All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH v3 0/2] bus: mhi: Fix MHI on big endian architectures
@ 2021-06-29  3:53 Paul Davey
  2021-06-29  3:53 ` [PATCH v3 1/2] bus: mhi: Fix pm_state conversion to string Paul Davey
                   ` (2 more replies)
  0 siblings, 3 replies; 8+ messages in thread
From: Paul Davey @ 2021-06-29  3:53 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: Hemant Kumar, Bhaumik Bhatt, linux-arm-msm, linux-kernel, Paul Davey

I encountered some problems getting the MHI driver to work on an Octeon
3 platform these seem to all be related to endianness issues.  The modem
interface appears to require the DMA structures to be in little endian,
however the MHI core driver was assembling them in native endianness.

Using little endian explicitly allows the interface to function as
expected.

Changes in v3:
  - removed change of doorbell helper functions db_val type from
    dma_addr_t to __le64 favouring doing conversion only when writing to
    context wp fields.  
Changes in v2:
  - use __fls instead of find_last_bit in pm_state conversion patch as
    requested by Hemant Kumar <hemantk@codeaurora.org>

Paul Davey (2):
  bus: mhi: Fix pm_state conversion to string
  bus: mhi: Fix MHI DMA structure endianness

 drivers/bus/mhi/core/debugfs.c  |  26 +++----
 drivers/bus/mhi/core/init.c     |  43 ++++++------
 drivers/bus/mhi/core/internal.h | 119 ++++++++++++++++----------------
 drivers/bus/mhi/core/main.c     |  22 +++---
 drivers/bus/mhi/core/pm.c       |   4 +-
 5 files changed, 109 insertions(+), 105 deletions(-)

-- 
2.32.0


^ permalink raw reply	[flat|nested] 8+ messages in thread

* [PATCH v3 1/2] bus: mhi: Fix pm_state conversion to string
  2021-06-29  3:53 [PATCH v3 0/2] bus: mhi: Fix MHI on big endian architectures Paul Davey
@ 2021-06-29  3:53 ` Paul Davey
  2021-07-01  1:50   ` Hemant Kumar
  2021-07-16  5:35   ` Manivannan Sadhasivam
  2021-06-29  3:53 ` [PATCH v3 2/2] bus: mhi: Fix MHI DMA structure endianness Paul Davey
  2021-07-16  5:43 ` [PATCH v3 0/2] bus: mhi: Fix MHI on big endian architectures Manivannan Sadhasivam
  2 siblings, 2 replies; 8+ messages in thread
From: Paul Davey @ 2021-06-29  3:53 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: Hemant Kumar, Bhaumik Bhatt, linux-arm-msm, linux-kernel, Paul Davey

On big endian architectures the mhi debugfs files which report pm state
give "Invalid State" for all states.  This is caused by using
find_last_bit which takes an unsigned long* while the state is passed in
as an enum mhi_pm_state which will be of int size.

Fix by using __fls to pass the value of state instead of find_last_bit.

Signed-off-by: Paul Davey <paul.davey@alliedtelesis.co.nz>
---
 drivers/bus/mhi/core/init.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
index c81b377fca8f..0706eb046f2a 100644
--- a/drivers/bus/mhi/core/init.c
+++ b/drivers/bus/mhi/core/init.c
@@ -79,9 +79,12 @@ static const char * const mhi_pm_state_str[] = {
 
 const char *to_mhi_pm_state_str(enum mhi_pm_state state)
 {
-	int index = find_last_bit((unsigned long *)&state, 32);
+	int index;
 
-	if (index >= ARRAY_SIZE(mhi_pm_state_str))
+	if (state)
+		index = __fls(state);
+
+	if (!state || index >= ARRAY_SIZE(mhi_pm_state_str))
 		return "Invalid State";
 
 	return mhi_pm_state_str[index];
-- 
2.32.0


^ permalink raw reply related	[flat|nested] 8+ messages in thread

* [PATCH v3 2/2] bus: mhi: Fix MHI DMA structure endianness
  2021-06-29  3:53 [PATCH v3 0/2] bus: mhi: Fix MHI on big endian architectures Paul Davey
  2021-06-29  3:53 ` [PATCH v3 1/2] bus: mhi: Fix pm_state conversion to string Paul Davey
@ 2021-06-29  3:53 ` Paul Davey
  2021-07-01  1:52   ` Hemant Kumar
  2021-07-16  5:43 ` [PATCH v3 0/2] bus: mhi: Fix MHI on big endian architectures Manivannan Sadhasivam
  2 siblings, 1 reply; 8+ messages in thread
From: Paul Davey @ 2021-06-29  3:53 UTC (permalink / raw)
  To: Manivannan Sadhasivam
  Cc: Hemant Kumar, Bhaumik Bhatt, linux-arm-msm, linux-kernel, Paul Davey

The MHI driver does not work on big endian architectures.  The
controller never transitions into mission mode.  This appears to be due
to the modem device expecting the various contexts and transfer rings to
have fields in little endian order in memory, but the driver constructs
them in native endianness.

Fix MHI event, channel and command contexts and TRE handling macros to
use explicit conversion to little endian.  Mark fields in relevant
structures as little endian to document this requirement.

Signed-off-by: Paul Davey <paul.davey@alliedtelesis.co.nz>
---
Removed Tested-by as patch has changed somewhat from v2.

 drivers/bus/mhi/core/debugfs.c  |  26 +++----
 drivers/bus/mhi/core/init.c     |  36 +++++-----
 drivers/bus/mhi/core/internal.h | 119 ++++++++++++++++----------------
 drivers/bus/mhi/core/main.c     |  22 +++---
 drivers/bus/mhi/core/pm.c       |   4 +-
 5 files changed, 104 insertions(+), 103 deletions(-)

diff --git a/drivers/bus/mhi/core/debugfs.c b/drivers/bus/mhi/core/debugfs.c
index 858d7516410b..d818586c229d 100644
--- a/drivers/bus/mhi/core/debugfs.c
+++ b/drivers/bus/mhi/core/debugfs.c
@@ -60,16 +60,16 @@ static int mhi_debugfs_events_show(struct seq_file *m, void *d)
 		}
 
 		seq_printf(m, "Index: %d intmod count: %lu time: %lu",
-			   i, (er_ctxt->intmod & EV_CTX_INTMODC_MASK) >>
+			   i, (le32_to_cpu(er_ctxt->intmod) & EV_CTX_INTMODC_MASK) >>
 			   EV_CTX_INTMODC_SHIFT,
-			   (er_ctxt->intmod & EV_CTX_INTMODT_MASK) >>
+			   (le32_to_cpu(er_ctxt->intmod) & EV_CTX_INTMODT_MASK) >>
 			   EV_CTX_INTMODT_SHIFT);
 
-		seq_printf(m, " base: 0x%0llx len: 0x%llx", er_ctxt->rbase,
-			   er_ctxt->rlen);
+		seq_printf(m, " base: 0x%0llx len: 0x%llx", le64_to_cpu(er_ctxt->rbase),
+			   le64_to_cpu(er_ctxt->rlen));
 
-		seq_printf(m, " rp: 0x%llx wp: 0x%llx", er_ctxt->rp,
-			   er_ctxt->wp);
+		seq_printf(m, " rp: 0x%llx wp: 0x%llx", le64_to_cpu(er_ctxt->rp),
+			   le64_to_cpu(er_ctxt->wp));
 
 		seq_printf(m, " local rp: 0x%pK db: 0x%pad\n", ring->rp,
 			   &mhi_event->db_cfg.db_val);
@@ -106,18 +106,18 @@ static int mhi_debugfs_channels_show(struct seq_file *m, void *d)
 
 		seq_printf(m,
 			   "%s(%u) state: 0x%lx brstmode: 0x%lx pollcfg: 0x%lx",
-			   mhi_chan->name, mhi_chan->chan, (chan_ctxt->chcfg &
+			   mhi_chan->name, mhi_chan->chan, (le32_to_cpu(chan_ctxt->chcfg) &
 			   CHAN_CTX_CHSTATE_MASK) >> CHAN_CTX_CHSTATE_SHIFT,
-			   (chan_ctxt->chcfg & CHAN_CTX_BRSTMODE_MASK) >>
-			   CHAN_CTX_BRSTMODE_SHIFT, (chan_ctxt->chcfg &
+			   (le32_to_cpu(chan_ctxt->chcfg) & CHAN_CTX_BRSTMODE_MASK) >>
+			   CHAN_CTX_BRSTMODE_SHIFT, (le32_to_cpu(chan_ctxt->chcfg) &
 			   CHAN_CTX_POLLCFG_MASK) >> CHAN_CTX_POLLCFG_SHIFT);
 
-		seq_printf(m, " type: 0x%x event ring: %u", chan_ctxt->chtype,
-			   chan_ctxt->erindex);
+		seq_printf(m, " type: 0x%x event ring: %u", le32_to_cpu(chan_ctxt->chtype),
+			   le32_to_cpu(chan_ctxt->erindex));
 
 		seq_printf(m, " base: 0x%llx len: 0x%llx rp: 0x%llx wp: 0x%llx",
-			   chan_ctxt->rbase, chan_ctxt->rlen, chan_ctxt->rp,
-			   chan_ctxt->wp);
+			   le64_to_cpu(chan_ctxt->rbase), le64_to_cpu(chan_ctxt->rlen),
+			   le64_to_cpu(chan_ctxt->rp), le64_to_cpu(chan_ctxt->wp));
 
 		seq_printf(m, " local rp: 0x%pK local wp: 0x%pK db: 0x%pad\n",
 			   ring->rp, ring->wp,
diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
index 0706eb046f2a..77962829941a 100644
--- a/drivers/bus/mhi/core/init.c
+++ b/drivers/bus/mhi/core/init.c
@@ -293,17 +293,17 @@ int mhi_init_dev_ctxt(struct mhi_controller *mhi_cntrl)
 		if (mhi_chan->offload_ch)
 			continue;
 
-		tmp = chan_ctxt->chcfg;
+		tmp = le32_to_cpu(chan_ctxt->chcfg);
 		tmp &= ~CHAN_CTX_CHSTATE_MASK;
 		tmp |= (MHI_CH_STATE_DISABLED << CHAN_CTX_CHSTATE_SHIFT);
 		tmp &= ~CHAN_CTX_BRSTMODE_MASK;
 		tmp |= (mhi_chan->db_cfg.brstmode << CHAN_CTX_BRSTMODE_SHIFT);
 		tmp &= ~CHAN_CTX_POLLCFG_MASK;
 		tmp |= (mhi_chan->db_cfg.pollcfg << CHAN_CTX_POLLCFG_SHIFT);
-		chan_ctxt->chcfg = tmp;
+		chan_ctxt->chcfg = cpu_to_le32(tmp);
 
-		chan_ctxt->chtype = mhi_chan->type;
-		chan_ctxt->erindex = mhi_chan->er_index;
+		chan_ctxt->chtype = cpu_to_le32(mhi_chan->type);
+		chan_ctxt->erindex = cpu_to_le32(mhi_chan->er_index);
 
 		mhi_chan->ch_state = MHI_CH_STATE_DISABLED;
 		mhi_chan->tre_ring.db_addr = (void __iomem *)&chan_ctxt->wp;
@@ -328,14 +328,14 @@ int mhi_init_dev_ctxt(struct mhi_controller *mhi_cntrl)
 		if (mhi_event->offload_ev)
 			continue;
 
-		tmp = er_ctxt->intmod;
+		tmp = le32_to_cpu(er_ctxt->intmod);
 		tmp &= ~EV_CTX_INTMODC_MASK;
 		tmp &= ~EV_CTX_INTMODT_MASK;
 		tmp |= (mhi_event->intmod << EV_CTX_INTMODT_SHIFT);
-		er_ctxt->intmod = tmp;
+		er_ctxt->intmod = cpu_to_le32(tmp);
 
-		er_ctxt->ertype = MHI_ER_TYPE_VALID;
-		er_ctxt->msivec = mhi_event->irq;
+		er_ctxt->ertype = cpu_to_le32(MHI_ER_TYPE_VALID);
+		er_ctxt->msivec = cpu_to_le32(mhi_event->irq);
 		mhi_event->db_cfg.db_mode = true;
 
 		ring->el_size = sizeof(struct mhi_tre);
@@ -349,9 +349,9 @@ int mhi_init_dev_ctxt(struct mhi_controller *mhi_cntrl)
 		 * ring is empty
 		 */
 		ring->rp = ring->wp = ring->base;
-		er_ctxt->rbase = ring->iommu_base;
+		er_ctxt->rbase = cpu_to_le64(ring->iommu_base);
 		er_ctxt->rp = er_ctxt->wp = er_ctxt->rbase;
-		er_ctxt->rlen = ring->len;
+		er_ctxt->rlen = cpu_to_le64(ring->len);
 		ring->ctxt_wp = &er_ctxt->wp;
 	}
 
@@ -378,9 +378,9 @@ int mhi_init_dev_ctxt(struct mhi_controller *mhi_cntrl)
 			goto error_alloc_cmd;
 
 		ring->rp = ring->wp = ring->base;
-		cmd_ctxt->rbase = ring->iommu_base;
+		cmd_ctxt->rbase = cpu_to_le64(ring->iommu_base);
 		cmd_ctxt->rp = cmd_ctxt->wp = cmd_ctxt->rbase;
-		cmd_ctxt->rlen = ring->len;
+		cmd_ctxt->rlen = cpu_to_le64(ring->len);
 		ring->ctxt_wp = &cmd_ctxt->wp;
 	}
 
@@ -581,10 +581,10 @@ void mhi_deinit_chan_ctxt(struct mhi_controller *mhi_cntrl,
 	chan_ctxt->rp = 0;
 	chan_ctxt->wp = 0;
 
-	tmp = chan_ctxt->chcfg;
+	tmp = le32_to_cpu(chan_ctxt->chcfg);
 	tmp &= ~CHAN_CTX_CHSTATE_MASK;
 	tmp |= (MHI_CH_STATE_DISABLED << CHAN_CTX_CHSTATE_SHIFT);
-	chan_ctxt->chcfg = tmp;
+	chan_ctxt->chcfg = cpu_to_le32(tmp);
 
 	/* Update to all cores */
 	smp_wmb();
@@ -618,14 +618,14 @@ int mhi_init_chan_ctxt(struct mhi_controller *mhi_cntrl,
 		return -ENOMEM;
 	}
 
-	tmp = chan_ctxt->chcfg;
+	tmp = le32_to_cpu(chan_ctxt->chcfg);
 	tmp &= ~CHAN_CTX_CHSTATE_MASK;
 	tmp |= (MHI_CH_STATE_ENABLED << CHAN_CTX_CHSTATE_SHIFT);
-	chan_ctxt->chcfg = tmp;
+	chan_ctxt->chcfg = cpu_to_le32(tmp);
 
-	chan_ctxt->rbase = tre_ring->iommu_base;
+	chan_ctxt->rbase = cpu_to_le64(tre_ring->iommu_base);
 	chan_ctxt->rp = chan_ctxt->wp = chan_ctxt->rbase;
-	chan_ctxt->rlen = tre_ring->len;
+	chan_ctxt->rlen = cpu_to_le64(tre_ring->len);
 	tre_ring->ctxt_wp = &chan_ctxt->wp;
 
 	tre_ring->rp = tre_ring->wp = tre_ring->base;
diff --git a/drivers/bus/mhi/core/internal.h b/drivers/bus/mhi/core/internal.h
index 672052fe3b44..3bb353f678b9 100644
--- a/drivers/bus/mhi/core/internal.h
+++ b/drivers/bus/mhi/core/internal.h
@@ -209,14 +209,14 @@ extern struct bus_type mhi_bus_type;
 #define EV_CTX_INTMODT_MASK GENMASK(31, 16)
 #define EV_CTX_INTMODT_SHIFT 16
 struct mhi_event_ctxt {
-	__u32 intmod;
-	__u32 ertype;
-	__u32 msivec;
-
-	__u64 rbase __packed __aligned(4);
-	__u64 rlen __packed __aligned(4);
-	__u64 rp __packed __aligned(4);
-	__u64 wp __packed __aligned(4);
+	__le32 intmod;
+	__le32 ertype;
+	__le32 msivec;
+
+	__le64 rbase __packed __aligned(4);
+	__le64 rlen __packed __aligned(4);
+	__le64 rp __packed __aligned(4);
+	__le64 wp __packed __aligned(4);
 };
 
 #define CHAN_CTX_CHSTATE_MASK GENMASK(7, 0)
@@ -227,25 +227,25 @@ struct mhi_event_ctxt {
 #define CHAN_CTX_POLLCFG_SHIFT 10
 #define CHAN_CTX_RESERVED_MASK GENMASK(31, 16)
 struct mhi_chan_ctxt {
-	__u32 chcfg;
-	__u32 chtype;
-	__u32 erindex;
-
-	__u64 rbase __packed __aligned(4);
-	__u64 rlen __packed __aligned(4);
-	__u64 rp __packed __aligned(4);
-	__u64 wp __packed __aligned(4);
+	__le32 chcfg;
+	__le32 chtype;
+	__le32 erindex;
+
+	__le64 rbase __packed __aligned(4);
+	__le64 rlen __packed __aligned(4);
+	__le64 rp __packed __aligned(4);
+	__le64 wp __packed __aligned(4);
 };
 
 struct mhi_cmd_ctxt {
-	__u32 reserved0;
-	__u32 reserved1;
-	__u32 reserved2;
-
-	__u64 rbase __packed __aligned(4);
-	__u64 rlen __packed __aligned(4);
-	__u64 rp __packed __aligned(4);
-	__u64 wp __packed __aligned(4);
+	__le32 reserved0;
+	__le32 reserved1;
+	__le32 reserved2;
+
+	__le64 rbase __packed __aligned(4);
+	__le64 rlen __packed __aligned(4);
+	__le64 rp __packed __aligned(4);
+	__le64 wp __packed __aligned(4);
 };
 
 struct mhi_ctxt {
@@ -258,8 +258,8 @@ struct mhi_ctxt {
 };
 
 struct mhi_tre {
-	u64 ptr;
-	u32 dword[2];
+	__le64 ptr;
+	__le32 dword[2];
 };
 
 struct bhi_vec_entry {
@@ -277,57 +277,58 @@ enum mhi_cmd_type {
 /* No operation command */
 #define MHI_TRE_CMD_NOOP_PTR (0)
 #define MHI_TRE_CMD_NOOP_DWORD0 (0)
-#define MHI_TRE_CMD_NOOP_DWORD1 (MHI_CMD_NOP << 16)
+#define MHI_TRE_CMD_NOOP_DWORD1 (cpu_to_le32(MHI_CMD_NOP << 16))
 
 /* Channel reset command */
 #define MHI_TRE_CMD_RESET_PTR (0)
 #define MHI_TRE_CMD_RESET_DWORD0 (0)
-#define MHI_TRE_CMD_RESET_DWORD1(chid) ((chid << 24) | \
-					(MHI_CMD_RESET_CHAN << 16))
+#define MHI_TRE_CMD_RESET_DWORD1(chid) (cpu_to_le32((chid << 24) | \
+					(MHI_CMD_RESET_CHAN << 16)))
 
 /* Channel stop command */
 #define MHI_TRE_CMD_STOP_PTR (0)
 #define MHI_TRE_CMD_STOP_DWORD0 (0)
-#define MHI_TRE_CMD_STOP_DWORD1(chid) ((chid << 24) | \
-				       (MHI_CMD_STOP_CHAN << 16))
+#define MHI_TRE_CMD_STOP_DWORD1(chid) (cpu_to_le32((chid << 24) | \
+				       (MHI_CMD_STOP_CHAN << 16)))
 
 /* Channel start command */
 #define MHI_TRE_CMD_START_PTR (0)
 #define MHI_TRE_CMD_START_DWORD0 (0)
-#define MHI_TRE_CMD_START_DWORD1(chid) ((chid << 24) | \
-					(MHI_CMD_START_CHAN << 16))
+#define MHI_TRE_CMD_START_DWORD1(chid) (cpu_to_le32((chid << 24) | \
+					(MHI_CMD_START_CHAN << 16)))
 
-#define MHI_TRE_GET_CMD_CHID(tre) (((tre)->dword[1] >> 24) & 0xFF)
-#define MHI_TRE_GET_CMD_TYPE(tre) (((tre)->dword[1] >> 16) & 0xFF)
+#define MHI_TRE_GET_DWORD(tre, word) (le32_to_cpu((tre)->dword[(word)]))
+#define MHI_TRE_GET_CMD_CHID(tre) ((MHI_TRE_GET_DWORD(tre, 1) >> 24) & 0xFF)
+#define MHI_TRE_GET_CMD_TYPE(tre) ((MHI_TRE_GET_DWORD(tre, 1) >> 16) & 0xFF)
 
 /* Event descriptor macros */
-#define MHI_TRE_EV_PTR(ptr) (ptr)
-#define MHI_TRE_EV_DWORD0(code, len) ((code << 24) | len)
-#define MHI_TRE_EV_DWORD1(chid, type) ((chid << 24) | (type << 16))
-#define MHI_TRE_GET_EV_PTR(tre) ((tre)->ptr)
-#define MHI_TRE_GET_EV_CODE(tre) (((tre)->dword[0] >> 24) & 0xFF)
-#define MHI_TRE_GET_EV_LEN(tre) ((tre)->dword[0] & 0xFFFF)
-#define MHI_TRE_GET_EV_CHID(tre) (((tre)->dword[1] >> 24) & 0xFF)
-#define MHI_TRE_GET_EV_TYPE(tre) (((tre)->dword[1] >> 16) & 0xFF)
-#define MHI_TRE_GET_EV_STATE(tre) (((tre)->dword[0] >> 24) & 0xFF)
-#define MHI_TRE_GET_EV_EXECENV(tre) (((tre)->dword[0] >> 24) & 0xFF)
-#define MHI_TRE_GET_EV_SEQ(tre) ((tre)->dword[0])
-#define MHI_TRE_GET_EV_TIME(tre) ((tre)->ptr)
-#define MHI_TRE_GET_EV_COOKIE(tre) lower_32_bits((tre)->ptr)
-#define MHI_TRE_GET_EV_VEID(tre) (((tre)->dword[0] >> 16) & 0xFF)
-#define MHI_TRE_GET_EV_LINKSPEED(tre) (((tre)->dword[1] >> 24) & 0xFF)
-#define MHI_TRE_GET_EV_LINKWIDTH(tre) ((tre)->dword[0] & 0xFF)
+#define MHI_TRE_EV_PTR(ptr) (cpu_to_le64(ptr))
+#define MHI_TRE_EV_DWORD0(code, len) (cpu_to_le32((code << 24) | len))
+#define MHI_TRE_EV_DWORD1(chid, type) (cpu_to_le32((chid << 24) | (type << 16)))
+#define MHI_TRE_GET_EV_PTR(tre) (le64_to_cpu((tre)->ptr))
+#define MHI_TRE_GET_EV_CODE(tre) ((MHI_TRE_GET_DWORD(tre, 0) >> 24) & 0xFF)
+#define MHI_TRE_GET_EV_LEN(tre) (MHI_TRE_GET_DWORD(tre, 0) & 0xFFFF)
+#define MHI_TRE_GET_EV_CHID(tre) ((MHI_TRE_GET_DWORD(tre, 1) >> 24) & 0xFF)
+#define MHI_TRE_GET_EV_TYPE(tre) ((MHI_TRE_GET_DWORD(tre, 1) >> 16) & 0xFF)
+#define MHI_TRE_GET_EV_STATE(tre) ((MHI_TRE_GET_DWORD(tre, 0) >> 24) & 0xFF)
+#define MHI_TRE_GET_EV_EXECENV(tre) ((MHI_TRE_GET_DWORD(tre, 0) >> 24) & 0xFF)
+#define MHI_TRE_GET_EV_SEQ(tre) MHI_TRE_GET_DWORD(tre, 0)
+#define MHI_TRE_GET_EV_TIME(tre) (MHI_TRE_GET_EV_PTR(tre))
+#define MHI_TRE_GET_EV_COOKIE(tre) lower_32_bits(MHI_TRE_GET_EV_PTR(tre))
+#define MHI_TRE_GET_EV_VEID(tre) ((MHI_TRE_GET_DWORD(tre, 0) >> 16) & 0xFF)
+#define MHI_TRE_GET_EV_LINKSPEED(tre) ((MHI_TRE_GET_DWORD(tre, 1) >> 24) & 0xFF)
+#define MHI_TRE_GET_EV_LINKWIDTH(tre) (MHI_TRE_GET_DWORD(tre, 0) & 0xFF)
 
 /* Transfer descriptor macros */
-#define MHI_TRE_DATA_PTR(ptr) (ptr)
-#define MHI_TRE_DATA_DWORD0(len) (len & MHI_MAX_MTU)
-#define MHI_TRE_DATA_DWORD1(bei, ieot, ieob, chain) ((2 << 16) | (bei << 10) \
-	| (ieot << 9) | (ieob << 8) | chain)
+#define MHI_TRE_DATA_PTR(ptr) (cpu_to_le64(ptr))
+#define MHI_TRE_DATA_DWORD0(len) (cpu_to_le32(len & MHI_MAX_MTU))
+#define MHI_TRE_DATA_DWORD1(bei, ieot, ieob, chain) (cpu_to_le32((2 << 16) | (bei << 10) \
+	| (ieot << 9) | (ieob << 8) | chain))
 
 /* RSC transfer descriptor macros */
-#define MHI_RSCTRE_DATA_PTR(ptr, len) (((u64)len << 48) | ptr)
-#define MHI_RSCTRE_DATA_DWORD0(cookie) (cookie)
-#define MHI_RSCTRE_DATA_DWORD1 (MHI_PKT_TYPE_COALESCING << 16)
+#define MHI_RSCTRE_DATA_PTR(ptr, len) (cpu_to_le64(((u64)len << 48) | ptr))
+#define MHI_RSCTRE_DATA_DWORD0(cookie) (cpu_to_le32(cookie))
+#define MHI_RSCTRE_DATA_DWORD1 (cpu_to_le32(MHI_PKT_TYPE_COALESCING << 16))
 
 enum mhi_pkt_type {
 	MHI_PKT_TYPE_INVALID = 0x0,
@@ -499,7 +500,7 @@ struct state_transition {
 struct mhi_ring {
 	dma_addr_t dma_handle;
 	dma_addr_t iommu_base;
-	u64 *ctxt_wp; /* point to ctxt wp */
+	__le64 *ctxt_wp; /* point to ctxt wp */
 	void *pre_aligned;
 	void *base;
 	void *rp;
diff --git a/drivers/bus/mhi/core/main.c b/drivers/bus/mhi/core/main.c
index 8ac73f9e92a6..507401cd30df 100644
--- a/drivers/bus/mhi/core/main.c
+++ b/drivers/bus/mhi/core/main.c
@@ -114,7 +114,7 @@ void mhi_ring_er_db(struct mhi_event *mhi_event)
 	struct mhi_ring *ring = &mhi_event->ring;
 
 	mhi_event->db_cfg.process_db(mhi_event->mhi_cntrl, &mhi_event->db_cfg,
-				     ring->db_addr, *ring->ctxt_wp);
+				     ring->db_addr, le64_to_cpu(*ring->ctxt_wp));
 }
 
 void mhi_ring_cmd_db(struct mhi_controller *mhi_cntrl, struct mhi_cmd *mhi_cmd)
@@ -123,7 +123,7 @@ void mhi_ring_cmd_db(struct mhi_controller *mhi_cntrl, struct mhi_cmd *mhi_cmd)
 	struct mhi_ring *ring = &mhi_cmd->ring;
 
 	db = ring->iommu_base + (ring->wp - ring->base);
-	*ring->ctxt_wp = db;
+	*ring->ctxt_wp = cpu_to_le64(db);
 	mhi_write_db(mhi_cntrl, ring->db_addr, db);
 }
 
@@ -140,7 +140,7 @@ void mhi_ring_chan_db(struct mhi_controller *mhi_cntrl,
 	 * before letting h/w know there is new element to fetch.
 	 */
 	dma_wmb();
-	*ring->ctxt_wp = db;
+	*ring->ctxt_wp = cpu_to_le64(db);
 
 	mhi_chan->db_cfg.process_db(mhi_cntrl, &mhi_chan->db_cfg,
 				    ring->db_addr, db);
@@ -432,7 +432,7 @@ irqreturn_t mhi_irq_handler(int irq_number, void *dev)
 	struct mhi_event_ctxt *er_ctxt =
 		&mhi_cntrl->mhi_ctxt->er_ctxt[mhi_event->er_index];
 	struct mhi_ring *ev_ring = &mhi_event->ring;
-	dma_addr_t ptr = er_ctxt->rp;
+	dma_addr_t ptr = le64_to_cpu(er_ctxt->rp);
 	void *dev_rp;
 
 	if (!is_valid_ring_ptr(ev_ring, ptr)) {
@@ -537,14 +537,14 @@ static void mhi_recycle_ev_ring_element(struct mhi_controller *mhi_cntrl,
 
 	/* Update the WP */
 	ring->wp += ring->el_size;
-	ctxt_wp = *ring->ctxt_wp + ring->el_size;
+	ctxt_wp = le64_to_cpu(*ring->ctxt_wp) + ring->el_size;
 
 	if (ring->wp >= (ring->base + ring->len)) {
 		ring->wp = ring->base;
 		ctxt_wp = ring->iommu_base;
 	}
 
-	*ring->ctxt_wp = ctxt_wp;
+	*ring->ctxt_wp = cpu_to_le64(ctxt_wp);
 
 	/* Update the RP */
 	ring->rp += ring->el_size;
@@ -799,7 +799,7 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl,
 	struct device *dev = &mhi_cntrl->mhi_dev->dev;
 	u32 chan;
 	int count = 0;
-	dma_addr_t ptr = er_ctxt->rp;
+	dma_addr_t ptr = le64_to_cpu(er_ctxt->rp);
 
 	/*
 	 * This is a quick check to avoid unnecessary event processing
@@ -938,7 +938,7 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl,
 		mhi_recycle_ev_ring_element(mhi_cntrl, ev_ring);
 		local_rp = ev_ring->rp;
 
-		ptr = er_ctxt->rp;
+		ptr = le64_to_cpu(er_ctxt->rp);
 		if (!is_valid_ring_ptr(ev_ring, ptr)) {
 			dev_err(&mhi_cntrl->mhi_dev->dev,
 				"Event ring rp points outside of the event ring\n");
@@ -968,7 +968,7 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl,
 	int count = 0;
 	u32 chan;
 	struct mhi_chan *mhi_chan;
-	dma_addr_t ptr = er_ctxt->rp;
+	dma_addr_t ptr = le64_to_cpu(er_ctxt->rp);
 
 	if (unlikely(MHI_EVENT_ACCESS_INVALID(mhi_cntrl->pm_state)))
 		return -EIO;
@@ -1009,7 +1009,7 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl,
 		mhi_recycle_ev_ring_element(mhi_cntrl, ev_ring);
 		local_rp = ev_ring->rp;
 
-		ptr = er_ctxt->rp;
+		ptr = le64_to_cpu(er_ctxt->rp);
 		if (!is_valid_ring_ptr(ev_ring, ptr)) {
 			dev_err(&mhi_cntrl->mhi_dev->dev,
 				"Event ring rp points outside of the event ring\n");
@@ -1531,7 +1531,7 @@ static void mhi_mark_stale_events(struct mhi_controller *mhi_cntrl,
 	/* mark all stale events related to channel as STALE event */
 	spin_lock_irqsave(&mhi_event->lock, flags);
 
-	ptr = er_ctxt->rp;
+	ptr = le64_to_cpu(er_ctxt->rp);
 	if (!is_valid_ring_ptr(ev_ring, ptr)) {
 		dev_err(&mhi_cntrl->mhi_dev->dev,
 			"Event ring rp points outside of the event ring\n");
diff --git a/drivers/bus/mhi/core/pm.c b/drivers/bus/mhi/core/pm.c
index bbf6cd04861e..66a79ad07ff2 100644
--- a/drivers/bus/mhi/core/pm.c
+++ b/drivers/bus/mhi/core/pm.c
@@ -218,7 +218,7 @@ int mhi_ready_state_transition(struct mhi_controller *mhi_cntrl)
 			continue;
 
 		ring->wp = ring->base + ring->len - ring->el_size;
-		*ring->ctxt_wp = ring->iommu_base + ring->len - ring->el_size;
+		*ring->ctxt_wp = cpu_to_le64(ring->iommu_base + ring->len - ring->el_size);
 		/* Update all cores */
 		smp_wmb();
 
@@ -420,7 +420,7 @@ static int mhi_pm_mission_mode_transition(struct mhi_controller *mhi_cntrl)
 			continue;
 
 		ring->wp = ring->base + ring->len - ring->el_size;
-		*ring->ctxt_wp = ring->iommu_base + ring->len - ring->el_size;
+		*ring->ctxt_wp = cpu_to_le64(ring->iommu_base + ring->len - ring->el_size);
 		/* Update to all cores */
 		smp_wmb();
 
-- 
2.32.0


^ permalink raw reply related	[flat|nested] 8+ messages in thread

* Re: [PATCH v3 1/2] bus: mhi: Fix pm_state conversion to string
  2021-06-29  3:53 ` [PATCH v3 1/2] bus: mhi: Fix pm_state conversion to string Paul Davey
@ 2021-07-01  1:50   ` Hemant Kumar
  2021-07-16  5:35   ` Manivannan Sadhasivam
  1 sibling, 0 replies; 8+ messages in thread
From: Hemant Kumar @ 2021-07-01  1:50 UTC (permalink / raw)
  To: Paul Davey, Manivannan Sadhasivam
  Cc: Bhaumik Bhatt, linux-arm-msm, linux-kernel

On Tue, 2021-06-29 at 15:53 +1200, Paul Davey wrote:
> On big endian architectures the mhi debugfs files which report pm
> state
> give "Invalid State" for all states.  This is caused by using
> find_last_bit which takes an unsigned long* while the state is passed
> in
> as an enum mhi_pm_state which will be of int size.
> 
> Fix by using __fls to pass the value of state instead of
> find_last_bit.
> 
> Signed-off-by: Paul Davey <paul.davey@alliedtelesis.co.nz>

Reviewed-by: Hemant Kumar <hemantk@codeaurora.org>
-- 
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project


^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH v3 2/2] bus: mhi: Fix MHI DMA structure endianness
  2021-06-29  3:53 ` [PATCH v3 2/2] bus: mhi: Fix MHI DMA structure endianness Paul Davey
@ 2021-07-01  1:52   ` Hemant Kumar
  2021-07-16  0:35     ` Bhaumik Bhatt
  0 siblings, 1 reply; 8+ messages in thread
From: Hemant Kumar @ 2021-07-01  1:52 UTC (permalink / raw)
  To: Paul Davey, Manivannan Sadhasivam
  Cc: Bhaumik Bhatt, linux-arm-msm, linux-kernel

Hi Paul,
On Tue, 2021-06-29 at 15:53 +1200, Paul Davey wrote:
> The MHI driver does not work on big endian architectures.  The
> controller never transitions into mission mode.  This appears to be
> due
> to the modem device expecting the various contexts and transfer rings
> to
> have fields in little endian order in memory, but the driver
> constructs
> them in native endianness.
> 
> Fix MHI event, channel and command contexts and TRE handling macros
> to
> use explicit conversion to little endian.  Mark fields in relevant
> structures as little endian to document this requirement.
> 
> Signed-off-by: Paul Davey <paul.davey@alliedtelesis.co.nz>
> ---
> Removed Tested-by as patch has changed somewhat from v2.
Bhaumik is going to help testing V2 on x86 platform. Change looks good
to me. Will ack it once testing is done.
> [..]
> 

Thanks,
Hemant
-- 
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project


^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH v3 2/2] bus: mhi: Fix MHI DMA structure endianness
  2021-07-01  1:52   ` Hemant Kumar
@ 2021-07-16  0:35     ` Bhaumik Bhatt
  0 siblings, 0 replies; 8+ messages in thread
From: Bhaumik Bhatt @ 2021-07-16  0:35 UTC (permalink / raw)
  To: Hemant Kumar, Paul Davey
  Cc: Manivannan Sadhasivam, linux-arm-msm, linux-kernel,
	hemantk=codeaurora.org

Hi Paul,

On 2021-06-30 06:52 PM, Hemant Kumar wrote:
> Hi Paul,
> On Tue, 2021-06-29 at 15:53 +1200, Paul Davey wrote:
>> The MHI driver does not work on big endian architectures.  The
>> controller never transitions into mission mode.  This appears to be
>> due
>> to the modem device expecting the various contexts and transfer rings
>> to
>> have fields in little endian order in memory, but the driver
>> constructs
>> them in native endianness.
>> 
>> Fix MHI event, channel and command contexts and TRE handling macros
>> to
>> use explicit conversion to little endian.  Mark fields in relevant
>> structures as little endian to document this requirement.
>> 
>> Signed-off-by: Paul Davey <paul.davey@alliedtelesis.co.nz>
>> ---
>> Removed Tested-by as patch has changed somewhat from v2.
> Bhaumik is going to help testing V2 on x86 platform. Change looks good
> to me. Will ack it once testing is done.
>>  [..]
>> 
> 
> Thanks,
> Hemant

Just following up on this one. I have been seeing problems with my setup 
and
will try to get to this test next week.

Thanks,
Bhaumik
---
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora 
Forum,
a Linux Foundation Collaborative Project

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH v3 1/2] bus: mhi: Fix pm_state conversion to string
  2021-06-29  3:53 ` [PATCH v3 1/2] bus: mhi: Fix pm_state conversion to string Paul Davey
  2021-07-01  1:50   ` Hemant Kumar
@ 2021-07-16  5:35   ` Manivannan Sadhasivam
  1 sibling, 0 replies; 8+ messages in thread
From: Manivannan Sadhasivam @ 2021-07-16  5:35 UTC (permalink / raw)
  To: Paul Davey; +Cc: Hemant Kumar, Bhaumik Bhatt, linux-arm-msm, linux-kernel

On Tue, Jun 29, 2021 at 03:53:56PM +1200, Paul Davey wrote:
> On big endian architectures the mhi debugfs files which report pm state
> give "Invalid State" for all states.  This is caused by using
> find_last_bit which takes an unsigned long* while the state is passed in
> as an enum mhi_pm_state which will be of int size.
> 
> Fix by using __fls to pass the value of state instead of find_last_bit.
> 
> Signed-off-by: Paul Davey <paul.davey@alliedtelesis.co.nz>

Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>

Thanks,
Mani

> ---
>  drivers/bus/mhi/core/init.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/bus/mhi/core/init.c b/drivers/bus/mhi/core/init.c
> index c81b377fca8f..0706eb046f2a 100644
> --- a/drivers/bus/mhi/core/init.c
> +++ b/drivers/bus/mhi/core/init.c
> @@ -79,9 +79,12 @@ static const char * const mhi_pm_state_str[] = {
>  
>  const char *to_mhi_pm_state_str(enum mhi_pm_state state)
>  {
> -	int index = find_last_bit((unsigned long *)&state, 32);
> +	int index;
>  
> -	if (index >= ARRAY_SIZE(mhi_pm_state_str))
> +	if (state)
> +		index = __fls(state);
> +
> +	if (!state || index >= ARRAY_SIZE(mhi_pm_state_str))
>  		return "Invalid State";
>  
>  	return mhi_pm_state_str[index];
> -- 
> 2.32.0
> 

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: [PATCH v3 0/2] bus: mhi: Fix MHI on big endian architectures
  2021-06-29  3:53 [PATCH v3 0/2] bus: mhi: Fix MHI on big endian architectures Paul Davey
  2021-06-29  3:53 ` [PATCH v3 1/2] bus: mhi: Fix pm_state conversion to string Paul Davey
  2021-06-29  3:53 ` [PATCH v3 2/2] bus: mhi: Fix MHI DMA structure endianness Paul Davey
@ 2021-07-16  5:43 ` Manivannan Sadhasivam
  2 siblings, 0 replies; 8+ messages in thread
From: Manivannan Sadhasivam @ 2021-07-16  5:43 UTC (permalink / raw)
  To: Paul Davey; +Cc: Hemant Kumar, Bhaumik Bhatt, linux-arm-msm, linux-kernel

On Tue, Jun 29, 2021 at 03:53:55PM +1200, Paul Davey wrote:
> I encountered some problems getting the MHI driver to work on an Octeon
> 3 platform these seem to all be related to endianness issues.  The modem
> interface appears to require the DMA structures to be in little endian,
> however the MHI core driver was assembling them in native endianness.
> 
> Using little endian explicitly allows the interface to function as
> expected.
> 

Please add fixes tag and Cc stable for all the patches.

Thanks,
Mani

> Changes in v3:
>   - removed change of doorbell helper functions db_val type from
>     dma_addr_t to __le64 favouring doing conversion only when writing to
>     context wp fields.  
> Changes in v2:
>   - use __fls instead of find_last_bit in pm_state conversion patch as
>     requested by Hemant Kumar <hemantk@codeaurora.org>
> 
> Paul Davey (2):
>   bus: mhi: Fix pm_state conversion to string
>   bus: mhi: Fix MHI DMA structure endianness
> 
>  drivers/bus/mhi/core/debugfs.c  |  26 +++----
>  drivers/bus/mhi/core/init.c     |  43 ++++++------
>  drivers/bus/mhi/core/internal.h | 119 ++++++++++++++++----------------
>  drivers/bus/mhi/core/main.c     |  22 +++---
>  drivers/bus/mhi/core/pm.c       |   4 +-
>  5 files changed, 109 insertions(+), 105 deletions(-)
> 
> -- 
> 2.32.0
> 

^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2021-07-16  5:43 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-06-29  3:53 [PATCH v3 0/2] bus: mhi: Fix MHI on big endian architectures Paul Davey
2021-06-29  3:53 ` [PATCH v3 1/2] bus: mhi: Fix pm_state conversion to string Paul Davey
2021-07-01  1:50   ` Hemant Kumar
2021-07-16  5:35   ` Manivannan Sadhasivam
2021-06-29  3:53 ` [PATCH v3 2/2] bus: mhi: Fix MHI DMA structure endianness Paul Davey
2021-07-01  1:52   ` Hemant Kumar
2021-07-16  0:35     ` Bhaumik Bhatt
2021-07-16  5:43 ` [PATCH v3 0/2] bus: mhi: Fix MHI on big endian architectures Manivannan Sadhasivam

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.