Some MIPS defconfigs that don't define 'CONFIG_PCI_DRIVERS_GENERIC' but
define 'CONFIG_PCI_DRIVERS_LEGACY' or none of them, can fail when they are
built since definition for 'pci_remap_iospace' is being done in include
file 'arch/mips/include/asm/pci.h' and the specific function implemented
in 'arch/mips/pci/pci-generic.c'. MIPS PCI drivers that don't use generic
PCI core APIs don't really need 'pci_remap_iospace' to be defined at all.
Hence, change definition for 'pci_remap_iospace' to be dependent on the
preprocessor 'CONFIG_PCI_DRIVERS_GENERIC' definition to avoid possible
build problems.
Fixes: 9f76779f24 ("MIPS: implement architecture-specific 'pci_remap_iospace()'")
CC: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Sergio Paracuellos <sergio.paracuellos@gmail.com>
Link: https://lore.kernel.org/r/20211008095311.26475-1-sergio.paracuellos@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
'cpc' and 'mc' nodes correspond with the MIPS 'Cluster Power Controller'
and 'MIPS Common Device Memory Map' which are present in some MIPS related
boards. There is already bindings documentation for these two located in:
- Documentation/devicetree/bindings/power/mti,mips-cpc.yaml
- Documentation/devicetree/bindings/bus/mti,mips-cdmm.yaml
Hence, properly update compatible strings and align nodes with already
mainlined bindings documentation. Also, move their definition to a proper
place since both of them are not related with the palmbus at all.
Signed-off-by: Sergio Paracuellos <sergio.paracuellos@gmail.com>
Link: https://lore.kernel.org/r/20211002060706.30511-1-sergio.paracuellos@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The value to adjust in the bridge register RALINK_PCI_IOBASE must take into
account the raw value from DT, not only the translated linux port number.
As long as io_offset is zero, the two are the same, but if you were to use
multiple host bridge in the system, or pick a different bus address in DT,
you can have a nonzero io_offset. At this means to take into account the
bus address which is used to calculate this offset, substracting it from
the IO resource start address.
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Sergio Paracuellos <sergio.paracuellos@gmail.com>
Link: https://lore.kernel.org/r/20210925203224.10419-7-sergio.paracuellos@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
pci_remap_iospace() was originally meant as an architecture specific helper
but it moved into generic code after all architectures had the same
requirements. MIPS has different requirements so it should not be shared.
The way for doing this will be using a macro 'pci_remap_iospace' defined
for those architectures that need a special treatment. Hence, put core API
function inside preprocesor conditional code for 'pci_remap_iospace'
definition.
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Sergio Paracuellos <sergio.paracuellos@gmail.com>
Link: https://lore.kernel.org/r/20210925203224.10419-5-sergio.paracuellos@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This header defines a number of symbols/macros that are not used and can
be removed. In addition, it defines a number of synonym symbols that are
used interchangeably in the source. Only one is now selected.
In addition to the above, a number of comments that have no meaning for
this chip have been removed.
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Link: https://lore.kernel.org/r/20210927214527.5341-1-Larry.Finger@lwfinger.net
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Remove mutex 'usb_vendor_req_mutex'.
It was used to protect a shared buffer for USB requests and, since that
buffer is removed in previous patch, this mutex is now useless.
Furthermore, because it was used to serialize the calls to the Core USB
API, we thoroughly tested the enabling of concurrent firing of USB requests
without the mutex and found no problems of any kind in common use cases.
Co-developed-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Link: https://lore.kernel.org/r/20210924122705.3781-17-fmdefrancesco@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Create and call new usb_write() instead of usbctrl_vendorreq() from
inside rtw_write{8,16,32,N}().
In old code, rtw_write{8,16,32,N}() called usbctrl_vendorreq() which in
turn uses usb_control_msg() from within a "while" loop to build a control
URB, send it off and wait for completion. usbctrl_vendorreq() was used
for both receiving and sending messages, depending on the "requesttype"
argument which is passed by callers.
Compared to usbctrl_vendorreq(), which manages both reads and writes
from and to the USB endpoint, the new usb_write() manages only writes.
For this purpose it uses the newer USB Core usb_control_msg_send() API.
The latter is preferred according both to suggestions by Greg Kroah-Hartman
and also to its actual design.
A noteworthy feature of usb_control_msg_send() is that the data pointer
can be made to a reference on the stack because it does not have the
restriction that usb_control_msg() has where the data pointer must be to
dynamically allocated memory.
usbctrl_vendorreq() used a "while" loop that we considered unnecessary
so that it is not in the new usb_write(). Furthermore, the latter has no
redundant checking, less obvious comments, no debug prints, and it manages
errors before success case. All in all, usb_write() is simpler than
usbctrl_vendorreq() and uses less lines of code.
Suggested-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Co-developed-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
Link: https://lore.kernel.org/r/20210924122705.3781-15-fmdefrancesco@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Create and call new usb_read() instead of usbctrl_vendorreq() from
inside rtw_read{8,16,32}().
In old code, rtw_read{8,16,32}() called usbctrl_vendorreq() which in
turn uses usb_control_msg() from within a "while" loop to build a control
URB, send it off and wait for completion. usbctrl_vendorreq() was used for
both receiving and sending messages, depending on the "requesttype"
argument which was passed by callers.
Compared to usbctrl_vendorreq(), which managed both reads and writes
from and to the USB endpoint, the new usb_read() manages only reads. For
this purpose it uses the newer USB Core usb_control_msg_recv() API. The
latter is preferred according both to a suggestion by Greg Kroah-Hartman
and also to its actual design.
Two noteworthy features of usb_control_msg_recv() are that (1) the data
pointer can be made to a reference on the stack because it does not have
the restriction that usb_control_msg() has where the data pointer must be
to dynamically allocated memory, and that (2) the whole message must be
properly received from the device in order for this function to be
successfuli (if a device returns less than the expected amount of data,
then the function will fail).
usbctrl_vendorreq() uses a "while" loop that we considered unnecessary
so that it is not in the new usb_read(). Furthermore, the latter has no
redundant checking, less obvious comments, and it manages errors before
success cases. All in all, usb_read() is simpler than
usbctrl_vendorreq() and uses less lines of code.
Suggested-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Co-developed-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
Link: https://lore.kernel.org/r/20210924122705.3781-14-fmdefrancesco@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Change the type of "data" from __le32 to __le16. The size of the data
that usbctrl_vendorreq() will read is two bytes in little endian order,
so the most suitable type is __le16.
With the old code, since the two most significant bytes of data are not
initialized, KMSan can likely detect the reading of uninitialized data,
so this change can prevent the checker from complaining.
Co-developed-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
Link: https://lore.kernel.org/r/20210924122705.3781-12-fmdefrancesco@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Change the type of "data" from __le32 to __le16 in rtw_write16(). The
argument "val", which is u16, after being conditionally swapped to little
endian, is assigned to "data"; therefore, __le16 is the most suitable type
for "data". Remove the bitwise AND of "val" with 0xffff because it is
redundant. Use cpu_to_le16() because "data" is __le16.
Co-developed-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
Link: https://lore.kernel.org/r/20210924122705.3781-9-fmdefrancesco@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
