| Message ID | CAH10aOhPy37DzogqHtq4AWov+w5BDyhFXmcCcvDxjD7=4ZnwqA@mail.gmail.com |
|---|---|
| State | New |
| Headers | show |
On Mon, Sep 02, 2013 at 06:27:57PM +0800, Jesse Sung wrote: > Adam Lee (1): > Andre Guedes (9): > Fengguang Wu (1): > Johan Hedberg (19): > Marcel Holtmann (1): > Tedd Ho-Jeong An (1): > Wen-chien Jesse Sung (2): 34 patches? Ouch, is this really the only way this can be fixed? -apw
2013/9/2 Andy Whitcroft <apw@canonical.com>: > On Mon, Sep 02, 2013 at 06:27:57PM +0800, Jesse Sung wrote: > >> Adam Lee (1): >> Andre Guedes (9): >> Fengguang Wu (1): >> Johan Hedberg (19): >> Marcel Holtmann (1): >> Tedd Ho-Jeong An (1): >> Wen-chien Jesse Sung (2): > > 34 patches? Ouch, is this really the only way this can be fixed? Besides two sauce patches for broadcom fw loader and one for 7260 fw loader, rest of them are bug fixes or for indroducing __hci_cmd_sync() and __hci_cmd_sync_ev(). Since the 7260 driver depends on these helper functions, we'll need most of them. And these patches benefits broadcom loader too since it works better with these functions utilized. Thanks, Jesse
Discussion of this has been moved off-list. If we do agree to pull these patches in, that submission *will* happen on-list. Brad On 09/02/2013 03:27 AM, Jesse Sung wrote: > Hi, > > Currently the broadcom firmware loader in the kernel loads firmware > with inserting > certain delay between two commands. This approach does not work well with new > platforms(eg. Lynx Point LP). Kernel 3.10 introduced an approach to > send hci commands synchronously. The firmware loader can make sure > that previous command is completed > before sending out another. It works better and there's no need to > worry about if > the delay is too small or too large. > > There's also a demand of supporting intel 7260 bluetooth module. It is > supported by > kernel 3.10, and the support depends on the approach mentioned above. > > Related changes and some bug fixes are cherry-picked from 3.10 in this > pull request. > Basically these are all clean merges, and the broadcom firmware > loader, which is an > ubuntu sauce currently, is ported to use this new approach. > > This is already tested on several machines, but since the amount of > commits is quite large, > I'd like to have your comments on this. If this looks good for SRU, > I'll do another pull request against quantal. > > Thanks. > > -- > > BugLink: https://launchpad.net/bugs/1065400 > BugLink: https://launchpad.net/bugs/1209124 > > The following changes since commit 942bf619e6a9494a775cb1df0a7d993272d05c7f: > > UBUNTU: Ubuntu-3.8.0-30.44 (2013-08-22 15:16:45 -0500) > > are available in the git repository at: > > git://kernel.ubuntu.com/jesse/ubuntu-raring.git bt-update > > for you to fetch changes up to d7ac7f500318d455077d3a34dedf7d42bb521811: > > Bluetooth: fix wrong use of PTR_ERR() in btusb (2013-08-30 21:05:21 +0800) > > ---------------------------------------------------------------- > Adam Lee (1): > Bluetooth: fix wrong use of PTR_ERR() in btusb > > Andre Guedes (9): > Bluetooth: Fix __hci_req_sync > Bluetooth: Return ENODATA in hci_req_run > Bluetooth: Check hci_req_run returning value in __hci_req_sync > Bluetooth: HCI request error handling > Bluetooth: Make hci_req_add returning void > Bluetooth: Check req->err in hci_req_add > Bluetooth: Fix HCI request framework > Bluetooth: Fix hci_inquiry ioctl usage > Bluetooth: Remove unneeded hci_req_cmd_status function > > Fengguang Wu (1): > Bluetooth: hci_get_cmd_complete() can be static > > Johan Hedberg (19): > Bluetooth: Add LE Local Features reading support > Bluetooth: Add support for reading LE White List Size > Bluetooth: Add support for reading LE supported states > Bluetooth: Rename hci_request to hci_req_sync > Bluetooth: Fix __hci_req_sync() handling of empty requests > Bluetooth: Split HCI init sequence into three stages > Bluetooth: Add initial skeleton for asynchronous HCI requests > Bluetooth: Refactor HCI command skb creation > Bluetooth: Introduce new hci_req_add function > Bluetooth: Fix stand-alone HCI command handling > Bluetooth: Add request cmd_complete and cmd_status functions > Bluetooth: Use async requests internally in hci_req_sync > Bluetooth: Remove unused hdev->init_last_cmd > Bluetooth: Remove empty HCI event handlers > Bluetooth: Fix endianness handling of cmd_status/complete opcodes > Bluetooth: Track received events in hdev > Bluetooth: Add __hci_cmd_sync() helper function > Bluetooth: Add support for custom event terminated commands > Bluetooth: Add __hci_cmd_sync_ev function > > Marcel Holtmann (1): > Bluetooth: Add driver setup stage for early init > > Tedd Ho-Jeong An (1): > Bluetooth: Add support for Intel Bluetooth device [8087:07dc] > > Wen-chien Jesse Sung (2): > UBUNTU: SAUCE: Bluetooth: Use setup callback for patchram loading > UBUNTU: SAUCE: Bluetooth: Remove load_firmware callback > > drivers/bluetooth/btusb.c | 443 > +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-------- > include/net/bluetooth/bluetooth.h | 11 +++ > include/net/bluetooth/hci.h | 18 ++++ > include/net/bluetooth/hci_core.h | 32 +++++-- > net/bluetooth/hci_core.c | 828 > ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++------------------------ > net/bluetooth/hci_event.c | 532 > ++++++++++++++----------------------------------------------------------------------------------------------- > net/bluetooth/hci_sock.c | 5 ++ > 7 files changed, 1249 insertions(+), 620 deletions(-) >
Hi, Currently the broadcom firmware loader in the kernel loads firmware with inserting certain delay between two commands. This approach does not work well with new platforms(eg. Lynx Point LP). Kernel 3.10 introduced an approach to send hci commands synchronously. The firmware loader can make sure that previous command is completed before sending out another. It works better and there's no need to worry about if the delay is too small or too large. There's also a demand of supporting intel 7260 bluetooth module. It is supported by kernel 3.10, and the support depends on the approach mentioned above. Related changes and some bug fixes are cherry-picked from 3.10 in this pull request. Basically these are all clean merges, and the broadcom firmware loader, which is an ubuntu sauce currently, is ported to use this new approach. This is already tested on several machines, but since the amount of commits is quite large, I'd like to have your comments on this. If this looks good for SRU, I'll do another pull request against quantal. Thanks. -- BugLink: https://launchpad.net/bugs/1065400 BugLink: https://launchpad.net/bugs/1209124 The following changes since commit 942bf619e6a9494a775cb1df0a7d993272d05c7f: UBUNTU: Ubuntu-3.8.0-30.44 (2013-08-22 15:16:45 -0500) are available in the git repository at: git://kernel.ubuntu.com/jesse/ubuntu-raring.git bt-update for you to fetch changes up to d7ac7f500318d455077d3a34dedf7d42bb521811: Bluetooth: fix wrong use of PTR_ERR() in btusb (2013-08-30 21:05:21 +0800) ---------------------------------------------------------------- Adam Lee (1): Bluetooth: fix wrong use of PTR_ERR() in btusb Andre Guedes (9): Bluetooth: Fix __hci_req_sync Bluetooth: Return ENODATA in hci_req_run Bluetooth: Check hci_req_run returning value in __hci_req_sync Bluetooth: HCI request error handling Bluetooth: Make hci_req_add returning void Bluetooth: Check req->err in hci_req_add Bluetooth: Fix HCI request framework Bluetooth: Fix hci_inquiry ioctl usage Bluetooth: Remove unneeded hci_req_cmd_status function Fengguang Wu (1): Bluetooth: hci_get_cmd_complete() can be static Johan Hedberg (19): Bluetooth: Add LE Local Features reading support Bluetooth: Add support for reading LE White List Size Bluetooth: Add support for reading LE supported states Bluetooth: Rename hci_request to hci_req_sync Bluetooth: Fix __hci_req_sync() handling of empty requests Bluetooth: Split HCI init sequence into three stages Bluetooth: Add initial skeleton for asynchronous HCI requests Bluetooth: Refactor HCI command skb creation Bluetooth: Introduce new hci_req_add function Bluetooth: Fix stand-alone HCI command handling Bluetooth: Add request cmd_complete and cmd_status functions Bluetooth: Use async requests internally in hci_req_sync Bluetooth: Remove unused hdev->init_last_cmd Bluetooth: Remove empty HCI event handlers Bluetooth: Fix endianness handling of cmd_status/complete opcodes Bluetooth: Track received events in hdev Bluetooth: Add __hci_cmd_sync() helper function Bluetooth: Add support for custom event terminated commands Bluetooth: Add __hci_cmd_sync_ev function Marcel Holtmann (1): Bluetooth: Add driver setup stage for early init Tedd Ho-Jeong An (1): Bluetooth: Add support for Intel Bluetooth device [8087:07dc] Wen-chien Jesse Sung (2): UBUNTU: SAUCE: Bluetooth: Use setup callback for patchram loading UBUNTU: SAUCE: Bluetooth: Remove load_firmware callback drivers/bluetooth/btusb.c | 443 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-------- include/net/bluetooth/bluetooth.h | 11 +++ include/net/bluetooth/hci.h | 18 ++++ include/net/bluetooth/hci_core.h | 32 +++++-- net/bluetooth/hci_core.c | 828 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++------------------------ net/bluetooth/hci_event.c | 532 ++++++++++++++----------------------------------------------------------------------------------------------- net/bluetooth/hci_sock.c | 5 ++ 7 files changed, 1249 insertions(+), 620 deletions(-)