[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [PATCH v2 18/40] xen/mpu: introduce helper access_protection_region
Hi Penny, On 13/01/2023 05:28, Penny Zheng wrote: Each EL2 MPU protection region could be configured using PRBAR<n>_EL2 and PRLAR<n>_EL2. This commit introduces a new helper access_protection_region() to access EL2 MPU protection region, including both read/write operations. As explained in section G1.3.18 of the reference manual for AArch64v8R, a set of system register PRBAR<n>_EL2 and PRLAR<n>_EL2 provide access to the EL2 MPU region which is determined by the value of 'n' and PRSELR_EL2.REGION as PRSELR_EL2.REGION<7:4>:n.(n = 0, 1, 2, ... , 15) For example to access regions from 16 to 31: - Set PRSELR_EL2 to 0b1xxxx - Region 16 configuration is accessible through PRBAR0_EL2 and PRLAR0_EL2 - Region 17 configuration is accessible through PRBAR1_EL2 and PRLAR1_EL2 - Region 18 configuration is accessible through PRBAR2_EL2 and PRLAR2_EL2 - ... - Region 31 configuration is accessible through PRBAR15_EL2 and PRLAR15_EL2 Signed-off-by: Penny Zheng <penny.zheng@xxxxxxx> Signed-off-by: Wei Chen <wei.chen@xxxxxxx> --- xen/arch/arm/mm_mpu.c | 151 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 151 insertions(+) diff --git a/xen/arch/arm/mm_mpu.c b/xen/arch/arm/mm_mpu.c index c9e17ab6da..f2b494449c 100644 --- a/xen/arch/arm/mm_mpu.c +++ b/xen/arch/arm/mm_mpu.c @@ -46,6 +46,157 @@ uint64_t __ro_after_init next_transient_region_idx; /* Maximum number of supported MPU memory regions by the EL2 MPU. */ uint64_t __ro_after_init max_xen_mpumap;+/* Write a MPU protection region */+#define WRITE_PROTECTION_REGION(sel, pr, prbar_el2, prlar_el2) ({ \ + uint64_t _sel = sel; \ + const pr_t *_pr = pr; \ + asm volatile( \ + "msr "__stringify(PRSELR_EL2)", %0;" /* Selects the region */ \ This is an open-coding version of WRITE_SYSREG(). Can we use it instead? + "dsb sy;" \ What is this dsb for? Also is the 'sy' really necessary? + "msr "__stringify(prbar_el2)", %1;" /* Write PRBAR<n>_EL2 */ \ WRITE_SYSREG()? + "msr "__stringify(prlar_el2)", %2;" /* Write PRLAR<n>_EL2 */ \ WRITE_SYSREG()? + "dsb sy;" \ Same about dsb. But I would consider to move the dsb and selection part outside of the macro. So they could outside of the switch and reduce the code generation. + : : "r" (_sel), "r" (_pr->prbar.bits), "r" (_pr->prlar.bits)); \ +}) + +/* Read a MPU protection region */ +#define READ_PROTECTION_REGION(sel, prbar_el2, prlar_el2) ({ \ My comment on READ_PROTECTION also applies here. But you would want to use READ_SYSREG() for 'mrs'. + uint64_t _sel = sel; \ + pr_t _pr; \ + asm volatile( \ + "msr "__stringify(PRSELR_EL2)", %2;" /* Selects the region */ \ + "dsb sy;" \ + "mrs %0, "__stringify(prbar_el2)";" /* Read PRBAR<n>_EL2 */ \ + "mrs %1, "__stringify(prlar_el2)";" /* Read PRLAR<n>_EL2 */ \ + "dsb sy;" \ + : "=r" (_pr.prbar.bits), "=r" (_pr.prlar.bits) : "r" (_sel)); \ + _pr; \ +}) + +/* + * Access MPU protection region, including both read/write operations. + * Armv8-R AArch64 at most supports 255 MPU protection regions. + * See section G1.3.18 of the reference manual for Armv8-R AArch64, + * PRBAR<n>_EL2 and PRLAR<n>_EL2 provide access to the EL2 MPU region + * determined by the value of 'n' and PRSELR_EL2.REGION as + * PRSELR_EL2.REGION<7:4>:n(n = 0, 1, 2, ... , 15) + * For example to access regions from 16 to 31 (0b10000 to 0b11111): + * - Set PRSELR_EL2 to 0b1xxxx + * - Region 16 configuration is accessible through PRBAR0_ELx and PRLAR0_ELx + * - Region 17 configuration is accessible through PRBAR1_ELx and PRLAR1_ELx + * - Region 18 configuration is accessible through PRBAR2_ELx and PRLAR2_ELx + * - ... + * - Region 31 configuration is accessible through PRBAR15_ELx and PRLAR15_ELx + * + * @read: if it is read operation. + * @pr_read: mpu protection region returned by read op. + * @pr_write: const mpu protection region passed through write op. + * @sel: mpu protection region selector + */ +static void access_protection_region(bool read, pr_t *pr_read, + const pr_t *pr_write, uint64_t sel) I would rather prefer if we introduce two helpers (one for the read operation, the other for the write operation). This would make the code a bit easier to read. +{ + switch ( sel & 0xf ) + { + case 0: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR0_EL2, PRLAR0_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR0_EL2, PRLAR0_EL2); + break; + case 1: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR1_EL2, PRLAR1_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR1_EL2, PRLAR1_EL2); + break; + case 2: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR2_EL2, PRLAR2_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR2_EL2, PRLAR2_EL2); + break; + case 3: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR3_EL2, PRLAR3_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR3_EL2, PRLAR3_EL2); + break; + case 4: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR4_EL2, PRLAR4_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR4_EL2, PRLAR4_EL2); + break; + case 5: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR5_EL2, PRLAR5_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR5_EL2, PRLAR5_EL2); + break; + case 6: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR6_EL2, PRLAR6_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR6_EL2, PRLAR6_EL2); + break; + case 7: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR7_EL2, PRLAR7_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR7_EL2, PRLAR7_EL2); + break; + case 8: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR8_EL2, PRLAR8_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR8_EL2, PRLAR8_EL2); + break; + case 9: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR9_EL2, PRLAR9_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR9_EL2, PRLAR9_EL2); + break; + case 10: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR10_EL2, PRLAR10_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR10_EL2, PRLAR10_EL2); + break; + case 11: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR11_EL2, PRLAR11_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR11_EL2, PRLAR11_EL2); + break; + case 12: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR12_EL2, PRLAR12_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR12_EL2, PRLAR12_EL2); + break; + case 13: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR13_EL2, PRLAR13_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR13_EL2, PRLAR13_EL2); + break; + case 14: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR14_EL2, PRLAR14_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR14_EL2, PRLAR14_EL2); + break; + case 15: + if ( read ) + *pr_read = READ_PROTECTION_REGION(sel, PRBAR15_EL2, PRLAR15_EL2); + else + WRITE_PROTECTION_REGION(sel, pr_write, PRBAR15_EL2, PRLAR15_EL2); + break; What if the caller pass a number higher than 15? + } +} + /* TODO: Implementation on the first usage */ void dump_hyp_walk(vaddr_t addr) { Cheers, -- Julien Grall
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |