The LBA28 capacity (at offsets 60/61 of identification) is supposed to
express the maximum size supported by LBA28 commands. If the device is
larger than this, we have to cap it to 2^28-1.

At least NetBSD happens to be using this value to determine whether to use
LBA28 or LBA48 for its commands, using LBA28 for sectors that don't need
LBA48. This commit thus fixes NetBSD access to disks larger than 128GiB.

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
---
 hw/ide/core.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/hw/ide/core.c b/hw/ide/core.c
index fd69ca3167..e28f8aad61 100644
--- a/hw/ide/core.c
+++ b/hw/ide/core.c
@@ -98,8 +98,12 @@ static void put_le16(uint16_t *p, unsigned int v)
 static void ide_identify_size(IDEState *s)
 {
     uint16_t *p = (uint16_t *)s->identify_data;
-    put_le16(p + 60, s->nb_sectors);
-    put_le16(p + 61, s->nb_sectors >> 16);
+    int64_t nb_sectors_lba28 = s->nb_sectors;
+    if (nb_sectors_lba28 >= 1 << 28) {
+        nb_sectors_lba28 = (1 << 28) - 1;
+    }
+    put_le16(p + 60, nb_sectors_lba28);
+    put_le16(p + 61, nb_sectors_lba28 >> 16);
     put_le16(p + 100, s->nb_sectors);
     put_le16(p + 101, s->nb_sectors >> 16);
     put_le16(p + 102, s->nb_sectors >> 32);
-- 
2.32.0

Ping?

Samuel Thibault, le mar. 24 août 2021 12:43:44 +0200, a ecrit:
> The LBA28 capacity (at offsets 60/61 of identification) is supposed to
> express the maximum size supported by LBA28 commands. If the device is
> larger than this, we have to cap it to 2^28-1.
> 
> At least NetBSD happens to be using this value to determine whether to use
> LBA28 or LBA48 for its commands, using LBA28 for sectors that don't need
> LBA48. This commit thus fixes NetBSD access to disks larger than 128GiB.
> 
> Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
> ---
>  hw/ide/core.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/hw/ide/core.c b/hw/ide/core.c
> index fd69ca3167..e28f8aad61 100644
> --- a/hw/ide/core.c
> +++ b/hw/ide/core.c
> @@ -98,8 +98,12 @@ static void put_le16(uint16_t *p, unsigned int v)
>  static void ide_identify_size(IDEState *s)
>  {
>      uint16_t *p = (uint16_t *)s->identify_data;
> -    put_le16(p + 60, s->nb_sectors);
> -    put_le16(p + 61, s->nb_sectors >> 16);
> +    int64_t nb_sectors_lba28 = s->nb_sectors;
> +    if (nb_sectors_lba28 >= 1 << 28) {
> +        nb_sectors_lba28 = (1 << 28) - 1;
> +    }
> +    put_le16(p + 60, nb_sectors_lba28);
> +    put_le16(p + 61, nb_sectors_lba28 >> 16);
>      put_le16(p + 100, s->nb_sectors);
>      put_le16(p + 101, s->nb_sectors >> 16);
>      put_le16(p + 102, s->nb_sectors >> 32);
> -- 
> 2.32.0
> 

Ping?

Samuel Thibault, le mar. 24 août 2021 12:43:44 +0200, a ecrit:
> The LBA28 capacity (at offsets 60/61 of identification) is supposed to
> express the maximum size supported by LBA28 commands. If the device is
> larger than this, we have to cap it to 2^28-1.
> 
> At least NetBSD happens to be using this value to determine whether to use
> LBA28 or LBA48 for its commands, using LBA28 for sectors that don't need
> LBA48. This commit thus fixes NetBSD access to disks larger than 128GiB.
> 
> Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
> ---
>  hw/ide/core.c | 8 ++++++--
>  1 file changed, 6 insertions(+), 2 deletions(-)
> 
> diff --git a/hw/ide/core.c b/hw/ide/core.c
> index fd69ca3167..e28f8aad61 100644
> --- a/hw/ide/core.c
> +++ b/hw/ide/core.c
> @@ -98,8 +98,12 @@ static void put_le16(uint16_t *p, unsigned int v)
>  static void ide_identify_size(IDEState *s)
>  {
>      uint16_t *p = (uint16_t *)s->identify_data;
> -    put_le16(p + 60, s->nb_sectors);
> -    put_le16(p + 61, s->nb_sectors >> 16);
> +    int64_t nb_sectors_lba28 = s->nb_sectors;
> +    if (nb_sectors_lba28 >= 1 << 28) {
> +        nb_sectors_lba28 = (1 << 28) - 1;
> +    }
> +    put_le16(p + 60, nb_sectors_lba28);
> +    put_le16(p + 61, nb_sectors_lba28 >> 16);
>      put_le16(p + 100, s->nb_sectors);
>      put_le16(p + 101, s->nb_sectors >> 16);
>      put_le16(p + 102, s->nb_sectors >> 32);
> -- 
> 2.32.0
>