Return-Path: MIME-Version: 1.0 In-Reply-To: <20101116225418.GA15101@vigoh> References: <1289394446-14021-1-git-send-email-pavan_savoy@ti.com> <20101116225418.GA15101@vigoh> Date: Wed, 17 Nov 2010 11:04:59 +0530 Message-ID: Subject: Re: [PATCH v5] Bluetooth: btwilink driver From: Pavan Savoy To: "Gustavo F. Padovan" Cc: marcel@holtmann.org, linux-bluetooth@vger.kernel.org, linux-kernel@vger.kernel.org Content-Type: text/plain; charset=UTF-8 List-ID: Gustavo, On Wed, Nov 17, 2010 at 4:24 AM, Gustavo F. Padovan wrote: > Hi Pavan, > > * pavan_savoy@ti.com [2010-11-10 08:07:26 -0500]: > >> From: Pavan Savoy >> >> Marcel, >> >> Thanks for the comments... >> This patch contains, >> v5 comments :- >> declaration and assiging of variables and private data fixed up. >> proper casting. >> removed redundant un-necessary checks in send_frame. >> HCI_RUNNING fixes in terms of test_and_set/clear bit instead of set and >> clear. >> removed redundant checks for hdev, skb being NULL. >> removed module_param of reset, also WiLink don't need HCI_RESET anyways. >> removed ti_st_register_dev function and functionality moved to _probe. >> module_init/exit function names fixed up. >> >> stat byte counter increments and tx_complete is similar to hci_ldisc. >> Also I have not implemented the flush routine, since the functionality >> which needs to be done in flush routine is done in the underlying driver >> which is the shared transport driver and moreover the btwilink driver by >> itself doesn't maintains queue or data relevant to transport, so nothing >> to do. >> >> And Yes, I have verified this driver with multiple up/down reset on >> hci0. >> Also I generally test a2dp/ftp to verify large data transfers. > > Before re-submit a patch you have to fix all issues reported by the > reviewer or reply to patch thread why do you think you right and don't > want to change that. That is not happening with your patches, you are > not fixing all the stuff before re-submiting and it is not the first > time. =C2=A0If you do it right we can review it fast and your code goes > earlier to mainline. > You should also answer the questions first =C2=A0like "Where is the > ti_st_proto.write coming from?" You just ignored the > review and submitted a new patch. This is the reason, I tend to keep the patch comments in the mail. I have mentioned here the 1. comments I have taken care of, 2. few comments which I don't understand why it is like that and which are not taken care of. Also the question on ti_st_proto.write, I have answered it twice in mail, once to you and once to marcel, for more clarity I have even added it in the code comments, Please have a look @ line >> + /* ti_st_proto.write is filled up by the underlying shared >> + * transport driver upon registration >> + */ As to why this function is not EXPORT_SYMBOL and just an function pointer, Yes I had it as function pointer - But since something like _read is callback from the protocol driver perspective - to maintain uniformity I have this as function pointer. (Note: comments to other drivers which are based on ST driver intended read/write to be pointers which register/unregister to be EXPORTs). Ok, apart from this there was an open question as why I am checking for only 2 error conditions, it is because the underlying driver only can send those 2 errors and nothing else (st_register has few errors it can throw...) I understand my problem with test_and_set_bit, I will correct it, But I still feel I do nothing critical before test/set so as to return error if already opened - But will correct it and do it at the beginning ... Also what should I be doing regarding the deferred stats update? I checked up hci_ldisc which is what this code is pretty much based on, and see that it does pretty much the same thing .... hci_uart_tx_complete is called after the tty->ops->write correct ? I also now understand the issue in module_init regarding the platform_driver registration and will fix that. Thanks, Pavan >> >> v4 comments :- >> module init now returns what platform_driver_register returns. >> type casting of void* private data has been removed >> >> v3 comments :- >> Lizardo, >> I have taken care of most of the comments you had. >> Have re-wrote some of the code commenting you've mentioned. >> Thanks for the comments, >> >> The other few like -EPERM for platform driver registration is to keep >> it similar to other drivers, type casting is maintained just to feel saf= e >> and have style similar to other drivers. >> BT_WILINK in Kconfig is similar to BT_MRVL. >> I hope those aren't too critical. >> >> -- patch description -- >> >> 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. >> >> Kconfig and Makefile modifications to enable the Bluetooth >> driver for Texas Instrument's WiLink 7 chipset. >> >> Signed-off-by: Pavan Savoy >> --- >> =C2=A0drivers/bluetooth/Kconfig =C2=A0 =C2=A0| =C2=A0 10 + >> =C2=A0drivers/bluetooth/Makefile =C2=A0 | =C2=A0 =C2=A01 + >> =C2=A0drivers/bluetooth/btwilink.c | =C2=A0379 +++++++++++++++++++++++++= +++++++++++++++++ >> =C2=A03 files changed, 390 insertions(+), 0 deletions(-) >> =C2=A0create mode 100644 drivers/bluetooth/btwilink.c >> >> diff --git a/drivers/bluetooth/Kconfig b/drivers/bluetooth/Kconfig >> index 02deef4..8e0de9a 100644 >> --- a/drivers/bluetooth/Kconfig >> +++ b/drivers/bluetooth/Kconfig >> @@ -219,4 +219,14 @@ config BT_ATH3K >> =C2=A0 =C2=A0 =C2=A0 =C2=A0 Say Y here to compile support for "Atheros f= irmware download driver" >> =C2=A0 =C2=A0 =C2=A0 =C2=A0 into the kernel or say M to compile it as mo= dule (ath3k). >> >> +config BT_WILINK >> + =C2=A0 =C2=A0 tristate "Texas Instruments WiLink7 driver" >> + =C2=A0 =C2=A0 depends on TI_ST >> + =C2=A0 =C2=A0 help >> + =C2=A0 =C2=A0 =C2=A0 This enables the Bluetooth driver for Texas Instr= ument's BT/FM/GPS >> + =C2=A0 =C2=A0 =C2=A0 combo devices. This makes use of shared transport= line discipline >> + =C2=A0 =C2=A0 =C2=A0 core driver to communicate with the BT core of th= e combo chip. >> + >> + =C2=A0 =C2=A0 =C2=A0 Say Y here to compile support for Texas Instrumen= t's WiLink7 driver >> + =C2=A0 =C2=A0 =C2=A0 into the kernel or say M to compile it as module. >> =C2=A0endmenu >> diff --git a/drivers/bluetooth/Makefile b/drivers/bluetooth/Makefile >> index 71bdf13..f4460f4 100644 >> --- a/drivers/bluetooth/Makefile >> +++ b/drivers/bluetooth/Makefile >> @@ -18,6 +18,7 @@ obj-$(CONFIG_BT_HCIBTSDIO) =C2=A0+=3D btsdio.o >> =C2=A0obj-$(CONFIG_BT_ATH3K) =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 +=3D ath3k.o >> =C2=A0obj-$(CONFIG_BT_MRVL) =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0+=3D btmrvl.o >> =C2=A0obj-$(CONFIG_BT_MRVL_SDIO) =C2=A0 +=3D btmrvl_sdio.o >> +obj-$(CONFIG_BT_WILINK) =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0+=3D btwilink.o >> >> =C2=A0btmrvl-y =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 :=3D btmrvl_main.o >> =C2=A0btmrvl-$(CONFIG_DEBUG_FS) =C2=A0 =C2=A0+=3D btmrvl_debugfs.o >> diff --git a/drivers/bluetooth/btwilink.c b/drivers/bluetooth/btwilink.c >> new file mode 100644 >> index 0000000..1b1c4bc >> --- /dev/null >> +++ b/drivers/bluetooth/btwilink.c >> @@ -0,0 +1,379 @@ >> +/* >> + * =C2=A0Texas Instrument's Bluetooth Driver For Shared Transport. >> + * >> + * =C2=A0Bluetooth Driver acts as interface between HCI core and >> + * =C2=A0TI Shared Transport Layer. >> + * >> + * =C2=A0Copyright (C) 2009-2010 Texas Instruments >> + * =C2=A0Author: Raja Mani >> + * =C2=A0 Pavan Savoy >> + * >> + * =C2=A0This program is free software; you can redistribute it and/or = modify >> + * =C2=A0it under the terms of the GNU General Public License version 2= as >> + * =C2=A0published by the Free Software Foundation. >> + * >> + * =C2=A0This program is distributed in the hope that it will be useful= , >> + * =C2=A0but WITHOUT ANY WARRANTY; without even the implied warranty of >> + * =C2=A0MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. =C2=A0See= the >> + * =C2=A0GNU General Public License for more details. >> + * >> + * =C2=A0You should have received a copy of the GNU General Public Lice= nse >> + * =C2=A0along with this program; if not, write to the Free Software >> + * =C2=A0Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA =C2= =A002111-1307 =C2=A0USA >> + * >> + */ >> + >> +#include >> +#include >> +#include >> + >> +#include >> + >> +/* Bluetooth Driver Version */ >> +#define VERSION =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 "1.0" >> + >> +/* Number of seconds to wait for registration completion >> + * when ST returns PENDING status. >> + */ >> +#define BT_REGISTER_TIMEOUT =C2=A0 6000 =C2=A0 /* 6 sec */ >> + >> +/** >> + * struct ti_st - driver operation structure >> + * @hdev: hci device pointer which binds to bt driver >> + * @reg_status: ST registration callback status >> + * @st_write: write function provided by the ST driver >> + * =C2=A0 to be used by the driver during send_frame. >> + * @wait_reg_completion - completion sync between ti_st_open >> + * =C2=A0 and ti_st_registration_completion_cb. >> + */ >> +struct ti_st { >> + =C2=A0 =C2=A0 struct hci_dev *hdev; >> + =C2=A0 =C2=A0 char reg_status; >> + =C2=A0 =C2=A0 long (*st_write) (struct sk_buff *); >> + =C2=A0 =C2=A0 struct completion wait_reg_completion; >> +}; >> + >> +/* Increments HCI counters based on pocket ID (cmd,acl,sco) */ >> +static inline void ti_st_tx_complete(struct ti_st *hst, int pkt_type) >> +{ >> + =C2=A0 =C2=A0 struct hci_dev *hdev =3D hst->hdev; >> + >> + =C2=A0 =C2=A0 /* Update HCI stat counters */ >> + =C2=A0 =C2=A0 switch (pkt_type) { >> + =C2=A0 =C2=A0 case HCI_COMMAND_PKT: >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 hdev->stat.cmd_tx++; >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 break; >> + >> + =C2=A0 =C2=A0 case HCI_ACLDATA_PKT: >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 hdev->stat.acl_tx++; >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 break; >> + >> + =C2=A0 =C2=A0 case HCI_SCODATA_PKT: >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 hdev->stat.sco_tx++; >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 break; >> + =C2=A0 =C2=A0 } >> +} >> + >> +/* ------- Interfaces to Shared Transport ------ */ >> + >> +/* Called by ST layer to indicate protocol registration completion >> + * status.ti_st_open() function will wait for signal from this >> + * API when st_register() function returns ST_PENDING. >> + */ >> +static void st_registration_completion_cb(void *priv_data, char data) >> +{ >> + =C2=A0 =C2=A0 struct ti_st *lhst =3D priv_data; >> + >> + =C2=A0 =C2=A0 /* Save registration status for use in ti_st_open() */ >> + =C2=A0 =C2=A0 lhst->reg_status =3D data; >> + =C2=A0 =C2=A0 /* complete the wait in ti_st_open() */ >> + =C2=A0 =C2=A0 complete(&lhst->wait_reg_completion); >> +} >> + >> +/* Called by Shared Transport layer when receive data is >> + * available */ >> +static long st_receive(void *priv_data, struct sk_buff *skb) >> +{ >> + =C2=A0 =C2=A0 struct ti_st *lhst =3D priv_data; >> + =C2=A0 =C2=A0 int err; >> + >> + =C2=A0 =C2=A0 if (!skb) >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return -EFAULT; >> + >> + =C2=A0 =C2=A0 if (!lhst) { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 kfree_skb(skb); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return -EFAULT; >> + =C2=A0 =C2=A0 } >> + >> + =C2=A0 =C2=A0 skb->dev =3D (void *) lhst->hdev; >> + >> + =C2=A0 =C2=A0 /* Forward skb to HCI core layer */ >> + =C2=A0 =C2=A0 err =3D hci_recv_frame(skb); >> + =C2=A0 =C2=A0 if (err < 0) { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 BT_ERR("Unable to push skb t= o HCI core(%d)", err); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return err; >> + =C2=A0 =C2=A0 } >> + >> + =C2=A0 =C2=A0 lhst->hdev->stat.byte_rx +=3D skb->len; >> + >> + =C2=A0 =C2=A0 return 0; >> +} >> + >> +/* ------- Interfaces to HCI layer ------ */ >> +/* protocol structure registered with shared transport */ >> +static struct st_proto_s ti_st_proto =3D { >> + =C2=A0 =C2=A0 .type =3D ST_BT, >> + =C2=A0 =C2=A0 .recv =3D st_receive, >> + =C2=A0 =C2=A0 .reg_complete_cb =3D st_registration_completion_cb, >> +}; >> + >> +/* Called from HCI core to initialize the device */ >> +static int ti_st_open(struct hci_dev *hdev) >> +{ >> + =C2=A0 =C2=A0 unsigned long timeleft; >> + =C2=A0 =C2=A0 struct ti_st *hst; >> + =C2=A0 =C2=A0 int err; >> + >> + =C2=A0 =C2=A0 BT_DBG("%s %p", hdev->name, hdev); >> + >> + =C2=A0 =C2=A0 /* provide contexts for callbacks from ST */ >> + =C2=A0 =C2=A0 hst =3D hdev->driver_data; >> + =C2=A0 =C2=A0 ti_st_proto.priv_data =3D hst; >> + >> + =C2=A0 =C2=A0 err =3D st_register(&ti_st_proto); >> + =C2=A0 =C2=A0 if (err =3D=3D -EINPROGRESS) { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 /* Prepare wait-for-completi= on handler data structures. >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0* Needed to synchroniz= e this and >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0* st_registration_comp= letion_cb() functions. >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/ >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 init_completion(&hst->wait_r= eg_completion); >> + >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 /* Reset ST registration cal= lback status flag , this value >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0* will be updated in t= i_st_registration_completion_cb() >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0* function whenever it= called from ST driver. >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/ >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 hst->reg_status =3D -EINPROG= RESS; >> + >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 /* ST is busy with either pr= otocol registration or firmware >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0* download. Wait until= the registration callback is called >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/ >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 BT_DBG(" waiting for registr= ation completion signal from ST"); >> + >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 timeleft =3D wait_for_comple= tion_timeout >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = (&hst->wait_reg_completion, >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0msecs_to_jiffies(BT_REGISTER_TIMEOUT)); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 if (!timeleft) { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = BT_ERR("Timeout(%d sec),didn't get reg " >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 "completion signal = from ST", >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 BT_REGISTER_TIMEOUT= / 1000); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = return -ETIMEDOUT; >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 } >> + >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 /* Is ST registration callba= ck called with ERROR status? */ >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 if (hst->reg_status !=3D 0) = { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = BT_ERR("ST registration completed with invalid " >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 "status %d", hst->r= eg_status); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = return -EAGAIN; >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 } >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 err =3D 0; >> + =C2=A0 =C2=A0 } else if (err =3D=3D -EPERM) { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 BT_ERR("st_register failed %= d", err); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return err; >> + =C2=A0 =C2=A0 } > > > Again, you are checking for only EPERM and EINPROGRESS. You should check > for everything. I actually don't undertand why you have a special check > for EPERM. > >> + >> + =C2=A0 =C2=A0 /* ti_st_proto.write is filled up by the underlying shar= ed >> + =C2=A0 =C2=A0 =C2=A0* transport driver upon registration >> + =C2=A0 =C2=A0 =C2=A0*/ >> + =C2=A0 =C2=A0 hst->st_write =3D ti_st_proto.write; > > I do not like that, the underlying should export the write function. > >> + =C2=A0 =C2=A0 if (!hst->st_write) { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 BT_ERR("undefined ST write f= unction"); >> + >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 /* Undo registration with ST= */ >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 err =3D st_unregister(ST_BT)= ; >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 if (err) >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = BT_ERR("st_unregister() failed with error %d", err); >> + >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 hst->st_write =3D NULL; >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return err; >> + =C2=A0 =C2=A0 } >> + >> + =C2=A0 =C2=A0 /* Registration with ST layer is successful, >> + =C2=A0 =C2=A0 =C2=A0* hardware is ready to accept commands from HCI co= re. >> + =C2=A0 =C2=A0 =C2=A0*/ >> + =C2=A0 =C2=A0 if (test_and_set_bit(HCI_RUNNING, &hdev->flags)) { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 clear_bit(HCI_RUNNING, &hdev= ->flags); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 err =3D st_unregister(ST_BT)= ; >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 if (err) >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = BT_ERR("st_unregister() failed with error %d", err); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 hst->st_write =3D NULL; >> + =C2=A0 =C2=A0 } > > > What are you trying to do here? test_and_set_bit() result doesn't say > nothing about error and you shall put test_and_set_bit should be in the > beginning, to know if your device is already opened or not and then > clear_bit if some error ocurrs during the function. > >> + >> + =C2=A0 =C2=A0 return err; >> +} >> + >> +/* Close device */ >> +static int ti_st_close(struct hci_dev *hdev) >> +{ >> + =C2=A0 =C2=A0 int err; >> + =C2=A0 =C2=A0 struct ti_st *hst =3D hdev->driver_data; >> + >> + =C2=A0 =C2=A0 /* continue to unregister from transport */ >> + =C2=A0 =C2=A0 err =3D st_unregister(ST_BT); >> + =C2=A0 =C2=A0 if (err) >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 BT_ERR("st_unregister() fail= ed with error %d", err); >> + >> + =C2=A0 =C2=A0 hst->st_write =3D NULL; >> + >> + =C2=A0 =C2=A0 if (!test_and_clear_bit(HCI_RUNNING, &hdev->flags)) >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return 0; > > test_and_clear_bit should come first to check if your device is already > closed. > >> + >> + =C2=A0 =C2=A0 return err; >> +} >> + >> +static int ti_st_send_frame(struct sk_buff *skb) >> +{ >> + =C2=A0 =C2=A0 struct hci_dev *hdev; >> + =C2=A0 =C2=A0 struct ti_st *hst; >> + =C2=A0 =C2=A0 long len; >> + >> + =C2=A0 =C2=A0 hdev =3D (struct hci_dev *)skb->dev; >> + >> + =C2=A0 =C2=A0 if (!test_bit(HCI_RUNNING, &hdev->flags)) >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return -EBUSY; >> + >> + =C2=A0 =C2=A0 hst =3D hdev->driver_data; >> + >> + =C2=A0 =C2=A0 /* Prepend skb with frame type */ >> + =C2=A0 =C2=A0 memcpy(skb_push(skb, 1), &bt_cb(skb)->pkt_type, 1); >> + >> + =C2=A0 =C2=A0 BT_DBG(" %s: type %d len %d", hdev->name, bt_cb(skb)->pk= t_type, >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = skb->len); >> + >> + =C2=A0 =C2=A0 /* Insert skb to shared transport layer's transmit queue= . >> + =C2=A0 =C2=A0 =C2=A0* Freeing skb memory is taken care in shared trans= port layer, >> + =C2=A0 =C2=A0 =C2=A0* so don't free skb memory here. >> + =C2=A0 =C2=A0 =C2=A0*/ >> + =C2=A0 =C2=A0 len =3D hst->st_write(skb); >> + =C2=A0 =C2=A0 if (len < 0) { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 kfree_skb(skb); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 BT_ERR(" ST write failed (%l= d)", len); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return -EAGAIN; > > Why EAGAIN? > >> + =C2=A0 =C2=A0 } >> + >> + =C2=A0 =C2=A0 /* ST accepted our skb. So, Go ahead and do rest */ >> + =C2=A0 =C2=A0 hdev->stat.byte_tx +=3D len; >> + =C2=A0 =C2=A0 ti_st_tx_complete(hst, bt_cb(skb)->pkt_type); > > From Marcel in the other thread: > "What is the reason for this deferred stats update. That code looks > pretty much hackish to me." > >> + >> + =C2=A0 =C2=A0 return 0; >> +} >> + >> +static void ti_st_destruct(struct hci_dev *hdev) >> +{ >> + =C2=A0 =C2=A0 BT_DBG("%s", hdev->name); >> + >> + =C2=A0 =C2=A0 /* free ti_st memory */ > > get ride of the comment, it's pointless > >> + =C2=A0 =C2=A0 kfree(hdev->driver_data); >> + >> + =C2=A0 =C2=A0 return; > > No return; here > >> +} >> + >> +static int bt_ti_probe(struct platform_device *pdev) >> +{ >> + =C2=A0 =C2=A0 static struct ti_st *hst; >> + =C2=A0 =C2=A0 struct hci_dev *hdev; >> + =C2=A0 =C2=A0 int err; >> + >> + =C2=A0 =C2=A0 hst =3D kzalloc(sizeof(struct ti_st), GFP_KERNEL); >> + =C2=A0 =C2=A0 if (!hst) >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return -ENOMEM; >> + >> + =C2=A0 =C2=A0 /* Expose "hciX" device to user space */ >> + =C2=A0 =C2=A0 hdev =3D hci_alloc_dev(); >> + =C2=A0 =C2=A0 if (!hdev) >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return -ENOMEM; > > You are leaking hst, if hci_alloc_dev() fails. > >> + >> + =C2=A0 =C2=A0 BT_DBG("hdev %p", hdev); >> + >> + =C2=A0 =C2=A0 hst->hdev =3D hdev; >> + =C2=A0 =C2=A0 hdev->bus =3D HCI_UART; >> + =C2=A0 =C2=A0 hdev->driver_data =3D hst; >> + =C2=A0 =C2=A0 hdev->open =3D ti_st_open; >> + =C2=A0 =C2=A0 hdev->close =3D ti_st_close; >> + =C2=A0 =C2=A0 hdev->flush =3D NULL; >> + =C2=A0 =C2=A0 hdev->send =3D ti_st_send_frame; >> + =C2=A0 =C2=A0 hdev->destruct =3D ti_st_destruct; >> + =C2=A0 =C2=A0 hdev->owner =3D THIS_MODULE; >> + >> + =C2=A0 =C2=A0 err =3D hci_register_dev(hdev); >> + =C2=A0 =C2=A0 if (err < 0) { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 BT_ERR("Can't register HCI d= evice error %d", err); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 hci_free_dev(hdev); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return err; >> + =C2=A0 =C2=A0 } >> + >> + =C2=A0 =C2=A0 BT_DBG(" HCI device registered (hdev %p)", hdev); >> + >> + =C2=A0 =C2=A0 dev_set_drvdata(&pdev->dev, hst); >> + =C2=A0 =C2=A0 return err; >> +} >> + >> +static int bt_ti_remove(struct platform_device *pdev) >> +{ >> + =C2=A0 =C2=A0 struct hci_dev *hdev; >> + =C2=A0 =C2=A0 struct ti_st *hst =3D dev_get_drvdata(&pdev->dev); >> + >> + =C2=A0 =C2=A0 if (!hst) >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return -EFAULT; >> + >> + =C2=A0 =C2=A0 hdev =3D hst->hdev; >> + =C2=A0 =C2=A0 ti_st_close(hdev); >> + =C2=A0 =C2=A0 hci_unregister_dev(hdev); >> + >> + =C2=A0 =C2=A0 /* Free HCI device memory */ >> + =C2=A0 =C2=A0 hci_free_dev(hdev); >> + >> + =C2=A0 =C2=A0 /* Free driver data memory */ > > get ride of both commnets here. The name of the funcion is already > saying that. > >> + =C2=A0 =C2=A0 kfree(hst); >> + >> + =C2=A0 =C2=A0 dev_set_drvdata(&pdev->dev, NULL); >> + =C2=A0 =C2=A0 return 0; >> +} >> + >> +static struct platform_driver btwilink_driver =3D { >> + =C2=A0 =C2=A0 .probe =3D bt_ti_probe, >> + =C2=A0 =C2=A0 .remove =3D bt_ti_remove, >> + =C2=A0 =C2=A0 .driver =3D { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 .name =3D "btwilink", >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 .owner =3D THIS_MODULE, >> + =C2=A0 =C2=A0 }, >> +}; >> + >> +/* ------- Module Init/Exit interfaces ------ */ >> +static int __init btwilink_init(void) >> +{ >> + =C2=A0 =C2=A0 long ret; >> + >> + =C2=A0 =C2=A0 BT_INFO(" Bluetooth Driver for TI WiLink - Version %s", = VERSION); >> + >> + =C2=A0 =C2=A0 ret =3D platform_driver_register(&btwilink_driver); >> + =C2=A0 =C2=A0 if (ret !=3D 0) { >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 BT_ERR("btwilink platform dr= iver registration failed"); >> + =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 return ret; >> + =C2=A0 =C2=A0 } >> + =C2=A0 =C2=A0 return 0; > > Old issue again: > > From Marcel: please just do like we do with all other drivers; > > =C2=A0 =C2=A0 =C2=A0 =C2=A0BT_INFO(...) > > =C2=A0 =C2=A0 =C2=A0 =C2=A0return platform_driver_register(&btwilink_driv= er); > > -- > Gustavo F. Padovan > http://profusion.mobi > -- > To unsubscribe from this list: send the line "unsubscribe linux-bluetooth= " in > the body of a message to majordomo@vger.kernel.org > More majordomo info at =C2=A0http://vger.kernel.org/majordomo-info.html >