[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH 2/8] x86/IDT: Collect IDT related content idt.h


  • To: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
  • From: Jan Beulich <jbeulich@xxxxxxxx>
  • Date: Tue, 25 Feb 2025 09:27:55 +0100
  • Autocrypt: addr=jbeulich@xxxxxxxx; keydata= xsDiBFk3nEQRBADAEaSw6zC/EJkiwGPXbWtPxl2xCdSoeepS07jW8UgcHNurfHvUzogEq5xk hu507c3BarVjyWCJOylMNR98Yd8VqD9UfmX0Hb8/BrA+Hl6/DB/eqGptrf4BSRwcZQM32aZK 7Pj2XbGWIUrZrd70x1eAP9QE3P79Y2oLrsCgbZJfEwCgvz9JjGmQqQkRiTVzlZVCJYcyGGsD /0tbFCzD2h20ahe8rC1gbb3K3qk+LpBtvjBu1RY9drYk0NymiGbJWZgab6t1jM7sk2vuf0Py O9Hf9XBmK0uE9IgMaiCpc32XV9oASz6UJebwkX+zF2jG5I1BfnO9g7KlotcA/v5ClMjgo6Gl MDY4HxoSRu3i1cqqSDtVlt+AOVBJBACrZcnHAUSuCXBPy0jOlBhxPqRWv6ND4c9PH1xjQ3NP nxJuMBS8rnNg22uyfAgmBKNLpLgAGVRMZGaGoJObGf72s6TeIqKJo/LtggAS9qAUiuKVnygo 3wjfkS9A3DRO+SpU7JqWdsveeIQyeyEJ/8PTowmSQLakF+3fote9ybzd880fSmFuIEJldWxp Y2ggPGpiZXVsaWNoQHN1c2UuY29tPsJgBBMRAgAgBQJZN5xEAhsDBgsJCAcDAgQVAggDBBYC AwECHgECF4AACgkQoDSui/t3IH4J+wCfQ5jHdEjCRHj23O/5ttg9r9OIruwAn3103WUITZee e7Sbg12UgcQ5lv7SzsFNBFk3nEQQCACCuTjCjFOUdi5Nm244F+78kLghRcin/awv+IrTcIWF hUpSs1Y91iQQ7KItirz5uwCPlwejSJDQJLIS+QtJHaXDXeV6NI0Uef1hP20+y8qydDiVkv6l IreXjTb7DvksRgJNvCkWtYnlS3mYvQ9NzS9PhyALWbXnH6sIJd2O9lKS1Mrfq+y0IXCP10eS FFGg+Av3IQeFatkJAyju0PPthyTqxSI4lZYuJVPknzgaeuJv/2NccrPvmeDg6Coe7ZIeQ8Yj t0ARxu2xytAkkLCel1Lz1WLmwLstV30g80nkgZf/wr+/BXJW/oIvRlonUkxv+IbBM3dX2OV8 AmRv1ySWPTP7AAMFB/9PQK/VtlNUJvg8GXj9ootzrteGfVZVVT4XBJkfwBcpC/XcPzldjv+3 HYudvpdNK3lLujXeA5fLOH+Z/G9WBc5pFVSMocI71I8bT8lIAzreg0WvkWg5V2WZsUMlnDL9 mpwIGFhlbM3gfDMs7MPMu8YQRFVdUvtSpaAs8OFfGQ0ia3LGZcjA6Ik2+xcqscEJzNH+qh8V m5jjp28yZgaqTaRbg3M/+MTbMpicpZuqF4rnB0AQD12/3BNWDR6bmh+EkYSMcEIpQmBM51qM EKYTQGybRCjpnKHGOxG0rfFY1085mBDZCH5Kx0cl0HVJuQKC+dV2ZY5AqjcKwAxpE75MLFkr wkkEGBECAAkFAlk3nEQCGwwACgkQoDSui/t3IH7nnwCfcJWUDUFKdCsBH/E5d+0ZnMQi+G0A nAuWpQkjM1ASeQwSHEeAWPgskBQL
  • Cc: Roger Pau Monné <roger.pau@xxxxxxxxxx>, Xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • Delivery-date: Tue, 25 Feb 2025 08:28:05 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

On 24.02.2025 17:05, Andrew Cooper wrote:
> Logic concerning the IDT is somewhat different to the other system tables, and
> in particular ought not to be in asm/processor.h.  Collect it together a new
> header.
> 
> While doing so, make a few minor adjustments:
> 
>  * Make set_ist() use volatile rather than ACCESS_ONCE(), as
>    _write_gate_lower() already does, removing the need for xen/lib.h.

While I don't mind this, I'd still like to mention that one of the first things
I was told when starting to work on Linux was to avoid volatile about 
everywhere.

> --- /dev/null
> +++ b/xen/arch/x86/include/asm/idt.h
> @@ -0,0 +1,125 @@
> +/* SPDX-License-Identifier: GPL-2.0-only */
> +#ifndef X86_ASM_IDT_H
> +#define X86_ASM_IDT_H
> +
> +#include <xen/bug.h>
> +#include <xen/types.h>
> +
> +#include <asm/x86-defns.h>
> +
> +#define IST_NONE 0
> +#define IST_MCE  1
> +#define IST_NMI  2
> +#define IST_DB   3
> +#define IST_DF   4
> +#define IST_MAX  4
> +
> +typedef union {
> +    struct {
> +        uint64_t a, b;
> +    };
> +    struct {
> +        uint16_t addr0;
> +        uint16_t cs;
> +        uint8_t  ist; /* :3, 5 bits rsvd, but this yields far better code. */
> +        uint8_t  type:4, s:1, dpl:2, p:1;
> +        uint16_t addr1;
> +        uint32_t addr2;
> +        /* 32 bits rsvd. */
> +    };
> +} idt_entry_t;
> +
> +#define IDT_ENTRIES 256
> +extern idt_entry_t idt_table[];
> +extern idt_entry_t *idt_tables[];
> +
> +/*
> + * Set the Interrupt Stack Table used by a particular IDT entry.  Typically
> + * used on a live IDT, so volatile to disuade clever optimisations.
> + */
> +static inline void set_ist(volatile idt_entry_t *idt, unsigned int ist)
> +{
> +    /* IST is a 3 bit field, 32 bits into the IDT entry. */
> +    ASSERT(ist <= IST_MAX);
> +
> +    idt->ist = ist;
> +}
> +
> +static inline void enable_each_ist(idt_entry_t *idt)
> +{
> +    set_ist(&idt[X86_EXC_DF],  IST_DF);
> +    set_ist(&idt[X86_EXC_NMI], IST_NMI);
> +    set_ist(&idt[X86_EXC_MC],  IST_MCE);
> +    set_ist(&idt[X86_EXC_DB],  IST_DB);
> +}
> +
> +static inline void disable_each_ist(idt_entry_t *idt)
> +{
> +    set_ist(&idt[X86_EXC_DF],  IST_NONE);
> +    set_ist(&idt[X86_EXC_NMI], IST_NONE);
> +    set_ist(&idt[X86_EXC_MC],  IST_NONE);
> +    set_ist(&idt[X86_EXC_DB],  IST_NONE);
> +}
> +
> +/*
> + * Write the lower 64 bits of an IDT Entry. This relies on the upper 32
> + * bits of the address not changing, which is a safe assumption as all
> + * functions we are likely to load will live inside the 1GB
> + * code/data/bss address range.
> + */
> +static inline void _write_gate_lower(volatile idt_entry_t *gate,
> +                                     const idt_entry_t *new)
> +{
> +    ASSERT(gate->b == new->b);
> +    gate->a = new->a;
> +}

Would this better move down a few lines, immediately ahead of its two
use sites?

> +#define _set_gate(gate_addr,type,dpl,addr)               \

Moving this is questionable, as gates aren't limited to the IDT (in
principle; yes, we don't use call gates ourselves). However, as you
move it, my minimal request would be to add the missing blanks here.
Beyond that I wonder ...

> +do {                                                     \
> +    (gate_addr)->a = 0;                                  \
> +    smp_wmb(); /* disable gate /then/ rewrite */         \
> +    (gate_addr)->b =                                     \
> +        ((unsigned long)(addr) >> 32);                   \
> +    smp_wmb(); /* rewrite /then/ enable gate */          \
> +    (gate_addr)->a =                                     \
> +        (((unsigned long)(addr) & 0xFFFF0000UL) << 32) | \
> +        ((unsigned long)(dpl) << 45) |                   \
> +        ((unsigned long)(type) << 40) |                  \
> +        ((unsigned long)(addr) & 0xFFFFUL) |             \
> +        ((unsigned long)__HYPERVISOR_CS << 16) |         \
> +        (1UL << 47);                                     \
> +} while (0)

... whether using the other half of the union would allow this to
become a little more readable. (Then it would also rightfully live
here, seeing that the union is typedef-ed to idt_entry_t.) This then
may also extend to ...

> +static inline void _set_gate_lower(idt_entry_t *gate, unsigned long type,
> +                                   unsigned long dpl, void *addr)
> +{
> +    idt_entry_t idte;
> +    idte.b = gate->b;
> +    idte.a =
> +        (((unsigned long)(addr) & 0xFFFF0000UL) << 32) |
> +        ((unsigned long)(dpl) << 45) |
> +        ((unsigned long)(type) << 40) |
> +        ((unsigned long)(addr) & 0xFFFFUL) |
> +        ((unsigned long)__HYPERVISOR_CS << 16) |
> +        (1UL << 47);

... here and ...

> +    _write_gate_lower(gate, &idte);
> +}
> +
> +/*
> + * Update the lower half handler of an IDT entry, without changing any other
> + * configuration.
> + */
> +static inline void _update_gate_addr_lower(idt_entry_t *gate, void *addr)
> +{
> +    idt_entry_t idte;
> +    idte.a = gate->a;
> +
> +    idte.b = ((unsigned long)(addr) >> 32);
> +    idte.a &= 0x0000FFFFFFFF0000ULL;
> +    idte.a |= (((unsigned long)(addr) & 0xFFFF0000UL) << 32) |
> +        ((unsigned long)(addr) & 0xFFFFUL);

... here. Otoh you may have reasons to keep these like they are?

Could both _set_gate_lower() and _update_gate_addr_lower() have their
last parameters each be switched to pointer-to-const (they supposedly point
into .text after all)?

Jan



 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.