Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
In-Reply-To: <20180214214105-mutt-send-email-mst@kernel.org>
References: <20180214141850.4017-1-marcandre.lureau@redhat.com>
 <20180214141850.4017-9-marcandre.lureau@redhat.com> <20180214214105-mutt-send-email-mst@kernel.org>
From:   =?UTF-8?B?TWFyYy1BbmRyw6kgTHVyZWF1?= <marcandre.lureau@redhat.com>
Date:   Thu, 15 Feb 2018 19:24:34 +0100
Message-ID: <CAMxuvawaNA_yWE86Rv5abgLrdduYeJrc5Ezf63-F0HRhRjO=3Q@mail.gmail.com>
Subject: Re: [PATCH v14 8/9] fw_cfg: write vmcoreinfo details
To:     "Michael S. Tsirkin" <mst@redhat.com>
Cc:     Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Baoquan He <bhe@redhat.com>,
        Sergio Lopez Pascual <slp@redhat.com>,
        "Somlo, Gabriel" <somlo@cmu.edu>, xiaolong.ye@intel.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 8BIT
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi

On Thu, Feb 15, 2018 at 7:09 PM, Michael S. Tsirkin <mst@redhat.com> wrote:
> On Wed, Feb 14, 2018 at 03:18:49PM +0100, Marc-André Lureau wrote:
>> If the "etc/vmcoreinfo" fw_cfg file is present and we are not running
>> the kdump kernel, write the addr/size of the vmcoreinfo ELF note.
>>
>> The DMA operation is expected to run synchronously with today qemu,
>> but the specification states that it may become async, so we run
>> "control" field check in a loop for eventual changes.
>>
>> Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
>> ---
>>  drivers/firmware/qemu_fw_cfg.c | 144 ++++++++++++++++++++++++++++++++++++++++-
>>  1 file changed, 141 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/firmware/qemu_fw_cfg.c b/drivers/firmware/qemu_fw_cfg.c
>> index 37638b95cb45..69939e2529f2 100644
>> --- a/drivers/firmware/qemu_fw_cfg.c
>> +++ b/drivers/firmware/qemu_fw_cfg.c
>> @@ -34,11 +34,17 @@
>>  #include <linux/io.h>
>>  #include <linux/ioport.h>
>>  #include <linux/fw_cfg.h>
>> +#include <linux/delay.h>
>> +#include <linux/crash_dump.h>
>> +#include <linux/crash_core.h>
>>
>>  MODULE_AUTHOR("Gabriel L. Somlo <somlo@cmu.edu>");
>>  MODULE_DESCRIPTION("QEMU fw_cfg sysfs support");
>>  MODULE_LICENSE("GPL");
>>
>> +/* fw_cfg revision attribute, in /sys/firmware/qemu_fw_cfg top-level dir. */
>> +static u32 fw_cfg_rev;
>> +
>>  /* fw_cfg device i/o register addresses */
>>  static bool fw_cfg_is_mmio;
>>  static phys_addr_t fw_cfg_p_base;
>> @@ -59,6 +65,65 @@ static inline u16 fw_cfg_sel_endianness(u16 key)
>>               (u16 __force)cpu_to_le16(key);
>>  }
>>
>> +static inline bool fw_cfg_dma_enabled(void)
>> +{
>> +     return (fw_cfg_rev & FW_CFG_VERSION_DMA) && fw_cfg_reg_dma;
>> +}
>> +
>> +/* qemu fw_cfg device is sync today, but spec says it may become async */
>> +static void fw_cfg_wait_for_control(struct fw_cfg_dma_access *d)
>> +{
>> +     do {
>> +             u32 ctrl = be32_to_cpu(READ_ONCE(d->control));
>> +
>> +             if ((ctrl & ~FW_CFG_DMA_CTL_ERROR) == 0)
>> +                     return;
>> +
>> +             usleep_range(50, 100);
>
> I would just do cpu_relax() here.

ok, I didn't know that one.

>
>> +     } while (true);
>> +
>> +     /* do not reorder the read to d->control */
>> +     rmb();
>
> Hmm. I don't really understand the comment.
> Is this code ever reacheable? How does it help?

I thought that's what you suggested in v13 review, but true, I should
replace the return with a break to reach it. Is that what you expect
too? (my understanding is to make sure the READ_ONCE(control) in
wait_for_control happens before READ_ONCE(control) after in
dma_transfer)

>
>> +}
>> +
>> +static ssize_t fw_cfg_dma_transfer(void *address, u32 length, u32 control)
>> +{
>> +     phys_addr_t dma;
>> +     struct fw_cfg_dma_access *d = NULL;
>> +     ssize_t ret = length;
>> +
>> +     d = kmalloc(sizeof(*d), GFP_KERNEL);
>> +     if (!d) {
>> +             ret = -ENOMEM;
>> +             goto end;
>> +     }
>> +
>> +     /* fw_cfg device does not need IOMMU protection, so use physical addresses */
>> +     *d = (struct fw_cfg_dma_access) {
>> +             .address = cpu_to_be64(address ? virt_to_phys(address) : 0),
>> +             .length = cpu_to_be32(length),
>> +             .control = cpu_to_be32(control)
>> +     };
>> +
>> +     dma = virt_to_phys(d);
>> +
>> +     iowrite32be((u64)dma >> 32, fw_cfg_reg_dma);
>> +     /* force memory to sync before notifying device via MMIO */
>> +     wmb();
>> +     iowrite32be(dma, fw_cfg_reg_dma + 4);
>> +
>> +     fw_cfg_wait_for_control(d);
>> +
>> +     if (be32_to_cpu(READ_ONCE(d->control)) & FW_CFG_DMA_CTL_ERROR) {
>> +             ret = -EIO;
>> +     }
>> +
>> +end:
>> +     kfree(d);
>> +
>> +     return ret;
>> +}
>> +
>>  /* read chunk of given fw_cfg blob (caller responsible for sanity-check) */
>>  static inline void fw_cfg_read_blob(u16 key,
>>                                   void *buf, loff_t pos, size_t count)
>> @@ -87,6 +152,47 @@ static inline void fw_cfg_read_blob(u16 key,
>>       acpi_release_global_lock(glk);
>>  }
>>
>> +#ifdef CONFIG_CRASH_CORE
>> +/* write chunk of given fw_cfg blob (caller responsible for sanity-check) */
>> +static ssize_t fw_cfg_write_blob(u16 key,
>> +                              void *buf, loff_t pos, size_t count)
>> +{
>> +     u32 glk = -1U;
>> +     acpi_status status;
>> +     ssize_t ret = count;
>> +
>> +     /* If we have ACPI, ensure mutual exclusion against any potential
>> +      * device access by the firmware, e.g. via AML methods:
>> +      */
>> +     status = acpi_acquire_global_lock(ACPI_WAIT_FOREVER, &glk);
>> +     if (ACPI_FAILURE(status) && status != AE_NOT_CONFIGURED) {
>> +             /* Should never get here */
>> +             WARN(1, "%s: Failed to lock ACPI!\n", __func__);
>> +             return -EINVAL;
>> +     }
>> +
>> +     mutex_lock(&fw_cfg_dev_lock);
>> +     if (pos == 0) {
>> +             ret = fw_cfg_dma_transfer(buf, count, key << 16
>> +                                       | FW_CFG_DMA_CTL_SELECT
>> +                                       | FW_CFG_DMA_CTL_WRITE);
>> +     } else {
>> +             iowrite16(fw_cfg_sel_endianness(key), fw_cfg_reg_ctrl);
>> +             ret = fw_cfg_dma_transfer(NULL, pos, FW_CFG_DMA_CTL_SKIP);
>> +             if (ret < 0)
>> +                     goto end;
>> +             ret = fw_cfg_dma_transfer(buf, count, FW_CFG_DMA_CTL_WRITE);
>> +     }
>> +
>> +end:
>> +     mutex_unlock(&fw_cfg_dev_lock);
>> +
>> +     acpi_release_global_lock(glk);
>> +
>> +     return ret;
>> +}
>> +#endif /* CONFIG_CRASH_CORE */
>> +
>>  /* clean up fw_cfg device i/o */
>>  static void fw_cfg_io_cleanup(void)
>>  {
>> @@ -185,9 +291,6 @@ static int fw_cfg_do_platform_probe(struct platform_device *pdev)
>>       return 0;
>>  }
>>
>> -/* fw_cfg revision attribute, in /sys/firmware/qemu_fw_cfg top-level dir. */
>> -static u32 fw_cfg_rev;
>> -
>>  static ssize_t fw_cfg_showrev(struct kobject *k, struct attribute *a, char *buf)
>>  {
>>       return sprintf(buf, "%u\n", fw_cfg_rev);
>> @@ -210,6 +313,32 @@ struct fw_cfg_sysfs_entry {
>>       struct list_head list;
>>  };
>>
>> +#ifdef CONFIG_CRASH_CORE
>> +static ssize_t fw_cfg_write_vmcoreinfo(const struct fw_cfg_file *f)
>> +{
>> +     static struct fw_cfg_vmcoreinfo *data;
>> +     ssize_t ret;
>> +
>> +     data = kmalloc(sizeof(struct fw_cfg_vmcoreinfo), GFP_KERNEL);
>> +     if (!data)
>> +             return -ENOMEM;
>> +
>> +     *data = (struct fw_cfg_vmcoreinfo) {
>> +             .guest_format = cpu_to_le16(FW_CFG_VMCOREINFO_FORMAT_ELF),
>> +             .size = cpu_to_le32(VMCOREINFO_NOTE_SIZE),
>> +             .paddr = cpu_to_le64(paddr_vmcoreinfo_note())
>> +     };
>> +     /* spare ourself reading host format support for now since we
>> +      * don't know what else to format - host may ignore ours
>> +      */
>> +     ret = fw_cfg_write_blob(be16_to_cpu(f->select), data,
>> +                             0, sizeof(struct fw_cfg_vmcoreinfo));
>> +
>> +     kfree(data);
>> +     return ret;
>> +}
>> +#endif /* CONFIG_CRASH_CORE */
>> +
>>  /* get fw_cfg_sysfs_entry from kobject member */
>>  static inline struct fw_cfg_sysfs_entry *to_entry(struct kobject *kobj)
>>  {
>> @@ -450,6 +579,15 @@ static int fw_cfg_register_file(const struct fw_cfg_file *f)
>>       int err;
>>       struct fw_cfg_sysfs_entry *entry;
>>
>> +#ifdef CONFIG_CRASH_CORE
>> +     if (fw_cfg_dma_enabled() &&
>> +             strcmp(f->name, FW_CFG_VMCOREINFO_FILENAME) == 0 &&
>> +             !is_kdump_kernel()) {
>> +             if (fw_cfg_write_vmcoreinfo(f) < 0)
>> +                     pr_warn("fw_cfg: failed to write vmcoreinfo");
>> +     }
>> +#endif
>> +
>>       /* allocate new entry */
>>       entry = kzalloc(sizeof(*entry), GFP_KERNEL);
>>       if (!entry)
>> --
>> 2.16.1.73.g5832b7e9f2