- 13 May, 2016 40 commits
-
-
James Hogan authored
Performing an MTHC0 instruction without XPA being present will trigger a reserved instruction exception, therefore conditionalise the use of this instruction when building TLB handlers (build_update_entries()), and in __update_tlb(). This allows an XPA kernel to run on non XPA hardware without that instruction implemented, just like it can run on XPA capable hardware without XPA in use (with the noxpa kernel argument) or with XPA not configured in hardware. [paul.burton@imgtec.com: - Rebase atop other TLB work. - Add "mm" to subject. - Handle the __kmap_pgprot case.] Fixes: c5b36783 ("MIPS: Add support for XPA.") Signed-off-by:
James Hogan <james.hogan@imgtec.com> Signed-off-by:
Paul Burton <paul.burton@imgtec.com> Cc: Paul Gortmaker <paul.gortmaker@windriver.com> Cc: David Hildenbrand <dahi@linux.vnet.ibm.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: linux-mips@linux-mips.org Cc: linux-kernel@vger.kernel.org Patchwork: https://patchwork.linux-mips.org/patch/13124/Signed-off-by:
Ralf Baechle <ralf@linux-mips.org>
-
Paul Burton authored
We can simplify build_update_entries by unifying the code for the 36 bit physical addressing with MIPS32 case with the general case, by using pte_off_ variables in all cases & handling the trivial _PAGE_GLOBAL_SHIFT == 0 case in build_convert_pte_to_entrylo. This leaves XPA as the only special case. Signed-off-by:
-
Paul Burton authored
The XPA case in iPTE_SW or's in software mode bits to the pte_low value (which is what actually ends up in the high 32 bits of EntryLo...). It does this presuming that only bits in the upper 16 bits of the 32 bit pte_low value will be set. Make this assumption explicit with a BUG_ON. A similar assumption is made for the hardware mode bits, which are or'd in with a single ori instruction. Make that assumption explicit with a BUG_ON too. Signed-off-by:
-
Paul Burton authored
Rather than hardcode a scratch register for the XPA case in iPTE_SW, pass one through from the work registers allocated by the caller. This allows for the XPA path to function correctly regardless of the work registers in use. Without doing this there are cases (where KScratch registers are unavailable) in which iPTE_SW will incorrectly clobber $1 despite it already being in use for the PTE or PTE pointer. Signed-off-by:
-
James Hogan authored
For XPA kernels build_update_entries() uses $1 (at) as a scratch register, but doesn't arrange for it to be preserved, so it will always be clobbered by the TLB refill exception. Although this register normally has a very short lifetime that doesn't cross memory accesses, TLB refills due to instruction fetches (either on a page boundary or after preemption) could clobber live data, and its easy to reproduce the clobber with a little bit of assembler code. Note that the use of a hardware page table walker will partly mask the problem, as the TLB refill handler will not always be invoked. This is fixed by avoiding the use of the extra scratch register. The pte_high parts (going into the lower half of the EntryLo registers) are loaded and manipulated separately so as to keep the PTE pointer around for the other halves (instead of storing in the scratch register), and the pte_low parts (going into the high half of the EntryLo registers) are masked with 0x00ffffff using an ext instruction (instead of loading 0x00ffffff into the scratch register and AND'ing). [paul.burton@imgtec.com: - Rebase atop other TLB work. - Use ext instead of an sll, srl sequence. - Use cpu_has_xpa instead of #ifdefs. - Modify commit subject to include "mm".] Fixes: c5b36783 ("MIPS: Add support for XPA.") Signed-off-by:
-
Paul Burton authored
There are 2 distinct cases in which a kernel for a MIPS32 CPU (CONFIG_CPU_MIPS32=y) may use 64 bit physical addresses (CONFIG_PHYS_ADDR_T_64BIT=y): - 36 bit physical addressing as used by RMI Alchemy & Netlogic XLP/XLR CPUs. - MIPS32r5 eXtended Physical Addressing (XPA). These 2 cases are distinct in that they require different behaviour from the kernel - the EntryLo registers have different formats. Until Linux v4.1 we only supported the first case, with code conditional upon the 2 aforementioned Kconfig variables being set. Commit c5b36783 ("MIPS: Add support for XPA.") added support for the second case, but did so by modifying the code that existed for the first case rather than treating the 2 cases as distinct. Since the EntryLo registers have different formats this breaks the 36 bit Alchemy/XLP/XLR case. Fix this by splitting the 2 cases, with XPA cases now being conditional upon CONFIG_XPA and the non-XPA case matching the code as it existed prior to commit c5b36783 ("MIPS: Add support for XPA."). Signed-off-by:Manuel Lauss <manuel.lauss@gmail.com> Tested-by:
-
Paul Burton authored
The same definition for pte_page is duplicated for the MIPS32 PHYS_ADDR_T_64BIT case & the generic case. Unify them by moving a single definition outside of preprocessor conditionals. Signed-off-by:
-
Paul Burton authored
Ever since support for RI/XI was implemented by commit 6dd9344c ("MIPS: Implement Read Inhibit/eXecute Inhibit") we've had a mixture of _PAGE_READ & _PAGE_NO_READ bits. Rather than keep both around, switch away from using _PAGE_READ to determine page presence & instead invert the use to _PAGE_NO_READ. Wherever we formerly had no definition for _PAGE_NO_READ, change what was _PAGE_READ to _PAGE_NO_READ. The end result is that we consistently use _PAGE_NO_READ to determine whether a page is readable, regardless of whether RI/XI is implemented. Signed-off-by:
-
Paul Burton authored
asm/pgtable-bits.h has grown to become an unreadable mess of #ifdef directives defining bits conditionally upon other bits all at the preprocessing stage, for no good reason. Instead of having quite so many #ifdef's, simply use enums to provide sequential numbering for bit shifts, without having to keep track manually of what the last bit defined was. Masks are defined separately, after the shifts, which allows for most of their definitions to be reused for all systems rather than duplicated. This patch is not intended to make any behavioural change to the code - all bits should be used in the same way they were before this patch. Signed-off-by:
-
Paul Burton authored
asm/pgtable-bits.h is included in 2 assembly files and thus has to ifdef around C code, however nothing defined by the header is used in either of the assembly files that include it. Remove the redundant inclusions such that asm/pgtable-bits.h doesn't need to #ifdef around C code, for cleanliness and in preparation for later patches which will add more C. Signed-off-by:
-
James Hogan authored
The hardware page table walker (HTW) configuration is broken on XPA kernels where XPA couldn't be enabled (either nohtw or the hardware doesn't support it). This is because the PWSize.PTEW field (PTE width) was only set to 8 bytes (an extra shift of 1) in config_htw_params() if PageGrain.ELPA (enable large physical addressing) is set. On an XPA kernel though the size of PTEs is fixed at 8 bytes regardless of whether XPA could actually be enabled. Fix the initialisation of this field based on sizeof(pte_t) instead. Fixes: c5b36783 ("MIPS: Add support for XPA.") Signed-off-by:
-
James Hogan authored
XPA (eXtended Physical Addressing) should be detected as a combination of two architectural features: - Large Physical Address (as per Config3.LPA). With XPA this will be set on MIPS32r5 cores, but it may also be set for MIPS64r2 cores too. - MTHC0/MFHC0 instructions (as per Config5.MVH). With XPA this will be set, but it may also be set in VZ guest context even when Config3.LPA in the guest context has been cleared by the hypervisor. As such, XPA is only usable if both bits are set. Update CPU features to separate these two features, with cpu_has_xpa requiring both to be set. Signed-off-by:
-
Paul Burton authored
When emulating a jalr instruction with rd == $0, the code in isBranchInstr was incorrectly writing to GPR $0 which should actually always remain zeroed. This would lead to any further instructions emulated which use $0 operating on a bogus value until the task is next context switched, at which point the value of $0 in the task context would be restored to the correct zero by a store in SAVE_SOME. Fix this by not writing to rd if it is $0. Fixes: 102cedc3 ("MIPS: microMIPS: Floating point support.") Signed-off-by:
-
Paul Burton authored
The code in _sp_maddf (formerly ieee754sp_madd) appears to have been copied verbatim from ieee754sp_add, and although it's adding the unpacked "r" & "z" floats it kept using macros that operate on "x" & "y". This led to the addition being carried out incorrectly on some mismash of the product, accumulator & multiplicand fields. Typically this would lead to the assertions "ze == re" & "ze <= SP_EMAX" failing since ze & re hadn't been operated upon. Signed-off-by:
-
Paul Burton authored
A line incrementing the re variable was indented a level too deep in ieee754dp_mul, making the code unclear to read. Fix the indentation. This appears to have been copied verbatim along with the rest of the multiplication code to ieee754dp_maddf, now _dp_maddf, too so fix the indentation there too. Signed-off-by:
-
Paul Burton authored
A comment in ieee754dp_mul indicates that the code is about to perform a 32b x 32b multiplication & keep the high 32b of the result. It appears this was copied from the single-precision multiplication code, since the code actually goes on to perform a 64b x 64b multiplication & keep the high 64b of the result. Fix the comment to indicate 64b. It appears also that this comment was copied verbatim along with the rest of the multiplication code into ieee754dp_maddf, which has since been renamed _dp_maddf. Fix the same issue there. Signed-off-by:
-
Paul Burton authored
Introduce macros for handling the "z" argument to maddf & msubf, making its handling consistent with that of the "x" & "y" arguments rather than open-coding equivalents. Signed-off-by:
-
Paul Burton authored
The code for emulating MIPSr6 madd.d & msub.d instructions has previously been implemented as 2 different functions, namely ieee754dp_maddf & ieee754dp_msubf. The difference in behaviour of these 2 instructions is merely the sign of the product, so we can easily share the code implementing them. Do this for the double precision variant, removing the original ieee754dp_msubf in favor of reusing the code from ieee754dp_maddf. Signed-off-by:
-
Paul Burton authored
The code for emulating MIPSr6 madd.s & msub.s instructions has previously been implemented as 2 different functions, namely ieee754sp_maddf & ieee754sp_msubf. The difference in behaviour of these 2 instructions is merely the sign of the product, so we can easily share the code implementing them. Do this for the single precision variant, removing the original ieee754sp_msubf in favor of reusing the code from ieee754sp_maddf. Signed-off-by:
-
Paul Burton authored
Add support for emulating the MIPSr6 sel.fmt instruction, which was previously missing from the FPU emulation code. This instruction selects its result from 2 possible source registers, based upon bit 0 of the destination register, and is valid only for S (single) & D (double) data types. Signed-off-by:
-
Paul Burton authored
Declare the opcode for the MIPSr6 sel.fmt instruction, as fsel_op in order to match other FP op names. Signed-off-by:
-
Paul Burton authored
The conditions for branching when emulating the BC1EQZ & BC1NEZ instructions were backwards, leading to each of those instructions being treated as the other. Fix this by reversing the conditions, and clear up the code a little for readability & checkpatch. Fixes: c8a34581 ("MIPS: Emulate the BC1{EQ,NE}Z FPU instructions") Signed-off-by:
-
Paul Burton authored
The conditions for branching when emulating the BC1EQZ & BC1NEZ instructions were backwards, leading to each of those instructions being treated as the other. Fix this by reversing the conditions, and clear up the code a little for readability & checkpatch. Fixes: c909ca71 ("MIPS: math-emu: Emulate missing BC1{EQ,NE}Z instructions") Signed-off-by:
-
Rabin Vincent authored
Allow the builtin command line to be extended by what the bootloader passes in. For example, the bootloader can pass specific arguments depending on the boot mode, and these should override the defaults in the builtin cmdline. The default MIPS_CMDLINE_FROM_BOOTLOADER option prepends the bootloader's cmdline to the builtin cmdline so is not suitable for this purpose. Signed-off-by:
Rabin Vincent <rabinv@axis.com> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/13181/Signed-off-by:
-
Rabin Vincent authored
The leading spaces don't affect functionality but are unnecessary. Signed-off-by:
-
James Hogan authored
Various branches and jumps in noreorder parts of genex.S don't have their delay slot instructions indented conventionally with the extra space. Fix these, as well as various other inconsistent whitespace problems in this file, such as spaces used after some opcodes instead of a tab. Signed-off-by:
-
Paul Burton authored
Add support for extended ASIDs as determined by the Config4.AE bit. Since the only supported CPUs known to implement this are Netlogic XLP and MIPS I6400, select this variable ASID support based upon CONFIG_CPU_XLP and CONFIG_CPU_MIPSR6. Signed-off-by:
-
Paul Burton authored
In preparation for supporting variable ASID masks, retrieve ASID masks using functions in asm/cpu-info.h which accept struct cpuinfo_mips. This will allow those functions to determine the ASID mask based upon the CPU in a later patch. This also allows for the r3k & r8k cases to be handled in Kconfig, which is arguably cleaner than the previous #ifdefs. Signed-off-by:
-
James Hogan authored
Now that the at register ($1) is no longer saved by __kvm_mips_vcpu_run(), relax the noat assembler directive so that it only applies around code where at is restored before entering guest, and saved after exiting guest. Signed-off-by:
-
James Hogan authored
Update __kvm_mips_vcpu_run() to only save and restore callee saved registers. It is always called using the standard ABIs, so the caller will preserve any other registers that need preserving. Signed-off-by:
-
Paul Burton authored
In preparation for supporting varied widths of ASID mask in the kernel in general, switch KVM's guest ASIDs to a new KVM_ENTRYHI_ASID definition based on the 8-bit MIPS_ENTRYHI_ASID instead of ASID_MASK. It could potentially be used to support extended guest ASIDs in the future. Signed-off-by:
-
James Hogan authored
Add definitions for the ASID field in CP0_EntryHi (along with the soon to be used ASIDX field), and use them in a few previously hardcoded cases. Signed-off-by:
-
James Hogan authored
MIPS KVM uses different ASIDs for guest execution than for the host. The host ASID is saved on the stack when entering the guest with __kvm_mips_vcpu_run(), and restored again before returning back to the caller (exit to userland). - This does not take into account that pre-emption may have taken place during that time, which may have started a new ASID cycle and resulted in that process' ASID being invalidated and reused. - This does not take into account that the process may have migrated to a different CPU during that time, with a different ASID assignment since they are managed per-CPU. - It is actually redundant, since the host ASID will be restored correctly by kvm_arch_vcpu_put(), which is called almost immediately after kvm_arch_vcpu_ioctl_run() returns. Therefore drop this code from locore.S Signed-off-by:
-
John Crispin authored
The old address is no longer valid. Use the my new one instead. Signed-off-by:
John Crispin <john@phrozen.org> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/13201/Signed-off-by:
-
John Crispin authored
The old address is no longer valid. Use the my new one instead. Signed-off-by:
-
Paul Burton authored
Release 6 of the MIPS architecture introduced the bitswap instruction, which reverses the bits within each byte of a word. Make use of this instruction to implement the __arch_bitrev* functions, which should be faster for most MIPSr6 CPUs, reduces code size slightly and allows us to avoid the lookup table used by the generic implementation, saving 256 bytes in the kernel binary by dropping that. Signed-off-by:
-
Anna-Maria Gleixner authored
Since commit 3b9d6da6 ("cpu/hotplug: Fix rollback during error-out in __cpu_disable()") it is ensured that callbacks of CPU_ONLINE and CPU_DOWN_PREPARE are processed on the hotplugged CPU. Due to this work_on_cpu() calls are no longer required. Replace work_on_cpu() with a direct call of mips_cdmm_bus_up() or mips_cdmm_bus_down(). Description of those functions are adapted. Signed-off-by:
Anna-Maria Gleixner <anna-maria@linutronix.de> Cc: linux-mips@linux-mips.org Cc: linux-kernel@vger.kernel.org Cc: rt@linutronix.de Patchwork: https://patchwork.linux-mips.org/patch/13197/Signed-off-by:
-
Sergey Ryazanov authored
No one of supported MIPS machines has an IOMMU unit, so we can safely define PCI_DMA_BUS_IS_PHYS = 1. Also remove iommu flag from the pci controller structure, since it is useless. Signed-off-by:
Sergey Ryazanov <ryazanov.s.a@gmail.com> Cc: Linux MIPS <linux-mips@linux-mips.org> Patchwork: https://patchwork.linux-mips.org/patch/7604/Signed-off-by:
-
Leonid Yegoshin authored
SEGBITS is 40 bits or more, depending on CPU type. Introduces optional support for 48 bits of application virtual address space. Only 16K and 64K pages are supported. Enabling will result in a memory overhead of a small number of pages for small applications. For 64K pages a 3rd level of page tables is required which has some impact during software TLB refill. [ralf@linux-mips.org: Fixed things raised in the review of the version posted and changed kconfig to be a bit more userfriendly.] Signed-off-by:
Leonid Yegoshin <Leonid.Yegoshin@imgtec.com> Cc: aleksey.makarov@auriga.com Cc: james.hogan@imgtec.com Cc: paul.burton@imgtec.com Cc: david.daney@cavium.com Cc: peterz@infradead.org Cc: linux-mips@linux-mips.org Cc: linux-kernel@vger.kernel.org Cc: davidlohr@hp.com Cc: kirill@shutemov.name Cc: akpm@linux-foundation.org Cc: mingo@kernel.org Patchwork: https://patchwork.linux-mips.org/patch/10051/Signed-off-by:
-
Leonid Yegoshin authored
MIPS32 o32 ABI sigaction() processing on MIPS64 n64 kernel was incorrectly set to processing aka rt_sigaction() variant only. Signed-off-by:
-
