- 03 Sep, 2015 7 commits
-
-
Leonid Yegoshin authored
The MSA architecture specification allows for hardware to not implement unaligned vector memory accesses in some or all cases. A typical example of this is the I6400 core which does not implement unaligned vector memory access when the memory crosses a page boundary. The architecture also requires that such memory accesses complete successfully as far as userland is concerned, so the kernel is required to emulate them. This patch implements support for emulating unaligned MSA ld & st instructions by copying between the user memory & the tasks FP context in struct thread_struct, updating hardware registers from there as appropriate in order to avoid saving & restoring the entire vector context for each unaligned memory access. Tested both using an I6400 CPU and with a QEMU build hacked to produce AdEL exceptions for unaligned vector memory accesses. [paul.burton@imgtec.com: - Remove #ifdef's - Move msa_op into enum major_op rather than #define - Replace msa_{to,from}_wd with {read,write}_msa_wr_{b,h,w,l} and the format-agnostic wrappers, removing the custom endian mangling for big endian systems. - Restructure the msa_op case in emulate_load_store_insn to share more code between the load & store cases. - Avoid the need for a temporary union fpureg on the stack by simply reusing the already suitably aligned context in struct thread_struct. - Use sizeof(*fpr) rather than hardcoding 16 as the size for user memory checks & copies. - Stop recalculating the address of the unaligned vector memory access and rely upon the value read from BadVAddr as we do for other unaligned memory access instructions. - Drop the now unused val8 & val16 fields in union fpureg. - Rewrite commit message. - General formatting cleanups.] Signed-off-by:Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Cc: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Cc: Huacai Chen <chenhc@lemote.com> Cc: Maciej W. Rozycki <macro@linux-mips.org> Cc: linux-kernel@vger.kernel.org Cc: Jie Chen <chenj@lemote.com> Cc: Markos Chandras <markos.chandras@imgtec.com> Patchwork: https://patchwork.linux-mips.org/patch/10573/Signed-off-by:
Ralf Baechle <ralf@linux-mips.org>
-
Paul Burton authored
Introduce accessor functions allowing the kernel to access arbitrary vector registers using an arbitrary data format. The accessors are implemented in assembly, using macros to avoid massive duplication, in order to make use of the existing support for MSA with & without toolchain support. The accessors will be used in a later patch. Signed-off-by:
-
Leonid Yegoshin authored
Declare a struct describing the MSA MI10 instruction format used for ld & st instructions, for use by subsequent patches. Signed-off-by:
-
Bjorn Helgaas authored
"__weak" is defined in include/linux/compiler-gcc.h. We shouldn't need an arch-specific definition. Remove the "__weak" definition from arch/mips/include/asm/linkage.h. Signed-off-by:
Bjorn Helgaas <bhelgaas@google.com> Reviewed-by:
James Hogan <james.hogan@imgtec.com> Cc: Andrew Bresticker <abrestic@chromium.org> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/10689/Signed-off-by:
-
Bjorn Helgaas authored
Weak header file declarations are error-prone because they make every definition weak, and the linker chooses one based on link order (see 10629d71 ("PCI: Remove __weak annotation from pcibios_get_phb_of_node decl")). mips_cdmm_phys_base() is defined only in arch/mips/mti-malta/malta-memory.c so there's no problem with multiple definitions. But it works better to have a weak default implementation and allow a strong function to override it. Then we don't have to test whether a definition is present, and if there are ever multiple strong definitions, we get a link error instead of calling a random definition. Add a weak mips_cdmm_phys_base() definition and remove the weak annotation from the declaration in arch/mips/include/asm/cdmm.h. Signed-off-by:
-
Bjorn Helgaas authored
Weak header file declarations are error-prone because they make every definition weak, and the linker chooses one based on link order (see 10629d71 ("PCI: Remove __weak annotation from pcibios_get_phb_of_node decl")). The most elegant solution is to have a weak default implementation and allow a strong function to override it. Then we don't have to test whether a definition is present, and if there are ever multiple strong definitions, we get a link error instead of calling a random definition. Add a weak get_c0_fdc_int() definition with the default code and remove the weak annotation from the declaration. Signed-off-by:
-
Bjorn Helgaas authored
Weak header file declarations are error-prone because they make every definition weak, and the linker chooses one based on link order (see 10629d71 ("PCI: Remove __weak annotation from pcibios_get_phb_of_node decl")). get_c0_compare_int() is defined in several files. Each definition is weak, so I assume Kconfig prevents two or more from being included. The caller contains default code used when get_c0_compare_int() isn't defined at all. Add a weak get_c0_compare_int() definition with the default code and remove the weak annotation from the declaration. Then the platform implementations will be strong and will override the weak default. If multiple platforms are ever configured in, we'll get a link error instead of calling a random platform's implementation. Signed-off-by:
-
- 26 Aug, 2015 33 commits
-
-
Bjorn Helgaas authored
Weak header file declarations are error-prone because they make every definition weak, and the linker chooses one based on link order (see 10629d71 ("PCI: Remove __weak annotation from pcibios_get_phb_of_node decl")). That's not a problem for vpe_run() because Kconfig ensures there's never more than one definition: - vpe_run() is defined in arch/mips/kernel/vpe-mt.c if CONFIG_MIPS_VPE_LOADER_MT=y - vpe_run() is defined in arch/mips/mti-malta/malta-amon.c if CONFIG_MIPS_CMP=y - CONFIG_MIPS_VPE_LOADER_MT cannot be set if CONFIG_MIPS_CMP=y But it's simpler to verify correctness if we remove "weak" from the picture and test the config symbols directly. Remove "weak" from the vpe_run() declaration and use #if to test whether a definition should be present. Signed-off-by:
-
Bjorn Helgaas authored
vpe_run() is a weak symbol. If there's no definition of it, its value is zero. If vpe_run is zero, return failure early. We're going to fail anyway, so there's no point in getting a VPE and attempting to load it. Signed-off-by:
-
Bjorn Helgaas authored
Weak header file declarations are error-prone because they make every definition weak, and the linker chooses one based on link order (see 10629d71 ("PCI: Remove __weak annotation from pcibios_get_phb_of_node decl")). platform_maar_init() is defined in: - arch/mips/mm/init.c (where it is marked "weak") - arch/mips/mti-malta/malta-memory.c (without annotation) The "weak" attribute on the platform_maar_init() extern declaration applies to the platform-specific definition in arch/mips/mti-malta/malta-memory.c, so both definitions are weak, and which one we get depends on link order. Remove the "weak" attribute from the declaration. That makes the malta definition strong, so it will always be preferred if it is present. Signed-off-by:
-
Bjorn Helgaas authored
There's only one implementation of mips_cpc_phys_base(), and it's only used within the same file, so it doesn't need to be weak, and it doesn't need an extern declaration. Remove the extern mips_cpc_phys_base() declaration and make it static. [ralf@linux-mips.org: Fixed conflict.] Signed-off-by:
-
Paul Burton authored
The debug.h header provided some MIPS-specific debug macros, which are no longer used at all. Remove them. Signed-off-by:
-
Paul Burton authored
The db_assert call checks whether the bus_num pointer is non-NULL, but does so after said pointer has been dereferenced by the assignment on the previous line. Thus the check is pointless & likely to have been optimised out by the compiler anyway. The check_args function is static & only ever called from the local file with bus_num being a pointer to an on-stack variable, so the check seems somewhat overzealous anyway. Simply remove it. Signed-off-by:
-
Thomas Gleixner authored
The irq argument of most interrupt flow handlers is unused or merily used instead of a local variable. The handlers which need the irq argument can retrieve the irq number from the irq descriptor. Search and update was done with coccinelle and the invaluable help of Julia Lawall. Signed-off-by:
Thomas Gleixner <tglx@linutronix.de> Cc: Julia Lawall <Julia.Lawall@lip6.fr> Cc: Jiang Liu <jiang.liu@linux.intel.com> Cc: linux-mips@linux-mips.org Cc: LKML <linux-kernel@vger.kernel.org> Patchwork: https://patchwork.linux-mips.org/patch/10706/Signed-off-by:
-
Thomas Gleixner authored
The irq argument of most interrupt flow handlers is unused or merily used instead of a local variable. The handlers which need the irq argument can retrieve the irq number from the irq descriptor. Search and update was done with coccinelle and the invaluable help of Julia Lawall. Signed-off-by:
-
Thomas Gleixner authored
The gpio interrupt handling of octeon contains a homebrewn flow handler which calls either handle_level_irq or handle_edge_irq depending on the trigger type. Thats an extra conditional and call in the interrupt handling path. The proper way to handle different types and therefor different flows is to update the handler in the irq_set_type() callback. Remove the extra indirection and add the handler update to octeon_irq_ciu_gpio_set_type(). At mapping time it defaults to handle_level_irq which gets updated if the device tree contains a different trigger type. Signed-off-by:
-
Thomas Gleixner authored
The various interrupt flow handlers in ath79 are cascading interrupt handlers. They all have a disable_irq_nosync()/enable_irq() pair around the generic_handle_irq() call. The value of this disable/enable is zero because its a complete noop: disable_irq_nosync() merily increments the disable count without actually masking the interrupt. enable_irq() soleley decrements the disable count without touching the interrupt chip. The interrupt cannot arrive again because the complete call chain runs with interrupts disabled. Remove it. Signed-off-by:
-
Thomas Gleixner authored
bcsr_csc_handler() is a cascading interrupt handler. It has a disable_irq_nosync()/enable_irq() pair around the generic_handle_irq() call. The value of this disable/enable is zero because its a complete noop: disable_irq_nosync() merily increments the disable count without actually masking the interrupt. enable_irq() soleley decrements the disable count without touching the interrupt chip. The interrupt cannot arrive again because the complete call chain runs with interrupts disabled. Remove it. [ralf@linux-mips.org: Fold in followup fix from Thomas Gleixner.] Signed-off-by:
-
Thomas Gleixner authored
Use irq_set_handler_locked() as it avoids a redundant lookup of the irq descriptor. Search and replacement was done with coccinelle. Signed-off-by:
-
Thomas Gleixner authored
Hand in irq_data and avoid the redundant lookup of irq_desc. Originally-from: Jiang Liu <jiang.liu@linux.intel.com> Signed-off-by:
-
Jiang Liu authored
This is a preparatory patch for moving irq_data struct members. Signed-off-by:
Jiang Liu <jiang.liu@linux.intel.com> Signed-off-by:
-
Thomas Gleixner authored
Chained irq handlers usually set up handler data as well. We now have a function to set both under irq_desc->lock. Replace the two calls with one. Search and conversion was done with coccinelle. Reported-by:
Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by:
-
Thomas Gleixner authored
Chained irq handlers usually set up handler data as well. We now have a function to set both under irq_desc->lock. Replace the two calls with one. Search and conversion was done with coccinelle. Reported-by:
-
Thomas Gleixner authored
Chained irq handlers usually set up handler data as well. We now have a function to set both under irq_desc->lock. Replace the two calls with one. Search and conversion was done with coccinelle. Reported-by:
-
Thomas Gleixner authored
Chained irq handlers usually set up handler data as well. We now have a function to set both under irq_desc->lock. Replace the two calls with one. Search and conversion was done with coccinelle. Reported-by:
-
Markos Chandras authored
Set up the I6400 FTLB probability similar to P5600 and proAptiv. Signed-off-by:
Markos Chandras <markos.chandras@imgtec.com> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/10652/Signed-off-by:
-
Markos Chandras authored
R6 has dropped the MMUExtDef field from the config4 register and it now returns 0. However, the return value means nothing in that case and the only supported configuration for R6 is the VTLB+FTLB (MMUextDef == 3). As a result, rework the code so that the correct value is set for R6 cores. Signed-off-by:
-
Markos Chandras authored
Add a default case for the FTLB enable/disable code. This will be used to detect that something went wrong in the set_ftlb_enable() function either because that function knows nothing about the running core, or simply because the core can't turn its FTLB on/off. Signed-off-by:
-
Markos Chandras authored
We are so early in the boot process where we really don't want to stall and wait for CP0 FTLB related changes become visible so just drop the cp0 hazard barrier. Signed-off-by:
-
Markos Chandras authored
CM3 uses a 64-bit counter and compare registers so add support for them in the GIC counter interrupt. Signed-off-by:
-
Markos Chandras authored
Previously, the GIC accessors were only accessing u32 registers but newer CMs may actually be 64-bit on MIPS64 cores. As a result of which, extended these accessors to support 64-bit reads and writes. Signed-off-by:
-
Markos Chandras authored
The CM cache error reporting code is not Malta specific and as such it should live in the mips-cm.c file. Moreover, CM2 and CM3 differ in the way cache errors are being recorded to the registers so extend the previous code to add support for the CM3 as well. Signed-off-by:
-
Markos Chandras authored
The GCR CPC base register is 64-bit on 64-bit processors so use the appropriate field. Signed-off-by:
-
Markos Chandras authored
The CMGCRBase register (CP0, 15, 3) register is 64-bit on MIPS64 so we change its type to unsigned long. Signed-off-by:
-
Markos Chandras authored
Previously, the CM accessors were only accessing CM registers as u32 types instead of using the native CM register with. However, newer CMs may actually be 64-bit on MIPS64 cores. Fortunately, current 64-bit CMs (CM3) hold all the useful configuration bits in the lower half of the 64-bit registers (at least most of them) so they can still be accessed using the current 32-bit accessors. Signed-off-by:
-
Markos Chandras authored
Allow platforms to perform platform-specific steps before configuring the L2 cache. This is necessary for platforms with CM3 since the L2 parameters no longer live in the Config2 register. Signed-off-by:
-
Paul Burton authored
Detect the L2 cache configuration from GCR_L2_CONFIG when a CM3 is present in the system, rather than from Config2 which does not expose the L2 configuration on I6400. Signed-off-by:
-
Paul Burton authored
Provide accessor functions for the GCR_L2_CONFIG register introduced with CM3, and define the bits included in the register. Signed-off-by:
-
Paul Burton authored
Provide a function to trivially return the version of the CM present in the system, or 0 if no CM is present. The mips_cm_revision() will be used later on to determine the CM register width, so it must not use the regular CM accessors to read the revision register since that will lead to build failures due to recursive inlines. Signed-off-by:
-
Markos Chandras authored
R6 does not support the MIPS MT ASE and the CMP/SMP options so restrict them in order to prevent users from selecting incompatible SMP configuration for R6 cores. We also disable the CPS/SMP option because its support hasn't been added to the CPS code yet. Signed-off-by:
-
