Message-ID: <CAGXu5j+BPHPCnJ+ont8CGuUjjEYB_d6K-YtVJgNA7uwkpbfDoQ@mail.gmail.com>
Date:	Mon, 14 May 2012 15:21:17 -0700
From:	Kees Cook <keescook@...omium.org>
To:	Anton Vorontsov <anton.vorontsov@...aro.org>
Cc:	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Colin Cross <ccross@...roid.com>,
	Arnd Bergmann <arnd@...db.de>,
	John Stultz <john.stultz@...aro.org>, arve@...roid.com,
	Rebecca Schultz Zavin <rebecca@...roid.com>,
	Jesper Juhl <jj@...osbits.net>,
	Randy Dunlap <rdunlap@...otime.net>,
	Stephen Boyd <sboyd@...eaurora.org>,
	Thomas Meyer <thomas@...3r.de>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Marco Stornelli <marco.stornelli@...il.com>,
	WANG Cong <xiyou.wangcong@...il.com>,
	linux-kernel@...r.kernel.org, devel@...verdev.osuosl.org,
	linaro-kernel@...ts.linaro.org, patches@...aro.org,
	kernel-team@...roid.com
Subject: Re: [PATCH 10/11] pstore/ram: Switch to persistent_ram routines

On Fri, May 11, 2012 at 5:18 PM, Anton Vorontsov
<anton.vorontsov@...aro.org> wrote:
> The patch switches pstore RAM backend to use persistent_ram routines,
> one step closer to the ECC support.
>
> Signed-off-by: Anton Vorontsov <anton.vorontsov@...aro.org>

As mentioned, I'm all for this consolidation. That said, some notes below...

> ---
>  fs/pstore/ram.c |  109 ++++++++++++++++++++++++++++++-------------------------
>  1 file changed, 60 insertions(+), 49 deletions(-)
>
> diff --git a/fs/pstore/ram.c b/fs/pstore/ram.c
> index b26b58e..cf0ad92 100644
> --- a/fs/pstore/ram.c
> +++ b/fs/pstore/ram.c
> @@ -62,7 +62,7 @@ MODULE_PARM_DESC(dump_oops,
>                "set to 1 to dump oopses, 0 to only dump panics (default 1)");
>
>  struct ramoops_context {
> -       void *virt_addr;
> +       struct persistent_ram_zone **przs;
>        phys_addr_t phys_addr;
>        unsigned long size;
>        size_t record_size;
> @@ -90,39 +90,56 @@ static ssize_t ramoops_pstore_read(u64 *id, enum pstore_type_id *type,
>                                   struct pstore_info *psi)
>  {
>        ssize_t size;
> -       char *rambuf;
>        struct ramoops_context *cxt = psi->data;
> +       struct persistent_ram_zone *prz;
>
>        if (cxt->read_count >= cxt->max_count)
>                return -EINVAL;
> +
>        *id = cxt->read_count++;
> +       prz = cxt->przs[*id];
> +
>        /* Only supports dmesg output so far. */
>        *type = PSTORE_TYPE_DMESG;
>        /* TODO(kees): Bogus time for the moment. */
>        time->tv_sec = 0;
>        time->tv_nsec = 0;
>
> -       rambuf = cxt->virt_addr + (*id * cxt->record_size);
> -       size = strnlen(rambuf, cxt->record_size);
> +       size = persistent_ram_old_size(prz);
>        *buf = kmalloc(size, GFP_KERNEL);
>        if (*buf == NULL)
>                return -ENOMEM;
> -       memcpy(*buf, rambuf, size);
> +       memcpy(*buf, persistent_ram_old(prz), size);
>
>        return size;
>  }
>
> +static size_t ramoops_write_kmsg_hdr(struct persistent_ram_zone *prz)
> +{
> +       char *hdr;
> +       struct timeval timestamp;
> +       size_t len;
> +
> +       do_gettimeofday(&timestamp);
> +       hdr = kasprintf(GFP_ATOMIC, RAMOOPS_KERNMSG_HDR "%lu.%lu\n",
> +               (long)timestamp.tv_sec, (long)timestamp.tv_usec);
> +       WARN_ON_ONCE(!hdr);
> +       len = hdr ? strlen(hdr) : 0;
> +       persistent_ram_write(prz, hdr, len);
> +       kfree(hdr);
> +
> +       return len;
> +}
> +
>  static int ramoops_pstore_write(enum pstore_type_id type,
>                                enum kmsg_dump_reason reason,
>                                u64 *id,
>                                unsigned int part,
>                                size_t size, struct pstore_info *psi)
>  {
> -       char *buf;
> -       size_t res;
> -       struct timeval timestamp;
>        struct ramoops_context *cxt = psi->data;
> -       size_t available = cxt->record_size;
> +       struct persistent_ram_zone *prz = cxt->przs[cxt->count];
> +       size_t hlen;
>
>        /* Currently ramoops is designed to only store dmesg dumps. */
>        if (type != PSTORE_TYPE_DMESG)
> @@ -147,22 +164,10 @@ static int ramoops_pstore_write(enum pstore_type_id type,
>        if (part != 1)
>                return -ENOSPC;
>
> -       buf = cxt->virt_addr + (cxt->count * cxt->record_size);
> -
> -       res = sprintf(buf, "%s", RAMOOPS_KERNMSG_HDR);
> -       buf += res;
> -       available -= res;
> -
> -       do_gettimeofday(&timestamp);
> -       res = sprintf(buf, "%lu.%lu\n", (long)timestamp.tv_sec, (long)timestamp.tv_usec);
> -       buf += res;
> -       available -= res;
> -
> -       if (size > available)
> -               size = available;
> -
> -       memcpy(buf, cxt->pstore.buf, size);
> -       memset(buf + size, '\0', available - size);
> +       hlen = ramoops_write_kmsg_hdr(prz);
> +       if (size + hlen > prz->buffer_size)
> +               size = prz->buffer_size - hlen;
> +       persistent_ram_write(prz, cxt->pstore.buf, size);
>
>        cxt->count = (cxt->count + 1) % cxt->max_count;
>
> @@ -172,14 +177,12 @@ static int ramoops_pstore_write(enum pstore_type_id type,
>  static int ramoops_pstore_erase(enum pstore_type_id type, u64 id,
>                                struct pstore_info *psi)
>  {
> -       char *buf;
>        struct ramoops_context *cxt = psi->data;
>
>        if (id >= cxt->max_count)
>                return -EINVAL;
>
> -       buf = cxt->virt_addr + (id * cxt->record_size);
> -       memset(buf, '\0', cxt->record_size);
> +       persistent_ram_free_old(cxt->przs[id]);

Hm, I don't think persistent_ram_free_old() is what's wanted here.
That appears to entirely release the region? I want to make sure the
memory is cleared first. And will this area come back on a write, or
does it stay released?

>
>        return 0;
>  }
> @@ -200,6 +203,7 @@ static int __init ramoops_probe(struct platform_device *pdev)
>        struct ramoops_platform_data *pdata = pdev->dev.platform_data;
>        struct ramoops_context *cxt = &oops_cxt;
>        int err = -EINVAL;
> +       int i;
>
>        /* Only a single ramoops area allowed at a time, so fail extra
>         * probes.
> @@ -237,32 +241,37 @@ static int __init ramoops_probe(struct platform_device *pdev)
>        cxt->record_size = pdata->record_size;
>        cxt->dump_oops = pdata->dump_oops;
>
> +       cxt->przs = kzalloc(sizeof(*cxt->przs) * cxt->max_count, GFP_KERNEL);
> +       if (!cxt->przs) {
> +               pr_err("failed to initialize a prz array\n");
> +               goto fail_przs;

This should be fail_out.

> +       }
> +
> +       for (i = 0; i < cxt->max_count; i++) {
> +               size_t sz = cxt->record_size;
> +               phys_addr_t start = cxt->phys_addr + sz * i;
> +
> +               cxt->przs[i] = persistent_ram_new(start, sz, 0);

persistent_ram_new() is marked as __init, so this is unsafe to call if
built as a module. I think persistent_ram_new() will need to lose the
__init marking, or I'm misunderstanding something.

> +               if (IS_ERR(cxt->przs[i])) {
> +                       err = PTR_ERR(cxt->przs[i]);
> +                       pr_err("failed to initialize a prz\n");

Since neither persistent_ram_new() nor persistent_ram_buffer_map()
report the location of the failure, I'd like to keep the error report
(removed below "pr_err("request mem region (0x%lx@...llx)
failed\n",...") for failures, so there is something actionable in
dmesg when the platform data is mismatched for the hardware.

> +                       goto fail_prz;

This should be fail_przs.

> +               }
> +       }
> +
>        cxt->pstore.data = cxt;
> -       cxt->pstore.bufsize = cxt->record_size;
> -       cxt->pstore.buf = kmalloc(cxt->pstore.bufsize, GFP_KERNEL);
>        spin_lock_init(&cxt->pstore.buf_lock);
> +       cxt->pstore.bufsize = cxt->przs[0]->buffer_size;
> +       cxt->pstore.buf = kmalloc(cxt->pstore.bufsize, GFP_KERNEL);

I don't see a reason to re-order these (nothing can use buf yet
because we haven't registered it with pstore yet).

>        if (!cxt->pstore.buf) {
>                pr_err("cannot allocate pstore buffer\n");
>                goto fail_clear;
>        }
>
> -       if (!request_mem_region(cxt->phys_addr, cxt->size, "ramoops")) {
> -               pr_err("request mem region (0x%lx@...llx) failed\n",
> -                       cxt->size, (unsigned long long)cxt->phys_addr);
> -               err = -EINVAL;
> -               goto fail_buf;
> -       }
> -
> -       cxt->virt_addr = ioremap(cxt->phys_addr,  cxt->size);
> -       if (!cxt->virt_addr) {
> -               pr_err("ioremap failed\n");
> -               goto fail_mem_region;
> -       }
> -
>        err = pstore_register(&cxt->pstore);
>        if (err) {
>                pr_err("registering with pstore failed\n");
> -               goto fail_iounmap;
> +               goto fail_pstore;

This should be fail_buf.

>        }
>
>        /*
> @@ -280,15 +289,17 @@ static int __init ramoops_probe(struct platform_device *pdev)
>
>        return 0;
>
> -fail_iounmap:
> -       iounmap(cxt->virt_addr);
> -fail_mem_region:
> -       release_mem_region(cxt->phys_addr, cxt->size);
> -fail_buf:
> +fail_pstore:

No reason to rename this from "fail_buf".

>        kfree(cxt->pstore.buf);
>  fail_clear:
>        cxt->pstore.bufsize = 0;
>        cxt->max_count = 0;
> +fail_przs:
> +       for (i = 0; cxt->przs[i]; i++)
> +               persistent_ram_free(cxt->przs[i]);

This can lead to a BUG, since persistent_ram_free() doesn't handle
NULL arguments.

> +       kfree(cxt->przs);
> +fail_prz:
> +       kfree(cxt->pstore.buf);

This target (fail_prz) should be removed, and the kfree is redundant
to fail_buf above.

>  fail_out:
>        return err;
>  }
> --
> 1.7.9.2
>

-Kees

-- 
Kees Cook
Chrome OS Security
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists