freebsd-dev/stand/kboot/hostdisk.c
Warner Losh 5328f9034f stand/kboot: Simplify
There's plenty of stack in kboot, so use it here rather than the
malloc/free dance.

Sponsored by:		Netflix
Reviewed by:		tsoome, kevans
Differential Revision:	https://reviews.freebsd.org/D39416
2023-05-01 15:02:53 -06:00

630 lines
14 KiB
C

/*-
* Copyright (C) 2014 Nathan Whitehorn
* All rights reserved.
* Copyright 2022 Netflix, Inc
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL TOOLS GMBH BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
* OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
* WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
* OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
* ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/types.h>
#include <sys/disk.h>
#include <stdarg.h>
#include <paths.h>
#include "host_syscall.h"
#include "kboot.h"
#include "bootstrap.h"
#ifdef LOADER_ZFS_SUPPORT
#include "libzfs.h"
#include <sys/zfs_bootenv.h>
#endif
static int hostdisk_init(void);
static int hostdisk_strategy(void *devdata, int flag, daddr_t dblk,
size_t size, char *buf, size_t *rsize);
static int hostdisk_open(struct open_file *f, ...);
static int hostdisk_close(struct open_file *f);
static int hostdisk_ioctl(struct open_file *f, u_long cmd, void *data);
static int hostdisk_print(int verbose);
static char *hostdisk_fmtdev(struct devdesc *vdev);
static bool hostdisk_match(struct devsw *devsw, const char *devspec);
static int hostdisk_parsedev(struct devdesc **idev, const char *devspec, const char **path);
struct devsw hostdisk = {
.dv_name = "/dev",
.dv_type = DEVT_HOSTDISK,
.dv_init = hostdisk_init,
.dv_strategy = hostdisk_strategy,
.dv_open = hostdisk_open,
.dv_close = hostdisk_close,
.dv_ioctl = hostdisk_ioctl,
.dv_print = hostdisk_print,
.dv_cleanup = nullsys,
.dv_fmtdev = hostdisk_fmtdev,
.dv_match = hostdisk_match,
.dv_parsedev = hostdisk_parsedev,
};
/*
* We need to walk through the /sys/block directories looking for
* block devices that we can use.
*/
#define SYSBLK "/sys/block"
#define HOSTDISK_MIN_SIZE (16ul << 20) /* 16MB */
typedef STAILQ_HEAD(, hdinfo) hdinfo_list_t;
typedef struct hdinfo {
STAILQ_ENTRY(hdinfo) hd_link; /* link in device list */
hdinfo_list_t hd_children;
struct hdinfo *hd_parent;
const char *hd_dev;
uint64_t hd_size; /* In bytes */
uint64_t hd_sectors;
uint64_t hd_sectorsize;
int hd_flags;
#define HDF_HAS_ZPOOL 1 /* We found a zpool here and uuid valid */
uint64_t hd_zfs_uuid;
} hdinfo_t;
#define dev2hd(d) ((hdinfo_t *)d->d_opendata)
#define hd_name(hd) ((hd->hd_dev + 5))
static hdinfo_list_t hdinfo = STAILQ_HEAD_INITIALIZER(hdinfo);
typedef bool fef_cb_t(struct host_dirent64 *, void *);
#define FEF_RECURSIVE 1
static bool
foreach_file(const char *dir, fef_cb_t cb, void *argp, u_int flags)
{
char dents[2048];
int fd, dentsize;
struct host_dirent64 *dent;
fd = host_open(dir, O_RDONLY, 0);
if (fd < 0) {
printf("Can't open %s\n", dir);/* XXX */
return (false);
}
while (1) {
dentsize = host_getdents64(fd, dents, sizeof(dents));
if (dentsize <= 0)
break;
for (dent = (struct host_dirent64 *)dents;
(char *)dent < dents + dentsize;
dent = (struct host_dirent64 *)((void *)dent + dent->d_reclen)) {
if (!cb(dent, argp))
break;
}
}
host_close(fd);
return (true);
}
static void
hostdisk_add_part(hdinfo_t *hd, const char *drv, uint64_t secs)
{
hdinfo_t *md;
char *dev;
printf("hd %s adding %s %ju\n", hd->hd_dev, drv, (uintmax_t)secs);
if ((md = calloc(1, sizeof(*md))) == NULL)
return;
if (asprintf(&dev, "/dev/%s", drv) == -1) {
printf("hostdisk: no memory\n");
free(md);
return;
}
md->hd_dev = dev;
md->hd_sectors = secs;
md->hd_sectorsize = hd->hd_sectorsize;
md->hd_size = md->hd_sectors * md->hd_sectorsize;
md->hd_parent = hd;
STAILQ_INSERT_TAIL(&hd->hd_children, md, hd_link);
}
static bool
hostdisk_one_part(struct host_dirent64 *dent, void *argp)
{
hdinfo_t *hd = argp;
char szfn[1024];
uint64_t sz;
/* Need to skip /dev/ at start of hd_name */
if (strncmp(dent->d_name, hd_name(hd), strlen(hd_name(hd))) != 0)
return (true);
/* Find out how big this is -- no size not a disk */
snprintf(szfn, sizeof(szfn), "%s/%s/%s/size", SYSBLK,
hd_name(hd), dent->d_name);
if (!file2u64(szfn, &sz))
return true;
hostdisk_add_part(hd, dent->d_name, sz);
return true;
}
static void
hostdisk_add_parts(hdinfo_t *hd)
{
char fn[1024];
snprintf(fn, sizeof(fn), "%s/%s", SYSBLK, hd_name(hd));
foreach_file(fn, hostdisk_one_part, hd, 0);
}
static void
hostdisk_add_drive(const char *drv, uint64_t secs)
{
hdinfo_t *hd = NULL;
char *dev = NULL;
char fn[1024];
if ((hd = calloc(1, sizeof(*hd))) == NULL)
return;
if (asprintf(&dev, "/dev/%s", drv) == -1) {
printf("hostdisk: no memory\n");
free(hd);
return;
}
hd->hd_dev = dev;
hd->hd_sectors = secs;
snprintf(fn, sizeof(fn), "%s/%s/queue/hw_sector_size",
SYSBLK, drv);
if (!file2u64(fn, &hd->hd_sectorsize))
goto err;
hd->hd_size = hd->hd_sectors * hd->hd_sectorsize;
if (hd->hd_size < HOSTDISK_MIN_SIZE)
goto err;
hd->hd_flags = 0;
STAILQ_INIT(&hd->hd_children);
printf("/dev/%s: %ju %ju %ju\n",
drv, hd->hd_size, hd->hd_sectors, hd->hd_sectorsize);
STAILQ_INSERT_TAIL(&hdinfo, hd, hd_link);
hostdisk_add_parts(hd);
return;
err:
free(dev);
free(hd);
return;
}
/* Find a disk / partition by its filename */
static hdinfo_t *
hostdisk_find(const char *fn)
{
hdinfo_t *hd, *md;
STAILQ_FOREACH(hd, &hdinfo, hd_link) {
if (strcmp(hd->hd_dev, fn) == 0)
return (hd);
STAILQ_FOREACH(md, &hd->hd_children, hd_link) {
if (strcmp(md->hd_dev, fn) == 0)
return (md);
}
}
return (NULL);
}
static bool
hostdisk_one_disk(struct host_dirent64 *dent, void *argp __unused)
{
char szfn[1024];
uint64_t sz;
/*
* Skip . and ..
*/
if (strcmp(dent->d_name, ".") == 0 ||
strcmp(dent->d_name, "..") == 0)
return (true);
/* Find out how big this is -- no size not a disk */
snprintf(szfn, sizeof(szfn), "%s/%s/size", SYSBLK,
dent->d_name);
if (!file2u64(szfn, &sz))
return (true);
hostdisk_add_drive(dent->d_name, sz);
return (true);
}
static void
hostdisk_fake_one_disk(char *override)
{
hdinfo_t *hd = NULL;
struct host_kstat sb;
if (host_stat(override, &sb) != 0)
return;
if (!HOST_S_ISREG(sb.st_mode))
return;
if (sb.st_size == 0)
return;
if ((hd = calloc(1, sizeof(*hd))) == NULL)
return;
if ((hd->hd_dev = strdup(override)) == NULL)
goto err;
hd->hd_size = sb.st_size;
hd->hd_sectorsize = 512; /* XXX configurable? */
hd->hd_sectors = hd->hd_size / hd->hd_sectorsize;
if (hd->hd_size < HOSTDISK_MIN_SIZE)
goto err;
hd->hd_flags = 0;
STAILQ_INIT(&hd->hd_children);
printf("%s: %ju %ju %ju\n",
hd->hd_dev, hd->hd_size, hd->hd_sectors, hd->hd_sectorsize);
STAILQ_INSERT_TAIL(&hdinfo, hd, hd_link);
return;
err:
free(__DECONST(void *, hd->hd_dev));
free(hd);
}
static void
hostdisk_find_block_devices(void)
{
char *override;
override=getenv("hostdisk_override");
if (override != NULL)
hostdisk_fake_one_disk(override);
else
foreach_file(SYSBLK, hostdisk_one_disk, NULL, 0);
}
static int
hostdisk_init(void)
{
hostdisk_find_block_devices();
return (0);
}
static int
hostdisk_strategy(void *devdata, int flag, daddr_t dblk, size_t size,
char *buf, size_t *rsize)
{
struct devdesc *desc = devdata;
daddr_t pos;
int n;
int64_t off;
uint64_t res;
uint32_t posl, posh;
pos = dblk * 512;
posl = pos & 0xffffffffu;
posh = (pos >> 32) & 0xffffffffu;
if ((off = host_llseek(desc->d_unit, posh, posl, &res, 0)) < 0) {
printf("Seek error on fd %d to %ju (dblk %ju) returns %jd\n",
desc->d_unit, (uintmax_t)pos, (uintmax_t)dblk, (intmax_t)off);
return (EIO);
}
n = host_read(desc->d_unit, buf, size);
if (n < 0)
return (EIO);
*rsize = n;
return (0);
}
static int
hostdisk_open(struct open_file *f, ...)
{
struct devdesc *desc;
const char *fn;
va_list vl;
va_start(vl, f);
desc = va_arg(vl, struct devdesc *);
va_end(vl);
fn = dev2hd(desc)->hd_dev;
desc->d_unit = host_open(fn, O_RDONLY, 0);
if (desc->d_unit <= 0) {
printf("hostdisk_open: couldn't open %s: %d\n", fn, errno);
return (ENOENT);
}
return (0);
}
static int
hostdisk_close(struct open_file *f)
{
struct devdesc *desc = f->f_devdata;
host_close(desc->d_unit);
return (0);
}
static int
hostdisk_ioctl(struct open_file *f, u_long cmd, void *data)
{
struct devdesc *desc = f->f_devdata;
hdinfo_t *hd = dev2hd(desc);
switch (cmd) {
case DIOCGSECTORSIZE:
*(u_int *)data = hd->hd_sectorsize;
break;
case DIOCGMEDIASIZE:
*(uint64_t *)data = hd->hd_size;
break;
default:
return (ENOTTY);
}
return (0);
}
static int
hostdisk_print(int verbose)
{
char line[80];
hdinfo_t *hd, *md;
int ret = 0;
printf("%s devices:", hostdisk.dv_name);
if (pager_output("\n") != 0)
return (1);
STAILQ_FOREACH(hd, &hdinfo, hd_link) {
snprintf(line, sizeof(line),
" %s: %ju X %ju: %ju bytes\n",
hd->hd_dev,
(uintmax_t)hd->hd_sectors,
(uintmax_t)hd->hd_sectorsize,
(uintmax_t)hd->hd_size);
if ((ret = pager_output(line)) != 0)
break;
STAILQ_FOREACH(md, &hd->hd_children, hd_link) {
snprintf(line, sizeof(line),
" %s: %ju X %ju: %ju bytes\n",
md->hd_dev,
(uintmax_t)md->hd_sectors,
(uintmax_t)md->hd_sectorsize,
(uintmax_t)md->hd_size);
if ((ret = pager_output(line)) != 0)
goto done;
}
}
done:
return (ret);
}
static char *
hostdisk_fmtdev(struct devdesc *vdev)
{
static char name[DEV_DEVLEN];
snprintf(name, sizeof(name), "%s:", dev2hd(vdev)->hd_dev);
return (name);
}
static bool
hostdisk_match(struct devsw *devsw, const char *devspec)
{
hdinfo_t *hd;
const char *colon;
char *cp;
colon = strchr(devspec, ':');
if (colon == NULL)
return false;
cp = strdup(devspec);
cp[colon - devspec] = '\0';
hd = hostdisk_find(cp);
free(cp);
return (hd != NULL);
}
static int
hostdisk_parsedev(struct devdesc **idev, const char *devspec, const char **path)
{
const char *cp;
struct devdesc *dev;
hdinfo_t *hd;
int len;
char *fn;
/* Must have a : in it */
cp = strchr(devspec, ':');
if (cp == NULL)
return (EINVAL);
/* XXX Stat the /dev or defer error handling to open(2) call? */
if (path != NULL)
*path = cp + 1;
len = cp - devspec;
fn = strdup(devspec);
fn[len] = '\0';
hd = hostdisk_find(fn);
if (hd == NULL) {
printf("Can't find hdinfo for %s\n", fn);
free(fn);
return (EINVAL);
}
free(fn);
dev = malloc(sizeof(*dev));
if (dev == NULL)
return (ENOMEM);
dev->d_unit = 0;
dev->d_dev = &hostdisk;
dev->d_opendata = hd;
*idev = dev;
return (0);
}
/* XXX refactor */
static bool
sanity_check_currdev(void)
{
struct stat st;
return (stat(PATH_DEFAULTS_LOADER_CONF, &st) == 0 ||
#ifdef PATH_BOOTABLE_TOKEN
stat(PATH_BOOTABLE_TOKEN, &st) == 0 || /* non-standard layout */
#endif
stat(PATH_KERNEL, &st) == 0);
}
static const char *
hostdisk_try_one(hdinfo_t *hd)
{
char *fn;
if (asprintf(&fn, "%s:", hd->hd_dev) == -1)
return (NULL);
set_currdev(fn);
printf("Trying %s\n", fn);
if (sanity_check_currdev())
return (fn);
printf("Failed %s\n", fn);
free(fn);
return (NULL);
}
const char *
hostdisk_gen_probe(void)
{
hdinfo_t *hd, *md;
const char *rv = NULL;
STAILQ_FOREACH(hd, &hdinfo, hd_link) {
/* try whole disk */
if (hd->hd_flags & HDF_HAS_ZPOOL)
continue;
rv = hostdisk_try_one(hd);
if (rv != NULL)
return (rv);
/* try all partitions */
STAILQ_FOREACH(md, &hd->hd_children, hd_link) {
if (md->hd_flags & HDF_HAS_ZPOOL)
continue;
rv = hostdisk_try_one(md);
if (rv != NULL)
return (rv);
}
}
return (NULL);
}
#ifdef LOADER_ZFS_SUPPORT
static bool
hostdisk_zfs_check_one(hdinfo_t *hd)
{
char *fn;
bool found = false;
uint64_t pool_uuid;
if (asprintf(&fn, "%s:", hd->hd_dev) == -1)
return (false);
pool_uuid = 0;
zfs_probe_dev(fn, &pool_uuid, false);
if (pool_uuid != 0) {
found = true;
hd->hd_flags |= HDF_HAS_ZPOOL;
hd->hd_zfs_uuid = pool_uuid;
}
free(fn);
return (found);
}
void
hostdisk_zfs_probe(void)
{
hdinfo_t *hd, *md;
STAILQ_FOREACH(hd, &hdinfo, hd_link) {
if (hostdisk_zfs_check_one(hd))
continue;
STAILQ_FOREACH(md, &hd->hd_children, hd_link) {
hostdisk_zfs_check_one(md);
}
}
}
/* This likely shoud move to libsa/zfs/zfs.c and be used by at least EFI booting */
static bool
probe_zfs_currdev(uint64_t pool_guid, uint64_t root_guid, bool setcurrdev)
{
char *devname;
struct zfs_devdesc currdev;
bool bootable;
currdev.dd.d_dev = &zfs_dev;
currdev.dd.d_unit = 0;
currdev.pool_guid = pool_guid;
currdev.root_guid = root_guid;
devname = devformat(&currdev.dd);
if (setcurrdev)
set_currdev(devname);
bootable = sanity_check_currdev();
if (bootable) {
char buf[VDEV_PAD_SIZE];
if (zfs_get_bootonce(&currdev, OS_BOOTONCE, buf, sizeof(buf)) == 0) {
printf("zfs bootonce: %s\n", buf);
if (setcurrdev)
set_currdev(buf);
setenv("zfs-bootonce", buf, 1);
}
(void)zfs_attach_nvstore(&currdev);
init_zfs_boot_options(devname);
}
return (bootable);
}
static bool
hostdisk_zfs_try_default(hdinfo_t *hd)
{
return (probe_zfs_currdev(hd->hd_zfs_uuid, 0, true));
}
bool
hostdisk_zfs_find_default(void)
{
hdinfo_t *hd, *md;
STAILQ_FOREACH(hd, &hdinfo, hd_link) {
if (hd->hd_flags & HDF_HAS_ZPOOL) {
if (hostdisk_zfs_try_default(hd))
return (true);
continue;
}
STAILQ_FOREACH(md, &hd->hd_children, hd_link) {
if (md->hd_flags & HDF_HAS_ZPOOL) {
if (hostdisk_zfs_try_default(md))
return (true);
}
}
}
return (false);
}
#endif