qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Daniel P. Berrangé" <berrange@redhat.com>
To: Zhuoying Cai <zycai@linux.ibm.com>
Cc: thuth@redhat.com, richard.henderson@linaro.org, david@redhat.com,
	pbonzini@redhat.com, walling@linux.ibm.com,
	jjherne@linux.ibm.com, jrossi@linux.ibm.com, pasic@linux.ibm.com,
	borntraeger@linux.ibm.com, farman@linux.ibm.com,
	iii@linux.ibm.com, eblake@redhat.com, armbru@redhat.com,
	qemu-s390x@nongnu.org, qemu-devel@nongnu.org
Subject: Re: [PATCH v3 03/28] hw/s390x/ipl: Create certificate store
Date: Fri, 6 Jun 2025 11:31:27 +0100	[thread overview]
Message-ID: <aELDf8--oCzcfhFI@redhat.com> (raw)
In-Reply-To: <20250604215657.528142-4-zycai@linux.ibm.com>

On Wed, Jun 04, 2025 at 05:56:31PM -0400, Zhuoying Cai wrote:
> Create a certificate store for boot certificates used for secure IPL.
> 
> Load certificates from the boot-certificate parameter of s390-ccw-virtio
> machine type option into the cert store.
> 
> Currently, only x509 certificates in DER format and uses SHA-256 hashing
> algorithm are supported, as these are the types required for secure boot
> on s390.
> 
> Signed-off-by: Zhuoying Cai <zycai@linux.ibm.com>
> ---
>  hw/s390x/cert-store.c       | 247 ++++++++++++++++++++++++++++++++++++
>  hw/s390x/cert-store.h       |  39 ++++++
>  hw/s390x/ipl.c              |   9 ++
>  hw/s390x/ipl.h              |   3 +
>  hw/s390x/meson.build        |   1 +
>  include/hw/s390x/ipl/qipl.h |   3 +
>  6 files changed, 302 insertions(+)
>  create mode 100644 hw/s390x/cert-store.c
>  create mode 100644 hw/s390x/cert-store.h
> 
> diff --git a/hw/s390x/cert-store.c b/hw/s390x/cert-store.c
> new file mode 100644
> index 0000000000..562fa22241
> --- /dev/null
> +++ b/hw/s390x/cert-store.c
> @@ -0,0 +1,247 @@
> +/*
> + * S390 certificate store implementation
> + *
> + * Copyright 2025 IBM Corp.
> + * Author(s): Zhuoying Cai <zycai@linux.ibm.com>
> + *
> + * SPDX-License-Identifier: GPL-2.0-or-later
> + */
> +
> +#include "qemu/osdep.h"
> +#include "cert-store.h"
> +#include "qemu/error-report.h"
> +#include "qemu/option.h"
> +#include "qemu/config-file.h"
> +#include "hw/s390x/ebcdic.h"
> +#include "hw/s390x/s390-virtio-ccw.h"
> +#include "qemu/cutils.h"
> +#include "crypto/x509-utils.h"
> +
> +static const char *s390_get_boot_certificates(void)
> +{
> +    return S390_CCW_MACHINE(qdev_get_machine())->boot_certificates;
> +}
> +
> +static size_t cert2buf(char *path, size_t max_size, char **cert_buf)
> +{
> +    size_t size;
> +
> +    /*
> +     * maximum allowed size of the certificate file to avoid consuming excessive memory
> +     * (malformed or maliciously large files)
> +     */
> +    if (!g_file_get_contents(path, cert_buf, &size, NULL) ||
> +        size == 0 || size > max_size) {

By the time this 'size > max_size' check is performed, the file
is already loaded into memory, which is fairly pointless. In
existing code loading certs we don't enforce any max size. This
data comes from the host admin who QEMU has to assume is trusted,
so the size check is not required.

> +        return 0;
> +    }
> +
> +    return size;
> +}
> +
> +static S390IPLCertificate *init_cert_x509_der(size_t size, char *raw)
> +{
> +    S390IPLCertificate *q_cert = NULL;
> +    int key_id_size;
> +    int hash_size;
> +    int is_der;
> +    int hash_type;
> +    Error *err = NULL;
> +
> +    is_der = qcrypto_check_x509_cert_fmt((uint8_t *)raw, size,
> +                                         QCRYPTO_CERT_FMT_DER, &err);
> +    /* return early if GNUTLS is not enabled */
> +    if (is_der == -ENOTSUP) {
> +        error_report("GNUTLS is not enabled");
> +        return NULL;
> +    }
> +    if (is_der != 0) {
> +        error_report("The certificate is not in DER format");
> +        return NULL;
> +    }

As mentioned in the other patch, we should exclusively accept PEM
format as the public interface. If we need DER format internllay,
gnutls_x509_cert_export can convert it into DER for us.

> +
> +    hash_type = qcrypto_get_x509_signature_algorithm((uint8_t *)raw, size, &err);
> +    if (hash_type != QCRYPTO_SIG_ALGO_RSA_SHA256) {
> +        error_report("The certificate does not use SHA-256 hashing");
> +        return NULL;
> +    }
> +
> +    key_id_size = qcrypto_get_x509_keyid_len(QCRYPTO_KEYID_FLAGS_SHA256);
> +    if (key_id_size == 0) {
> +        error_report("Failed to get certificate key ID size");
> +        return NULL;
> +    }
> +
> +    hash_size = qcrypto_get_x509_hash_len(QCRYPTO_HASH_ALGO_SHA256);
> +    if (hash_size == 0) {
> +        error_report("Failed to get certificate hash size");
> +        return NULL;
> +    }

Pointless method call when we have a QCRYPTO_HASH_DIGEST_LEN_SHA256
constant.

> +
> +    q_cert = g_new(S390IPLCertificate, 1);

g_new0 to guarantee zero initialization of all fields.

> +    q_cert->size = size;
> +    q_cert->key_id_size = key_id_size;
> +    q_cert->hash_size = hash_size;
> +    q_cert->raw = raw;
> +    q_cert->format = QCRYPTO_CERT_FMT_DER;
> +    q_cert->hash_type = QCRYPTO_SIG_ALGO_RSA_SHA256;
> +
> +    return q_cert;
> +}
> +
> +static int check_path_type(const char *path)
> +{
> +    struct stat path_stat;
> +
> +    if (stat(path, &path_stat) != 0) {
> +        perror("stat");
> +        return -1;
> +    }
> +
> +    if (S_ISDIR(path_stat.st_mode)) {
> +        return S_IFDIR;
> +    } else if (S_ISREG(path_stat.st_mode)) {
> +        return S_IFREG;
> +    } else {
> +        return -1;
> +    }
> +}

This helper isn't making the code any simpler - use the
stat() & S_ISXXX checks inline where needed.

> +static S390IPLCertificate *init_cert(char *paths)

s/paths/path/ as this is a single file being used.

> +{
> +    char *buf;
> +    char vc_name[VC_NAME_LEN_BYTES];
> +    g_autofree gchar *filename;

All g_autofree variables *must* be initialized at time
of declaration. 

> +    size_t size;
> +    S390IPLCertificate *qcert = NULL;
> +
> +    filename = g_path_get_basename(paths);
> +
> +    size = cert2buf(paths, CERT_MAX_SIZE, &buf);
> +    if (size == 0) {
> +        error_report("Failed to load certificate: %s", paths);
> +        g_free(buf);
> +        return NULL;
> +    }
> +
> +    qcert = init_cert_x509_der(size, buf);
> +    if (qcert == NULL) {
> +        error_report("Failed to initialize certificate: %s", paths);
> +        g_free(buf);
> +        return NULL;
> +    }
> +
> +    /*
> +     * Left justified certificate name with padding on the right with blanks.
> +     * Convert certificate name to EBCDIC.
> +     */
> +    strpadcpy(vc_name, VC_NAME_LEN_BYTES, filename, ' ');

What purpose does the 'vc_name' serve ? Are there expectations
on the user for naming of the cert fiels ?

> +    ebcdic_put(qcert->vc_name, vc_name, VC_NAME_LEN_BYTES);
> +
> +    return qcert;
> +}
> +
> +static void update_cert_store(S390IPLCertificateStore *cert_store,
> +                              S390IPLCertificate *qcert)
> +{
> +    size_t data_buf_size;
> +    size_t keyid_buf_size;
> +    size_t hash_buf_size;
> +    size_t cert_buf_size;
> +
> +    /* length field is word aligned for later DIAG use */
> +    keyid_buf_size = ROUND_UP(qcert->key_id_size, 4);
> +    hash_buf_size = ROUND_UP(qcert->hash_size, 4);
> +    cert_buf_size = ROUND_UP(qcert->size, 4);
> +    data_buf_size = keyid_buf_size + hash_buf_size + cert_buf_size;
> +
> +    if (cert_store->max_cert_size < data_buf_size) {
> +        cert_store->max_cert_size = data_buf_size;
> +    }
> +
> +    cert_store->certs[cert_store->count] = *qcert;
> +    cert_store->total_bytes += data_buf_size;
> +    cert_store->count++;
> +}
> +
> +static GPtrArray *get_cert_paths(void)
> +{
> +    const char *path;
> +    gchar **paths;

g_auto(GStrv) paths = NULL;

to automatically free this

> +    gchar **paths_copy;

and use this as the working variable

> +    int path_type;
> +    GDir *dir = NULL;
> +    gchar *cert_path;
> +    const gchar *filename;
> +    GPtrArray *cert_path_builder;
> +
> +    cert_path_builder = g_ptr_array_new();

g_autoptr(GPtrArray) cert_path_builder = g_ptr_array_new_fill(0, g_free)

so that this get freed automatically in error paths, along
with any paths stored within it.

> +
> +    path = s390_get_boot_certificates();
> +    if (path == NULL) {
> +        return cert_path_builder;
> +    }

Calling this variable 'path' is misleading given it is
a list of paths.

> +
> +    paths = g_strsplit(path, ":", -1);
> +    /* save the original pointer for freeing later */
> +    paths_copy = paths;
> +    while (*paths) {
> +        /* skip empty certificate path */
> +        if (!strcmp(*paths, "")) {

IMHO this should be reported as a fatal user configuration
error.

> +            paths += 1;
> +            continue;
> +        }
> +
> +        cert_path = NULL;
> +        path_type = check_path_type(*paths);

Just call stat() directly here, and report a fatal error
if returns non-zero instead of ignoring the error

> +        if (path_type == S_IFREG) {
> +            cert_path = g_strdup(*paths);
> +            g_ptr_array_add(cert_path_builder, cert_path);

Drop the intermediate cert_path variable as it
serves no purpose

> +        } else if (path_type == S_IFDIR) {
> +            dir = g_dir_open(*paths, 0, NULL);

Pass an "GError" object not NULL and treat failure as
fatal.

> +
> +            if (dir) {
> +                while ((filename = g_dir_read_name(dir))) {
> +                    cert_path = g_build_filename(*paths, filename, NULL);
> +                    g_ptr_array_add(cert_path_builder, (gpointer) cert_path);
> +                }

The (gpointer) cast serves no purpose.

> +
> +                g_dir_close(dir);
> +            }
> +        }


> +
> +        paths += 1;
> +    }
> +
> +    g_strfreev(paths_copy);

Redundant when using g_auto

> +    return cert_path_builder;

return g_steal_pointer(&cert_path_builder)


> +}
> +
> +void s390_ipl_create_cert_store(S390IPLCertificateStore *cert_store)
> +{
> +    GPtrArray *cert_path_builder;
> +
> +    cert_path_builder = get_cert_paths();
> +    if (cert_path_builder->len == 0) {
> +        g_ptr_array_free(cert_path_builder, true);

s/true/TRUE/ - GLib does not accept stdbool constants

> +        return;
> +    }
> +
> +    cert_store->max_cert_size = 0;
> +    cert_store->total_bytes = 0;
> +
> +    for (int i = 0; i < cert_path_builder->len; i++) {
> +        if (i > MAX_CERTIFICATES - 1) {
> +            printf("Warning: Maximum %d certificates are allowed,"
> +                            " ignoring certificate #%d and beyond\n",
> +                            MAX_CERTIFICATES, i + 1);

Is this limit mandated by some s390x hardware design spec ?

If so, this should be a fatal error, not a warning.

If not, then it would be better to accept all the user
provided certs, no matter how many.

> +            break;
> +        }



> +
> +        S390IPLCertificate *qcert = init_cert((char *) cert_path_builder->pdata[i]);
> +        if (qcert) {
> +            update_cert_store(cert_store, qcert);
> +        }
> +    }
> +
> +    g_ptr_array_free(cert_path_builder, true);
> +}

With regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|



  reply	other threads:[~2025-06-06 10:32 UTC|newest]

Thread overview: 51+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-06-04 21:56 [PATCH v3 00/28] Secure IPL Support for SCSI Scheme of virtio-blk/virtio-scsi Devices Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 01/28] Add boot-certificates to s390-ccw-virtio machine type option Zhuoying Cai
2025-06-06 14:00   ` Daniel P. Berrangé
2025-06-20 15:45     ` Zhuoying Cai
2025-06-24 15:03       ` Jared Rossi
2025-06-30 19:31         ` Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 02/28] crypto/x509-utils: Add helper functions for certificate store Zhuoying Cai
2025-06-06 10:03   ` Daniel P. Berrangé
2025-06-06 10:14   ` Daniel P. Berrangé
2025-06-17 10:58   ` Markus Armbruster
2025-06-17 14:57     ` Zhuoying Cai
2025-06-18  5:57       ` Markus Armbruster
2025-06-18 15:34         ` Zhuoying Cai
2025-06-23  6:15           ` Markus Armbruster
2025-06-23  8:00             ` Daniel P. Berrangé
2025-06-23  9:22               ` Markus Armbruster
2025-06-04 21:56 ` [PATCH v3 03/28] hw/s390x/ipl: Create " Zhuoying Cai
2025-06-06 10:31   ` Daniel P. Berrangé [this message]
2025-06-30 19:56     ` Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 04/28] s390x: Guest support for Certificate Store Facility (CS) Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 05/28] s390x/diag: Introduce DIAG 320 for certificate store facility Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 06/28] s390x/diag: Refactor address validation check from diag308_parm_check Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 07/28] s390x/diag: Implement DIAG 320 subcode 1 Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 08/28] crypto/x509-utils: Add helper functions for DIAG 320 subcode 2 Zhuoying Cai
2025-06-06 10:40   ` Daniel P. Berrangé
2025-06-06 10:43   ` Daniel P. Berrangé
2025-06-04 21:56 ` [PATCH v3 09/28] s390x/diag: Implement " Zhuoying Cai
2025-06-06 10:51   ` Daniel P. Berrangé
2025-06-04 21:56 ` [PATCH v3 10/28] s390x/diag: Introduce DIAG 508 for secure IPL operations Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 11/28] crypto: Add helper functions for DIAG 508 subcode 1 Zhuoying Cai
2025-06-06 10:56   ` Daniel P. Berrangé
2025-06-04 21:56 ` [PATCH v3 12/28] s390x/diag: Implement DIAG 508 subcode 1 for signature verification Zhuoying Cai
2025-06-06 10:58   ` Daniel P. Berrangé
2025-06-04 21:56 ` [PATCH v3 13/28] pc-bios/s390-ccw: Introduce IPL Information Report Block (IIRB) Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 14/28] pc-bios/s390-ccw: Define memory for IPLB and convert IPLB to pointers Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 15/28] hw/s390x/ipl: Add IPIB flags to IPL Parameter Block Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 16/28] hw/s390x/ipl: Set iplb->len to maximum length of " Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 17/28] s390x: Guest support for Secure-IPL Facility Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 18/28] pc-bios/s390-ccw: Refactor zipl_run() Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 19/28] pc-bios/s390-ccw: Refactor zipl_load_segment function Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 20/28] pc-bios/s390-ccw: Add signature verification for secure IPL in audit mode Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 21/28] s390x: Guest support for Secure-IPL Code Loading Attributes Facility (SCLAF) Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 22/28] pc-bios/s390-ccw: Add additional security checks for secure boot Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 23/28] Add secure-boot to s390-ccw-virtio machine type option Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 24/28] hw/s390x/ipl: Set IPIB flags for secure IPL Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 25/28] pc-bios/s390-ccw: Handle true secure IPL mode Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 26/28] pc-bios/s390-ccw: Handle secure boot with multiple boot devices Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 27/28] hw/s390x/ipl: Handle secure boot without specifying a boot device Zhuoying Cai
2025-06-04 21:56 ` [PATCH v3 28/28] docs: Add secure IPL documentation Zhuoying Cai
2025-06-06 11:04   ` Daniel P. Berrangé
2025-06-17 21:29   ` Collin Walling

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=aELDf8--oCzcfhFI@redhat.com \
    --to=berrange@redhat.com \
    --cc=armbru@redhat.com \
    --cc=borntraeger@linux.ibm.com \
    --cc=david@redhat.com \
    --cc=eblake@redhat.com \
    --cc=farman@linux.ibm.com \
    --cc=iii@linux.ibm.com \
    --cc=jjherne@linux.ibm.com \
    --cc=jrossi@linux.ibm.com \
    --cc=pasic@linux.ibm.com \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=qemu-s390x@nongnu.org \
    --cc=richard.henderson@linaro.org \
    --cc=thuth@redhat.com \
    --cc=walling@linux.ibm.com \
    --cc=zycai@linux.ibm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).