Topcliff: Update PCH_SPI driver to 2.6.35
2010/9/3 Masayuki Ohtak <[email protected]>:
Quoted 9 lines
SPI driver of Topcliff PCH
Topcliff PCH is the platform controller hub that is going to be used in Intel's upcoming general embedded platform. All IO peripherals in Topcliff PCH are actually devices sitting on AMBA bus. Topcliff PCH has SPI I/F. Using this I/F, it is able to access system devices connected to SPI.
Signed-off-by Masayuki Ohtake <[email protected]>
Hi Ohtak-san
Comments below. Also, please also fix the method that you're using to post patches. The email client you're currently using is posting the patches in HTML format, and is mangling them. git send-email can help you here.
Quoted 25 lines
--- drivers/spi/Kconfig | 19 + drivers/spi/Makefile | 1 + drivers/spi/spi_pch.c | 1813 +++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 1833 insertions(+), 0 deletions(-) create mode 100644 drivers/spi/spi_pch.cdrivers/spi/Kconfig +9
-0@@ -53,6 +53,25 @@ if SPI_MASTER comment "SPI Master Controller Drivers" +config PCH_SPI +# tristate "PCH SPI Controller" + boolean "PCH SPI Controller" + depends on PCI + help + This driver is for PCH(Platform controller Hub) SPI of Topcliff which + is an IOH(Input/Output Hub) for x86 embedded processor. + This driver can access PCH SPI bus device. +
drivers/spi/Kconfig if SPI_MASTER · line 74
Quoted 10 lines
drivers/spi/Kconfig +10
-0+config PCH_SPI_PLATFORM_DEVICE_COUNT + int "PCH SPI Bus count" + range 1 2 + depends on PCH_SPI + help + This driver is for PCH(Platform controller Hub) SPI of Topcliff which + is an IOH(Input/Output Hub) for x86 embedded processor. + The number of SPI buses/channels supported by the PCH SPI controller. + PCH SPI of Topcliff supports only one channel. +
As previously discussed, PCH_SPI_PLATFORM_DEVICE_COUNT needs to be removed (see below).
[...]
Quoted 48 lines
drivers/spi/Kconfig +47
-0+static int pch_spi_probe(struct pci_dev *pdev, const struct pci_device_id *id) +{ + + struct spi_master *master[PCH_MAX_DEV]; + + struct pch_spi_board_data *board_dat; + int retval, i; + int spi_alloc_num, master_num; + + dev_dbg(&pdev->dev, "%s ENTRY\n", __func__); + + /* allocate memory for private data */ + board_dat = kzalloc(sizeof(struct pch_spi_board_data), GFP_KERNEL); + + if (board_dat == NULL) { + dev_err(&pdev->dev, + " %s memory allocation for private data failed\n", + __func__); + retval = -ENOMEM; + goto err_kmalloc; + } + + dev_dbg(&pdev->dev, + "%s memory allocation for private data success\n", __func__); + + /* enable PCI device */ + retval = pci_enable_device(pdev); + if (retval != 0) { + dev_err(&pdev->dev, "%s pci_enable_device FAILED\n", __func__); + + goto err_pci_en_device; + } + + dev_dbg(&pdev->dev, "%s pci_enable_device returned=%d\n", + __func__, retval); + + board_dat->pdev = pdev; + + /* alllocate memory for SPI master */ + for (i = 0, spi_alloc_num = 0; i < PCH_MAX_DEV; i++, spi_alloc_num++) { + master[i] = spi_alloc_master(&pdev->dev, + sizeof(struct pch_spi_data)); + + if (master[i] == NULL) { + retval = -ENOMEM; + dev_err(&pdev->dev, "%s [ch=%d]Fail.\n", __func__, i); + goto err_spi_alloc_master;
drivers/spi/Kconfig if SPI_MASTER
Quoted 10 lines
drivers/spi/Kconfig +10
-0+ } + } + + dev_dbg(&pdev->dev, + "%s spi_alloc_master returned non NULL\n", __func__); + + /* initialize members of SPI master */ + for (i = 0, master_num = 0; i < PCH_MAX_DEV; + i++, master_num++) { + master[i]->bus_num = i;
This is not okay. The driver cannot assume that it provides the only SPI busses in the system. The bus number should either be set to -1 so that it gets auto-assigned, or the bus number should be passed in by board-specific setup code.
Quoted 49 lines
drivers/spi/Kconfig +49
-0+ master[i]->num_chipselect = PCH_MAX_CS; + master[i]->setup = pch_spi_setup; + dev_dbg(&pdev->dev, + "%s setup member of SPI master initialized\n", + __func__); + master[i]->transfer = pch_spi_transfer; + dev_dbg(&pdev->dev, + "%s transfer member of SPI master initialized\n", + __func__); + + board_dat->data[i] = spi_master_get_devdata(master[i]); + + board_dat->data[i]->master = master[i]; + board_dat->data[i]->n_curnt_chip = 255; + board_dat->data[i]->current_chip = NULL; + board_dat->data[i]->transfer_complete = false; + board_dat->data[i]->pkt_tx_buff = NULL; + board_dat->data[i]->pkt_rx_buff = NULL; + board_dat->data[i]->tx_index = 0; + board_dat->data[i]->rx_index = 0; + board_dat->data[i]->transfer_active = false; + board_dat->data[i]->board_dat = board_dat; + board_dat->suspend_sts = false; + + mutex_init(&board_dat->data[i]->mutex); + } + + + /* allocate resources for PCH SPI */ + retval = pch_spi_get_resources(board_dat); + if (retval != 0) { + dev_err(&pdev->dev, "%s fail(retval=%d)\n", + __func__, retval); + goto err_spi_get_resources; + } + + dev_dbg(&pdev->dev, "%s pch_spi_get_resources returned=%d\n", + __func__, retval); + + /* save private data in dev */ + pci_set_drvdata(pdev, (void *)board_dat); + dev_dbg(&pdev->dev, "%s invoked pci_set_drvdata\n", __func__); + + /* set master mode */ + for (i = 0; i < PCH_MAX_DEV; i++) { + pch_spi_set_master_mode(master[i]); + dev_dbg(&pdev->dev, + "%s invoked pch_spi_set_master_mode\n", __func__); + }
drivers/spi/Kconfig
Quoted 10 lines
drivers/spi/Kconfig +10
-0+ + /* Register the controller with the SPI core. */ + for (i = 0, master_num = 0; i < PCH_MAX_DEV; i++, master_num++) { + retval = spi_register_master(master[i]); + if (retval != 0) { + dev_err(&pdev->dev, + "%s spi_register_master FAILED\n", __func__); + goto err_spi_reg_master; + } + }
I *finally* understand what you're trying to do here. It looks like you have a single PCI device that has multiple independent SPI busses.
It also appears that this driver attempts to handle all of the
instances within a single context.
Doing it this way overcomplicates the driver because you need to have for() loops all over the driver to deal with the multiple instances.
Instead, the majority of the code should be structured to handle one and only one spi bus instance. Only at the top level pci probe, remove and PM code should there be for() loops for setting up or tearing down each spi bus instance one at a time.
In fact, it may be better for the pci_probe() hook to simply register one or more platform_devices; one for each instance of the SPI bus, and then have a separate platform_driver that binds to each instance.
drivers/spi/Kconfig
Quoted 10 lines
drivers/spi/Kconfig +10
-0+static int __init pch_spi_init(void) +{ +#ifdef PCH_SPIDEV_REG_OLD + int retval; + + retval = spi_register_board_info + (pch_spi_slaves, ARRAY_SIZE(pch_spi_slaves)); + if (retval) + return retval; +#endif
As we discussed previously, this driver must *not* register child spi devices with static code like this. Remove the above lines and the pch_spi_slaves array.
drivers/spi/Kconfig
Quoted 13 lines
drivers/spi/Kconfig +13
-0+ return pci_register_driver(&pch_spi_pcidev); +} + + +static void __exit pch_spi_exit(void) +{ + pci_unregister_driver(&pch_spi_pcidev); +} + +MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("PCH SPI PCI Driver"); +module_init(pch_spi_init); +module_exit(pch_spi_exit);
module_init() and modules_exit() should appear immediately after the functions they register as init and exit hooks.
Signature · 2 lines
Grant Likely, B.Sc., P.Eng. Secret Lab Technologies Ltd.
2010/9/3 Grant Likely <[email protected]>:
Quoted 7 lines
2010/9/3 Masayuki Ohtak <[email protected]>:
Quoted 1 line
SPI driver of Topcliff PCH
Instead, the majority of the code should be structured to handle one and only one spi bus instance. Only at the top level pci probe, remove and PM code should there be
for()loops for setting up or tearing down each spi bus instance one at a time.
I guess an elegant way to achieve this is to have the PCI driver spawn a simple platform device for each bus. Then you can even break the driver for that platform device into its own file and all.
In the MFDs we spawn child platform devices all the time, it's great fun!
Yours,
Linus Walleij
Quoted 9 lines · Linus Walleij
2010/9/3 Grant Likely <[email protected]>:
Quoted 7 lines
2010/9/3 Masayuki Ohtak <[email protected]>:
Quoted 1 line
SPI driver of Topcliff PCH
Instead, the majority of the code should be structured to handle one and only one spi bus instance. Only at the top level pci probe, remove and PM code should there be
for()loops for setting up or tearing down each spi bus instance one at a time.
Quoted 4 lines
I guess an elegant way to achieve this is to have the PCI driver spawn a simple platform device for each bus. Then you can even break the driver for that platform device into its own file and all.
Yes, that is exactly what I mean.
g.
Hi Grant,
Thank you for your comments.
We will modify for your comments.
Quoted 3 lines
The email client you're currently using is posting the patches in HTML format, and is mangling them. git send-email can help you here.
In our network environment, we cannot use send-email command. We have used Thunderbird with according to Documentation/email-clients.txt It may be the reason we are use Thunderbird with pre-format mode. Can't we use Thunderbird with pre-format mode ?
Thanks, Ohtake(OKISemi)
----- Original Message -----
From: "Grant Likely" <[email protected]>
To: "Masayuki Ohtak" <[email protected]>
Cc: <[email protected]>; "LKML" <[email protected]>; "David Brownell" <[email protected]>; <[email protected]>; <[email protected]>; <[email protected]>; <[email protected]>; <[email protected]>; "Tomoya MORINAGA" <[email protected]>
Sent: Friday, September 03, 2010 11:46 PM
Subject: Re: [MeeGo-Dev][PATCH] Topcliff: Update PCH_SPI driver to 2.6.35
Quoted 236 lines
2010/9/3 Masayuki Ohtak <[email protected]>:
Quoted 9 lines
SPI driver of Topcliff PCH
Topcliff PCH is the platform controller hub that is going to be used in Intel's upcoming general embedded platform. All IO peripherals in Topcliff PCH are actually devices sitting on AMBA bus. Topcliff PCH has SPI I/F. Using this I/F, it is able to access system devices connected to SPI.
Signed-off-by Masayuki Ohtake <[email protected]>Hi Ohtak-san
Comments below. Also, please also fix the method that you're using to post patches. The email client you're currently using is posting the patches in HTML format, and is mangling them. git send-email can help you here.
Quoted 25 lines
--- drivers/spi/Kconfig | 19 + drivers/spi/Makefile | 1 + drivers/spi/spi_pch.c | 1813 +++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 1833 insertions(+), 0 deletions(-) create mode 100644 drivers/spi/spi_pch.cdrivers/spi/Kconfig +9
-0@@ -53,6 +53,25 @@ if SPI_MASTER comment "SPI Master Controller Drivers" +config PCH_SPI +# tristate "PCH SPI Controller" + boolean "PCH SPI Controller" + depends on PCI + help + This driver is for PCH(Platform controller Hub) SPI of Topcliff which + is an IOH(Input/Output Hub) for x86 embedded processor. + This driver can access PCH SPI bus device. +drivers/spi/Kconfig if SPI_MASTER · line 74
Quoted 10 lines
drivers/spi/Kconfig +10
-0+config PCH_SPI_PLATFORM_DEVICE_COUNT + int "PCH SPI Bus count" + range 1 2 + depends on PCH_SPI + help + This driver is for PCH(Platform controller Hub) SPI of Topcliff which + is an IOH(Input/Output Hub) for x86 embedded processor. + The number of SPI buses/channels supported by the PCH SPI controller. + PCH SPI of Topcliff supports only one channel. +As previously discussed,
PCH_SPI_PLATFORM_DEVICE_COUNTneeds to be removed (see below).[...]
Quoted 48 lines
drivers/spi/Kconfig +47
-0+static int pch_spi_probe(struct pci_dev *pdev, const struct pci_device_id *id) +{ + + struct spi_master *master[PCH_MAX_DEV]; + + struct pch_spi_board_data *board_dat; + int retval, i; + int spi_alloc_num, master_num; + + dev_dbg(&pdev->dev, "%s ENTRY\n", __func__); + + /* allocate memory for private data */ + board_dat = kzalloc(sizeof(struct pch_spi_board_data), GFP_KERNEL); + + if (board_dat == NULL) { + dev_err(&pdev->dev, + " %s memory allocation for private data failed\n", + __func__); + retval = -ENOMEM; + goto err_kmalloc; + } + + dev_dbg(&pdev->dev, + "%s memory allocation for private data success\n", __func__); + + /* enable PCI device */ + retval = pci_enable_device(pdev); + if (retval != 0) { + dev_err(&pdev->dev, "%s pci_enable_device FAILED\n", __func__); + + goto err_pci_en_device; + } + + dev_dbg(&pdev->dev, "%s pci_enable_device returned=%d\n", + __func__, retval); + + board_dat->pdev = pdev; + + /* alllocate memory for SPI master */ + for (i = 0, spi_alloc_num = 0; i < PCH_MAX_DEV; i++, spi_alloc_num++) { + master[i] = spi_alloc_master(&pdev->dev, + sizeof(struct pch_spi_data)); + + if (master[i] == NULL) { + retval = -ENOMEM; + dev_err(&pdev->dev, "%s [ch=%d]Fail.\n", __func__, i); + goto err_spi_alloc_master;drivers/spi/Kconfig if SPI_MASTER
Quoted 10 lines
drivers/spi/Kconfig +10
-0+ } + } + + dev_dbg(&pdev->dev, + "%s spi_alloc_master returned non NULL\n", __func__); + + /* initialize members of SPI master */ + for (i = 0, master_num = 0; i < PCH_MAX_DEV; + i++, master_num++) { + master[i]->bus_num = i;This is not okay. The driver cannot assume that it provides the only SPI busses in the system. The bus number should either be set to -1 so that it gets auto-assigned, or the bus number should be passed in by board-specific setup code.
Quoted 49 lines
+ master[i]->num_chipselect =
PCH_MAX_CS;
+ master[i]->setup =pch_spi_setup;
+ dev_dbg(&pdev->dev,
+ "%s setup member of SPI master initialized\n",
+__func__);
+ master[i]->transfer =pch_spi_transfer;
+ dev_dbg(&pdev->dev,
+ "%s transfer member of SPI master initialized\n",
+__func__);drivers/spi/Kconfig +40
-0+ + board_dat->data[i] = spi_master_get_devdata(master[i]); + + board_dat->data[i]->master = master[i]; + board_dat->data[i]->n_curnt_chip = 255; + board_dat->data[i]->current_chip = NULL; + board_dat->data[i]->transfer_complete = false; + board_dat->data[i]->pkt_tx_buff = NULL; + board_dat->data[i]->pkt_rx_buff = NULL; + board_dat->data[i]->tx_index = 0; + board_dat->data[i]->rx_index = 0; + board_dat->data[i]->transfer_active = false; + board_dat->data[i]->board_dat = board_dat; + board_dat->suspend_sts = false; + + mutex_init(&board_dat->data[i]->mutex); + } + + + /* allocate resources for PCH SPI */ + retval = pch_spi_get_resources(board_dat); + if (retval != 0) { + dev_err(&pdev->dev, "%s fail(retval=%d)\n", + __func__, retval); + goto err_spi_get_resources; + } + + dev_dbg(&pdev->dev, "%s pch_spi_get_resources returned=%d\n", + __func__, retval); + + /* save private data in dev */ + pci_set_drvdata(pdev, (void *)board_dat); + dev_dbg(&pdev->dev, "%s invoked pci_set_drvdata\n", __func__); + + /* set master mode */ + for (i = 0; i < PCH_MAX_DEV; i++) { + pch_spi_set_master_mode(master[i]); + dev_dbg(&pdev->dev, + "%s invoked pch_spi_set_master_mode\n", __func__); + }drivers/spi/Kconfig
Quoted 10 lines
drivers/spi/Kconfig +10
-0+ + /* Register the controller with the SPI core. */ + for (i = 0, master_num = 0; i < PCH_MAX_DEV; i++, master_num++) { + retval = spi_register_master(master[i]); + if (retval != 0) { + dev_err(&pdev->dev, + "%s spi_register_master FAILED\n", __func__); + goto err_spi_reg_master; + } + }I *finally* understand what you're trying to do here. It looks like you have a single PCI device that has multiple independent SPI busses.
It also appears that this driver attempts to handle all of the
instances within a single context.Doing it this way overcomplicates the driver because you need to have
for()loops all over the driver to deal with the multiple instances.Instead, the majority of the code should be structured to handle one and only one spi bus instance. Only at the top level pci probe, remove and PM code should there be
for()loops for setting up or tearing down each spi bus instance one at a time.In fact, it may be better for the
pci_probe()hook to simply register one or moreplatform_devices; one for each instance of the SPI bus, and then have a separateplatform_driverthat binds to each instance.drivers/spi/Kconfig
Quoted 10 lines
drivers/spi/Kconfig +10
-0+static int __init pch_spi_init(void) +{ +#ifdef PCH_SPIDEV_REG_OLD + int retval; + + retval = spi_register_board_info + (pch_spi_slaves, ARRAY_SIZE(pch_spi_slaves)); + if (retval) + return retval; +#endifAs we discussed previously, this driver must *not* register child spi devices with static code like this. Remove the above lines and the
pch_spi_slavesarray.drivers/spi/Kconfig
Quoted 13 lines
+ return
pci_register_driver(&pch_spi_pcidev);drivers/spi/Kconfig +12
-0+} + + +static void __exit pch_spi_exit(void) +{ + pci_unregister_driver(&pch_spi_pcidev); +} + +MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("PCH SPI PCI Driver"); +module_init(pch_spi_init); +module_exit(pch_spi_exit);
module_init()andmodules_exit()should appear immediately after the functions they register as init and exit hooks.Signature · 2 lines
Grant Likely, B.Sc., P.Eng. Secret Lab Technologies Ltd.
Quoted 5 lines · Masayuki Ohtake
Hi Grant,
Thank you for your comments.
We will modify for your comments.
Quoted 7 lines
Quoted 3 lines
The email client you're currently using is posting the patches in HTML format, and is mangling them. git send-email can help you here.
In our network environment, we cannot use send-email command. We have used Thunderbird with according to
Documentation/email-clients.txtIt may be the reason we are use Thunderbird with pre-format mode. Can't we use Thunderbird with pre-format mode ?
Seems to be working now.
g.