Return-Path: From: "Savoy, Pavan" To: "Gustavo F. Padovan" CC: "linux-bluetooth@vger.kernel.org" , "marcel@holtmann.org" , "greg@kroah.com" , "linux-kernel@vger.kernel.org" Date: Fri, 8 Oct 2010 03:23:57 +0530 Subject: RE: [PATCH 1/2] drivers:bluetooth: TI_ST bluetooth driver Message-ID: <19F8576C6E063C45BE387C64729E739404AA21D25D@dbde02.ent.ti.com> References: <1286477237-7526-1-git-send-email-pavan_savoy@ti.com> <1286477237-7526-2-git-send-email-pavan_savoy@ti.com> <20101007184534.GB13602@vigoh> In-Reply-To: <20101007184534.GB13602@vigoh> Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 List-ID: Gustavo / Marcel, =20 > -----Original Message----- > From: Gustavo F. Padovan [mailto:pao@profusion.mobi] On Behalf Of Gustavo= F. > Padovan > Sent: Thursday, October 07, 2010 1:46 PM > To: Savoy, Pavan > Cc: linux-bluetooth@vger.kernel.org; marcel@holtmann.org; greg@kroah.com; > linux-kernel@vger.kernel.org > Subject: Re: [PATCH 1/2] drivers:bluetooth: TI_ST bluetooth driver >=20 > Hi Pavan, >=20 > Change the commit subject to "Bluetooth: TI_ST bluetooth driver" >=20 > * pavan_savoy@ti.com [2010-10-07 14:47:16 -0400]: >=20 > > From: Pavan Savoy > > > > This is the bluetooth protocol driver for the TI WiLink7 chipsets. > > Texas Instrument's WiLink chipsets combine wireless technologies > > like BT, FM, GPS and WLAN onto a single chip. > > > > This Bluetooth driver works on top of the TI_ST shared transport > > line discipline driver which also allows other drivers like > > FM V4L2 and GPS character driver to make use of the same UART interface= . > > > > Signed-off-by: Pavan Savoy > > --- > > drivers/bluetooth/bt_ti.c | 489 > +++++++++++++++++++++++++++++++++++++++++++++ > > 1 files changed, 489 insertions(+), 0 deletions(-) > > create mode 100644 drivers/bluetooth/bt_ti.c >=20 > We don't have filename with bt_.. in drivers/bluetooth/. Maybe ti_st.c > should be a better name, or something like that. >=20 > > > > diff --git a/drivers/bluetooth/bt_ti.c b/drivers/bluetooth/bt_ti.c > > new file mode 100644 > > index 0000000..dffbb56 > > --- /dev/null > > +++ b/drivers/bluetooth/bt_ti.c > > @@ -0,0 +1,489 @@ > > +/* > > + * Texas Instrument's Bluetooth Driver For Shared Transport. > > + * > > + * Bluetooth Driver acts as interface between HCI CORE and > > + * TI Shared Transport Layer. > > + * > > + * Copyright (C) 2009-2010 Texas Instruments > > + * Author: Raja Mani > > + * Pavan Savoy > > + * > > + * This program is free software; you can redistribute it and/or modi= fy > > + * it under the terms of the GNU General Public License version 2 as > > + * published by the Free Software Foundation. > > + * > > + * This program is distributed in the hope that it will be useful, > > + * but WITHOUT ANY WARRANTY; without even the implied warranty of > > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the > > + * GNU General Public License for more details. > > + * > > + * You should have received a copy of the GNU General Public License > > + * along with this program; if not, write to the Free Software > > + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-13= 07 > USA > > + * > > + */ > > + > > +#include > > +#include > > +#include > > + > > +#include > > + > > +/* Bluetooth Driver Version */ > > +#define VERSION "1.0" > > + > > +/* Defines number of seconds to wait for reg completion > > + * callback getting called from ST (in case,registration > > + * with ST returns PENDING status) > > + */ > > +#define BT_REGISTER_TIMEOUT 6000 /* 6 sec */ > > + > > +/* BT driver's local status */ > > +#define BT_DRV_RUNNING 0 > > +#define BT_ST_REGISTERED 1 > > + > > +/** > > + * struct hci_st - BT driver operation structure > > + * @hdev: hci device pointer which binds to bt driver > > + * @flags: used locally,to maintain various BT driver status > > + * @streg_cbdata: to hold ST registration callback status > > + * @st_write: write function pointer of ST driver > > + * @wait_for_btdrv_reg_completion - completion sync between hci_st_ope= n > > + * and hci_st_registration_completion_cb. > > + */ > > +struct hci_st { > > + struct hci_dev *hdev; > > + unsigned long flags; > > + char streg_cbdata; > > + long (*st_write) (struct sk_buff *); > > + struct completion wait_for_btdrv_reg_completion; > > +}; > > + > > +static int reset; > > + > > +/* Increments HCI counters based on pocket ID (cmd,acl,sco) */ > > +static inline void hci_st_tx_complete(struct hci_st *hst, int pkt_type= ) > > +{ > > + struct hci_dev *hdev; > > + hdev =3D hst->hdev; > > + > > + /* Update HCI stat counters */ > > + switch (pkt_type) { > > + case HCI_COMMAND_PKT: > > + hdev->stat.cmd_tx++; > > + break; > > + > > + case HCI_ACLDATA_PKT: > > + hdev->stat.acl_tx++; > > + break; > > + > > + case HCI_SCODATA_PKT: > > + hdev->stat.cmd_tx++; >=20 > it should be sco_tx here. >=20 > > + break; > > + } > > +} > > + > > +/* ------- Interfaces to Shared Transport ------ */ > > + > > +/* Called by ST layer to indicate protocol registration completion > > + * status.hci_st_open() function will wait for signal from this > > + * API when st_register() function returns ST_PENDING. > > + */ > > +static void hci_st_registration_completion_cb(void *priv_data, char da= ta) >=20 > That is not the hci layer, so rename this function (and the others) to > something that reflect where they are really doing. >=20 > > +{ > > + struct hci_st *lhst =3D (struct hci_st *)priv_data; > > + /* hci_st_open() function needs value of 'data' to know > > + * the registration status(success/fail),So have a back > > + * up of it. > > + */ > > + lhst->streg_cbdata =3D data; > > + > > + /* Got a feedback from ST for BT driver registration > > + * request.Wackup hci_st_open() function to continue > > + * it's open operation. > > + */ > > + complete(&lhst->wait_for_btdrv_reg_completion); > > +} > > + > > +/* Called by Shared Transport layer when receive data is > > + * available */ > > +static long hci_st_receive(void *priv_data, struct sk_buff *skb) > > +{ > > + int err; > > + int len; >=20 > you can put err and len in the same line. >=20 > > + struct hci_st *lhst =3D (struct hci_st *)priv_data; > > + > > + err =3D 0; > > + len =3D 0; >=20 > and no need to set them to 0 here. >=20 > > + > > + if (skb =3D=3D NULL) { > > + BT_ERR("Invalid SKB received from ST"); > > + return -EFAULT; > > + } >=20 > We need a empty line here. >=20 > > + if (!lhst) { > > + kfree_skb(skb); > > + BT_ERR("Invalid hci_st memory,freeing SKB"); > > + return -EFAULT; > > + } >=20 > And also here. Check the rest of the code for similar issues. >=20 > > + if (!test_bit(BT_DRV_RUNNING, &lhst->flags)) { > > + kfree_skb(skb); > > + BT_ERR("Device is not running,freeing SKB"); > > + return -EINVAL; > > + } >=20 > If you are here, your device is running, right? Or am I missing > something? >=20 > > + > > + len =3D skb->len; > > + skb->dev =3D (struct net_device *)lhst->hdev; > > + > > + /* Forward skb to HCI CORE layer */ > > + err =3D hci_recv_frame(skb); > > + if (err) { > > + kfree_skb(skb); > > + BT_ERR("Unable to push skb to HCI CORE(%d),freeing SKB", > > + err); > > + return err; > > + } > > + lhst->hdev->stat.byte_rx +=3D len; >=20 > actually you even don't need len, just use skb->len >=20 > > + > > + return 0; > > +} > > + > > +/* ------- Interfaces to HCI layer ------ */ > > + > > +/* Called from HCI core to initialize the device */ > > +static int hci_st_open(struct hci_dev *hdev) > > +{ > > + static struct st_proto_s hci_st_proto; > > + unsigned long timeleft; > > + struct hci_st *hst; > > + int err; > > + err =3D 0; > > + > > + BT_DBG("%s %p", hdev->name, hdev); > > + hst =3D hdev->driver_data; > > + > > + /* Populate BT driver info required by ST */ > > + memset(&hci_st_proto, 0, sizeof(hci_st_proto)); > > + > > + /* BT driver ID */ > > + hci_st_proto.type =3D ST_BT; > > + > > + /* Receive function which called from ST */ > > + hci_st_proto.recv =3D hci_st_receive; > > + > > + /* Packet match function may used in future */ > > + hci_st_proto.match_packet =3D NULL; >=20 > It is already NULL, you dua a memset. >=20 > > + > > + /* Callback to be called when registration is pending */ > > + hci_st_proto.reg_complete_cb =3D hci_st_registration_completion_cb; > > + > > + /* This is write function pointer of ST. BT driver will make use of > this > > + * for sending any packets to chip. ST will assign and give to us, so > > + * make it as NULL */ > > + hci_st_proto.write =3D NULL; >=20 > Same here. >=20 > > + > > + /* send in the hst to be received at registration complete callback > > + * and during st's receive > > + */ > > + hci_st_proto.priv_data =3D hst; > > + > > + /* Register with ST layer */ > > + err =3D st_register(&hci_st_proto); > > + if (err =3D=3D -EINPROGRESS) { > > + /* Prepare wait-for-completion handler data structures. > > + * Needed to syncronize this and st_registration_completion_cb() > > + * functions. > > + */ > > + init_completion(&hst->wait_for_btdrv_reg_completion); >=20 > I'm not liking that, but I'll leave for Marcel and others comment. Thanks for the comments, I will take care of them all, but I want to tell why I do this here. "st_register" function takes time. It takes about 2 seconds for us to download the firmware. (or longer based on UART baud-rate). This is intended to happen in the process context. Say hciconfig hci0 up takes a while longer than usual or if someone doing an ioctl(HCIDEVUP) and this is also the reason I have a callback, as and when firmware downloa= d completes, I call the Bluetooth driver's callback. and I await marcel's comment on this. > > + > > + /* Reset ST registration callback status flag , this value > > + * will be updated in hci_st_registration_completion_cb() > > + * function whenever it called from ST driver. > > + */ > > + hst->streg_cbdata =3D -EINPROGRESS; > > + > > + /* ST is busy with other protocol registration(may be busy with > > + * firmware download).So,Wait till the registration callback > > + * (passed as a argument to st_register() function) getting > > + * called from ST. > > + */ > > + BT_DBG(" %s waiting for reg completion signal from ST", > > + __func__); > > + > > + timeleft =3D > > + wait_for_completion_timeout > > + (&hst->wait_for_btdrv_reg_completion, > > + msecs_to_jiffies(BT_REGISTER_TIMEOUT)); > > + if (!timeleft) { > > + BT_ERR("Timeout(%d sec),didn't get reg" > > + "completion signal from ST", > > + BT_REGISTER_TIMEOUT / 1000); > > + return -ETIMEDOUT; > > + } > > + > > + /* Is ST registration callback called with ERROR value? */ > > + if (hst->streg_cbdata !=3D 0) { > > + BT_ERR("ST reg completion CB called with invalid" > > + "status %d", hst->streg_cbdata); > > + return -EAGAIN; > > + } > > + err =3D 0; > > + } else if (err =3D=3D -1) { >=20 > Use the proper error macro instead "-1" >=20 > > + BT_ERR("st_register failed %d", err); > > + return -EAGAIN; > > + } > > + > > + /* Do we have proper ST write function? */ > > + if (hci_st_proto.write !=3D NULL) { > > + /* We need this pointer for sending any Bluetooth pkts */ > > + hst->st_write =3D hci_st_proto.write; > > + } else { > > + BT_ERR("failed to get ST write func pointer"); > > + > > + /* Undo registration with ST */ > > + err =3D st_unregister(ST_BT); > > + if (err < 0) > > + BT_ERR("st_unregister failed %d", err); > > + > > + hst->st_write =3D NULL; > > + return -EAGAIN; > > + } > > + > > + /* Registration with ST layer is completed successfully, > > + * now chip is ready to accept commands from HCI CORE. > > + * Mark HCI Device flag as RUNNING > > + */ > > + set_bit(HCI_RUNNING, &hdev->flags); > > + > > + /* Registration with ST successful */ > > + set_bit(BT_ST_REGISTERED, &hst->flags); > > + > > + return err; > > +} > > + > > +/* Close device */ > > +static int hci_st_close(struct hci_dev *hdev) > > +{ > > + int err; > > + struct hci_st *hst; >=20 > Skip a line after declarations. >=20 > > + err =3D 0; >=20 > you can set err to 0 in the declaration if you really need that. >=20 > > + > > + hst =3D hdev->driver_data; > > + /* Unregister from ST layer */ > > + if (test_and_clear_bit(BT_ST_REGISTERED, &hst->flags)) { > > + err =3D st_unregister(ST_BT); > > + if (err !=3D 0) { > > + BT_ERR("st_unregister failed %d", err); > > + return -EBUSY; > > + } > > + } > > + > > + hst->st_write =3D NULL; > > + > > + /* ST layer would have moved chip to inactive state. > > + * So,clear HCI device RUNNING flag. > > + */ > > + if (!test_and_clear_bit(HCI_RUNNING, &hdev->flags)) > > + return 0; >=20 > Looks you are screwing up the flags here, if it fails on st_unregister() > and returns HCI_RUNNING should keep set? >=20 > > + > > + return err; >=20 > Rethink how you are doing error handling here, it should no be > complicated like that. >=20 > > +} > > + > > +/* Called from HCI CORE , Sends frames to Shared Transport */ > > +static int hci_st_send_frame(struct sk_buff *skb) > > +{ > > + struct hci_dev *hdev; > > + struct hci_st *hst; > > + long len; > > + > > + if (skb =3D=3D NULL) { > > + BT_ERR("Invalid skb received from HCI CORE"); > > + return -ENOMEM; > > + } > > + hdev =3D (struct hci_dev *)skb->dev; > > + if (!hdev) { > > + BT_ERR("SKB received for invalid HCI Device (hdev=3DNULL)"); > > + return -ENODEV; > > + } > > + if (!test_bit(HCI_RUNNING, &hdev->flags)) { > > + BT_ERR("Device is not running"); > > + return -EBUSY; > > + } > > + > > + hst =3D (struct hci_st *)hdev->driver_data; > > + > > + /* Prepend skb with frame type */ > > + memcpy(skb_push(skb, 1), &bt_cb(skb)->pkt_type, 1); > > + > > + BT_DBG(" %s: type %d len %d", hdev->name, bt_cb(skb)->pkt_type, > > + skb->len); > > + > > + /* Insert skb to shared transport layer's transmit queue. > > + * Freeing skb memory is taken care in shared transport layer, > > + * so don't free skb memory here. > > + */ > > + if (!hst->st_write) { > > + kfree_skb(skb); > > + BT_ERR(" Can't write to ST, st_write null?"); > > + return -EAGAIN; > > + } > > + len =3D hst->st_write(skb); > > + if (len < 0) { > > + /* Something went wrong in st write , free skb memory */ >=20 > IMHO we don't need comments like that, clearly we now that something > went wrong. >=20 > > + kfree_skb(skb); > > + BT_ERR(" ST write failed (%ld)", len); > > + return -EAGAIN; > > + } > > + > > + /* ST accepted our skb. So, Go ahead and do rest */ > > + hdev->stat.byte_tx +=3D len; > > + hci_st_tx_complete(hst, bt_cb(skb)->pkt_type); > > + > > + return 0; >=20 > goto might be better to handle error here. >=20 >=20 > -- > Gustavo F. Padovan > ProFUSION embedded systems - http://profusion.mobi