Linux cryptographic layer development
 help / color / mirror / Atom feed
From: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
To: herbert@gondor.apana.org.au
Cc: linux-crypto@vger.kernel.org, qat-linux@intel.com,
	marco.chiappero@intel.com,
	Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Subject: [PATCH 09/24] crypto: qat - handle retries due to collisions in adf_iov_putmsg()
Date: Wed, 10 Nov 2021 20:52:02 +0000	[thread overview]
Message-ID: <20211110205217.99903-10-giovanni.cabiddu@intel.com> (raw)
In-Reply-To: <20211110205217.99903-1-giovanni.cabiddu@intel.com>

From: Marco Chiappero <marco.chiappero@intel.com>

Rework __adf_iov_putmsg() to handle retries due to collisions
internally, removing the need for an external retry loop.
The functions __adf_iov_putmsg() and adf_iov_putmsg() have been merged
together maintaining the adf_iov_putmsg() name.

This will allow to use this function only for GEN2 devices, since
collision are peculiar of this generation and therefore should be
confined to the actual implementation of the transport/medium access.

Note that now adf_iov_putmsg() will retry to send a message only in case
of collisions and will now fail if an ACK is not received from the
remote function.

Signed-off-by: Marco Chiappero <marco.chiappero@intel.com>
Co-developed-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Signed-off-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
---
 drivers/crypto/qat/qat_common/adf_pf2vf_msg.c | 52 +++++++------------
 1 file changed, 19 insertions(+), 33 deletions(-)

diff --git a/drivers/crypto/qat/qat_common/adf_pf2vf_msg.c b/drivers/crypto/qat/qat_common/adf_pf2vf_msg.c
index 201744825e23..d98e3639c9d2 100644
--- a/drivers/crypto/qat/qat_common/adf_pf2vf_msg.c
+++ b/drivers/crypto/qat/qat_common/adf_pf2vf_msg.c
@@ -14,7 +14,7 @@
 					 ADF_PFVF_MSG_ACK_MAX_RETRY + \
 					 ADF_PFVF_MSG_COLLISION_DETECT_DELAY)
 
-static int __adf_iov_putmsg(struct adf_accel_dev *accel_dev, u32 msg, u8 vf_nr)
+static int adf_iov_putmsg(struct adf_accel_dev *accel_dev, u32 msg, u8 vf_nr)
 {
 	struct adf_accel_pci *pci_info = &accel_dev->accel_pci_dev;
 	struct adf_hw_device_data *hw_data = accel_dev->hw_device;
@@ -24,8 +24,9 @@ static int __adf_iov_putmsg(struct adf_accel_dev *accel_dev, u32 msg, u8 vf_nr)
 	u32 local_in_use_mask, local_in_use_pattern;
 	u32 remote_in_use_mask, remote_in_use_pattern;
 	struct mutex *lock;	/* lock preventing concurrent acces of CSR */
+	unsigned int retries = ADF_PFVF_MSG_MAX_RETRIES;
 	u32 int_bit;
-	int ret = 0;
+	int ret;
 
 	if (accel_dev->is_vf) {
 		pf2vf_offset = hw_data->get_pf2vf_offset(0);
@@ -45,20 +46,22 @@ static int __adf_iov_putmsg(struct adf_accel_dev *accel_dev, u32 msg, u8 vf_nr)
 		int_bit = ADF_PF2VF_INT;
 	}
 
+	msg &= ~local_in_use_mask;
+	msg |= local_in_use_pattern;
+
 	mutex_lock(lock);
 
+start:
+	ret = 0;
+
 	/* Check if the PFVF CSR is in use by remote function */
 	val = ADF_CSR_RD(pmisc_bar_addr, pf2vf_offset);
 	if ((val & remote_in_use_mask) == remote_in_use_pattern) {
 		dev_dbg(&GET_DEV(accel_dev),
 			"PFVF CSR in use by remote function\n");
-		ret = -EBUSY;
-		goto out;
+		goto retry;
 	}
 
-	msg &= ~local_in_use_mask;
-	msg |= local_in_use_pattern;
-
 	/* Attempt to get ownership of the PFVF CSR */
 	ADF_CSR_WR(pmisc_bar_addr, pf2vf_offset, msg | int_bit);
 
@@ -77,8 +80,7 @@ static int __adf_iov_putmsg(struct adf_accel_dev *accel_dev, u32 msg, u8 vf_nr)
 	if (val != msg) {
 		dev_dbg(&GET_DEV(accel_dev),
 			"Collision - PFVF CSR overwritten by remote function\n");
-		ret = -EIO;
-		goto out;
+		goto retry;
 	}
 
 	/* Finished with the PFVF CSR; relinquish it and leave msg in CSR */
@@ -86,31 +88,15 @@ static int __adf_iov_putmsg(struct adf_accel_dev *accel_dev, u32 msg, u8 vf_nr)
 out:
 	mutex_unlock(lock);
 	return ret;
-}
 
-/**
- * adf_iov_putmsg() - send PFVF message
- * @accel_dev:  Pointer to acceleration device.
- * @msg:	Message to send
- * @vf_nr:	VF number to which the message will be sent if on PF, ignored
- *		otherwise
- *
- * Function sends a message through the PFVF channel
- *
- * Return: 0 on success, error code otherwise.
- */
-static int adf_iov_putmsg(struct adf_accel_dev *accel_dev, u32 msg, u8 vf_nr)
-{
-	u32 count = 0;
-	int ret;
-
-	do {
-		ret = __adf_iov_putmsg(accel_dev, msg, vf_nr);
-		if (ret)
-			msleep(ADF_PFVF_MSG_RETRY_DELAY);
-	} while (ret && (count++ < ADF_PFVF_MSG_MAX_RETRIES));
-
-	return ret;
+retry:
+	if (--retries) {
+		msleep(ADF_PFVF_MSG_RETRY_DELAY);
+		goto start;
+	} else {
+		ret = -EBUSY;
+		goto out;
+	}
 }
 
 /**
-- 
2.33.1


  parent reply	other threads:[~2021-11-10 20:52 UTC|newest]

Thread overview: 27+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-10 20:51 [PATCH 00/24] crypto: qat - PFVF refactoring Giovanni Cabiddu
2021-11-10 20:51 ` [PATCH 01/24] crypto: qat - fix undetected PFVF timeout in ACK loop Giovanni Cabiddu
2021-11-10 20:51 ` [PATCH 02/24] crypto: qat - refactor PF top half for PFVF Giovanni Cabiddu
2021-11-10 20:51 ` [PATCH 03/24] crypto: qat - move vf2pf interrupt helpers Giovanni Cabiddu
2021-11-10 20:51 ` [PATCH 04/24] crypto: qat - move VF message handler to adf_vf2pf_msg.c Giovanni Cabiddu
2021-11-10 20:51 ` [PATCH 05/24] crypto: qat - move interrupt code out of the PFVF handler Giovanni Cabiddu
2021-11-10 20:51 ` [PATCH 06/24] crypto: qat - change PFVF ACK behaviour Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 07/24] crypto: qat - re-enable interrupts for legacy PFVF messages Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 08/24] crypto: qat - split PFVF message decoding from handling Giovanni Cabiddu
2021-11-10 20:52 ` Giovanni Cabiddu [this message]
2021-11-10 20:52 ` [PATCH 10/24] crypto: qat - relocate PFVF PF related logic Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 11/24] crypto: qat - relocate PFVF VF " Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 12/24] crypto: qat - relocate PFVF disabled function Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 13/24] crypto: qat - add pfvf_ops Giovanni Cabiddu
2021-11-12  9:43   ` kernel test robot
2021-11-12 10:20     ` Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 14/24] crypto: qat - differentiate between pf2vf and vf2pf offset Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 15/24] crypto: qat - abstract PFVF send function Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 16/24] crypto: qat - abstract PFVF receive logic Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 17/24] crypto: qat - reorganize PFVF code Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 18/24] crypto: qat - reorganize PFVF protocol definitions Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 19/24] crypto: qat - use enums for PFVF protocol codes Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 20/24] crypto: qat - pass the PF2VF responses back to the callers Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 21/24] crypto: qat - refactor pfvf version request messages Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 22/24] crypto: qat - do not rely on min version Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 23/24] crypto: qat - fix VF IDs in PFVF log messages Giovanni Cabiddu
2021-11-10 20:52 ` [PATCH 24/24] crypto: qat - improve logging of PFVF messages Giovanni Cabiddu

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20211110205217.99903-10-giovanni.cabiddu@intel.com \
    --to=giovanni.cabiddu@intel.com \
    --cc=herbert@gondor.apana.org.au \
    --cc=linux-crypto@vger.kernel.org \
    --cc=marco.chiappero@intel.com \
    --cc=qat-linux@intel.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox