? 0 : mss::conversions::BITS_PER_NIBBLE; + + for (size_t l_offset = 0; l_offset < mss::conversions::BITS_PER_NIBBLE; ++l_offset) + { + const size_t l_position_tmp = l_start + l_offset; + + if (l_bit_buffer.getBit(l_position_tmp)) + { + l_output.push_back(l_position_tmp); + } + } + + return l_output; +} + +// TODO: RTC: 157753 tparam R can be pulled from an PORT trait once we have it +// +/// @brief Restore symbol and chip marks according to BAD_DQ_BITMAP attribute, helper function for unit testing +/// @tparam MC the memory controller type +/// @tparam T, the fapi2 target type of the DIMM (derived) +/// @tparam R the maximum rank per DIMM +/// @tparam B the number of bytes per rank in the bad_dq_bitmap attribute +/// @param[in] i_target A target representing a DIMM +/// @param[in] i_bad_bits the bad bits values from the VPD, for the specified DIMM +/// @param[in,out] io_repairs_applied 8-bit mask, where a bit set means that rank had repairs applied +/// @param[in,out] io_repairs_exceeded 2-bit mask, where a bit set means that DIMM had more bad bits than could be repaired +/// @return FAPI2_RC_SUCCESS if and only if ok +/// +template< mss::mc_type MC = DEFAULT_MC_TYPE, fapi2::TargetType T, uint64_t R, uint64_t B > +inline fapi2::ReturnCode restore_repairs_helper( const fapi2::Target<T>& i_target, + const uint8_t i_bad_bits[R][B], + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) +{ + FAPI_INF("%s Restore repair marks from bad DQ data", mss::c_str(i_target)); + + using MCT = mss::mcbistMCTraits<MC>; + using MT = mss::mcbistTraits<MC, MCT::MC_TARGET_TYPE>; + + std::vector<uint64_t> l_ranks; + const auto l_dimm_idx = index(i_target); + + // gets all of the ranks to loop over + FAPI_TRY( mss::rank::ranks_on_dimm_helper<MC>(i_target, l_ranks) ); + + // loop through ranks + for (const auto l_rank : l_ranks) + { + const auto l_rank_idx = index(l_rank); + + repair_state_machine<fapi2::TARGET_TYPE_DIMM> l_machine; + + for (uint64_t l_byte = 0; l_byte < (MT::MAX_DQ_NIBBLES / mss::conversions::NIBBLES_PER_BYTE); ++l_byte) + { + for (size_t l_nibble = 0; l_nibble < mss::conversions::NIBBLES_PER_BYTE; ++l_nibble) + { + const auto l_bad_dq_vector = bad_bit_helper(i_bad_bits[l_rank_idx][l_byte], l_nibble); + FAPI_DBG("Total bad bits on DIMM:%d rank:%d nibble%d: %d", + l_dimm_idx, l_rank, (l_byte * NIBBLES_PER_BYTE) + l_nibble, l_bad_dq_vector.size()); + + // apply repairs and update repair machine state + // if there are no bad bits (l_bad_dq_vector.size() == 0) no action is necessary + if (l_bad_dq_vector.size() == 1) + { + // l_bad_dq_vector is per byte, so multiply up to get the bad dq's index + const uint64_t l_dq = l_bad_dq_vector[0] + (l_byte * BITS_PER_BYTE); + FAPI_TRY( l_machine.one_bad_dq(i_target, l_rank, l_dq, io_repairs_applied, io_repairs_exceeded) ); + } + else if (l_bad_dq_vector.size() > 1) + { + // l_bad_dq_vector is per byte, so multiply up to get the bad dq's index + const uint64_t l_dq = l_bad_dq_vector[0] + (l_byte * BITS_PER_BYTE); + FAPI_TRY( l_machine.multiple_bad_dq(i_target, l_rank, l_dq, io_repairs_applied, io_repairs_exceeded) ); + } + + // if repairs have been exceeded, we're done + if (io_repairs_exceeded.getBit(l_dimm_idx)) + { + FAPI_INF("Repairs exceeded on DIMM %s", c_str(i_target)); + return fapi2::FAPI2_RC_SUCCESS; + } + } // end loop through nibbles + } // end loop through bytes + } // end loop through ranks + +fapi_try_exit: + return fapi2::current_err; +} + +/// +/// @brief Restore symbol and chip marks according to BAD_DQ_BITMAP attribute +/// @tparam MC, the memory controller type +/// @tparam T, the fapi2 target type of the port (derived) +/// @param[in] i_target A target representing a port +/// @param[in,out] io_repairs_applied bit mask, where a bit set means a rank had repairs applied (bit0 = rank0, etc) +/// @param[in,out] io_repairs_exceeded bit mask, where a bit set means a DIMM had more bad bits than could be repaired (bit0 = DIMM0 etc) +/// @return FAPI2_RC_SUCCESS if and only if ok +/// +template< mss::mc_type MC = DEFAULT_MC_TYPE, fapi2::TargetType T > +inline fapi2::ReturnCode restore_repairs( const fapi2::Target<T>& i_target, + fapi2::buffer<uint8_t>& io_repairs_applied, + fapi2::buffer<uint8_t>& io_repairs_exceeded) +{ + uint8_t l_bad_bits[BAD_BITS_RANKS][BAD_DQ_BYTE_COUNT] = {}; + + io_repairs_applied = 0; + io_repairs_exceeded = 0; + + for (const auto& l_dimm : mss::find_targets<fapi2::TARGET_TYPE_DIMM>(i_target)) + { + FAPI_TRY( mss::get_bad_dq_bitmap<MC>(l_dimm, l_bad_bits) ); + + FAPI_TRY( (restore_repairs_helper<MC, fapi2::TARGET_TYPE_DIMM, BAD_BITS_RANKS, BAD_DQ_BYTE_COUNT>( + l_dimm, l_bad_bits, io_repairs_applied, io_repairs_exceeded)) ); + } + +fapi_try_exit: + return fapi2::current_err; +} }// ns mss |